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 -fexternal-templates @gol
178 -falt-external-templates @gol
179 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
180 -fno-implicit-templates @gol
181 -fno-implicit-inline-templates @gol
182 -fno-implement-inlines -fms-extensions @gol
183 -fno-nonansi-builtins -fno-operator-names @gol
184 -fno-optional-diags -fpermissive @gol
185 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
186 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
187 -fno-default-inline -Wabi -Wctor-dtor-privacy @gol
188 -Wnon-virtual-dtor -Wreorder @gol
189 -Weffc++ -Wno-deprecated @gol
190 -Wno-non-template-friend -Wold-style-cast @gol
191 -Woverloaded-virtual -Wno-pmf-conversions @gol
192 -Wsign-promo -Wsynth}
194 @item Objective-C Language Options
195 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
196 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
197 -fgnu-runtime -fnext-runtime -gen-decls @gol
198 -Wno-protocol -Wselector -Wundeclared-selector}
200 @item Language Independent Options
201 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
202 @gccoptlist{-fmessage-length=@var{n} @gol
203 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
205 @item Warning Options
206 @xref{Warning Options,,Options to Request or Suppress Warnings}.
207 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
208 -w -Wextra -Wall -Waggregate-return @gol
209 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
210 -Wconversion -Wno-deprecated-declarations @gol
211 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
212 -Werror -Werror-implicit-function-declaration @gol
213 -Wfloat-equal -Wformat -Wformat=2 @gol
214 -Wno-format-extra-args -Wformat-nonliteral @gol
215 -Wformat-security -Wno-format-y2k @gol
216 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
217 -Wimport -Wno-import -Winit-self -Winline @gol
218 -Wno-invalid-offsetof -Winvalid-pch @gol
219 -Wlarger-than-@var{len} -Wlong-long @gol
220 -Wmain -Wmissing-braces @gol
221 -Wmissing-format-attribute -Wmissing-noreturn @gol
222 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
223 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
224 -Wreturn-type -Wsequence-point -Wshadow @gol
225 -Wsign-compare -Wstrict-aliasing @gol
226 -Wswitch -Wswitch-default -Wswitch-enum @gol
227 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
228 -Wunknown-pragmas -Wunreachable-code @gol
229 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
230 -Wunused-value -Wunused-variable -Wwrite-strings}
232 @item C-only Warning Options
233 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
234 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
235 -Wstrict-prototypes -Wtraditional @gol
236 -Wdeclaration-after-statement}
238 @item Debugging Options
239 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
240 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
241 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
242 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
243 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
244 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
245 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
246 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
247 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
248 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
249 -ftest-coverage -ftime-report @gol
250 -g -g@var{level} -gcoff -gdwarf -gdwarf-1 -gdwarf-1+ -gdwarf-2 @gol
251 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
252 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
253 -print-multi-directory -print-multi-lib @gol
254 -print-prog-name=@var{program} -print-search-dirs -Q @gol
257 @item Optimization Options
258 @xref{Optimize Options,,Options that Control Optimization}.
259 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
260 -falign-labels=@var{n} -falign-loops=@var{n} @gol
261 -fbranch-probabilities -fprofile-values -fbranch-target-load-optimize @gol
262 -fbranch-target-load-optimize2 -fcaller-saves -fcprop-registers @gol
263 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
264 -fdelayed-branch -fdelete-null-pointer-checks @gol
265 -fexpensive-optimizations -ffast-math -ffloat-store @gol
266 -fforce-addr -fforce-mem -ffunction-sections @gol
267 -fgcse -fgcse-lm -fgcse-sm -floop-optimize -fcrossjumping @gol
268 -fif-conversion -fif-conversion2 @gol
269 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
270 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
271 -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
272 -fno-default-inline -fno-defer-pop @gol
273 -fno-function-cse -fno-guess-branch-probability @gol
274 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
275 -funsafe-math-optimizations -ffinite-math-only @gol
276 -fno-trapping-math -fno-zero-initialized-in-bss @gol
277 -fomit-frame-pointer -foptimize-register-move @gol
278 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
279 -freduce-all-givs -fregmove -frename-registers @gol
280 -freorder-blocks -freorder-functions @gol
281 -frerun-cse-after-loop -frerun-loop-opt @gol
282 -frounding-math -fschedule-insns -fschedule-insns2 @gol
283 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
284 -fsched-spec-load-dangerous -fsched2-use-superblocks @gol
285 -fsched2-use-traces -fsignaling-nans @gol
286 -fsingle-precision-constant -fssa -fssa-ccp -fssa-dce @gol
287 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
288 -funroll-all-loops -funroll-loops -fpeel-loops @gol
289 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
290 --param @var{name}=@var{value}
291 -O -O0 -O1 -O2 -O3 -Os}
293 @item Preprocessor Options
294 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
295 @gccoptlist{-A@var{question}=@var{answer} @gol
296 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
297 -C -dD -dI -dM -dN @gol
298 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
299 -idirafter @var{dir} @gol
300 -include @var{file} -imacros @var{file} @gol
301 -iprefix @var{file} -iwithprefix @var{dir} @gol
302 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
303 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
304 -P -fworking-directory -remap @gol
305 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
306 -Xpreprocessor @var{option}}
308 @item Assembler Option
309 @xref{Assembler Options,,Passing Options to the Assembler}.
310 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
313 @xref{Link Options,,Options for Linking}.
314 @gccoptlist{@var{object-file-name} -l@var{library} @gol
315 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
316 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
317 -Wl,@var{option} -Xlinker @var{option} @gol
320 @item Directory Options
321 @xref{Directory Options,,Options for Directory Search}.
322 @gccoptlist{-B@var{prefix} -I@var{dir} -I- -L@var{dir} -specs=@var{file}}
325 @c I wrote this xref this way to avoid overfull hbox. -- rms
326 @xref{Target Options}.
327 @gccoptlist{-V @var{version} -b @var{machine}}
329 @item Machine Dependent Options
330 @xref{Submodel Options,,Hardware Models and Configurations}.
332 @emph{M680x0 Options}
333 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
334 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
335 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
336 -malign-int -mstrict-align}
338 @emph{M68hc1x Options}
339 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
340 -mauto-incdec -minmax -mlong-calls -mshort @gol
341 -msoft-reg-count=@var{count}}
344 @gccoptlist{-mg -mgnu -munix}
347 @gccoptlist{-mcpu=@var{cpu-type} @gol
348 -mtune=@var{cpu-type} @gol
349 -mcmodel=@var{code-model} @gol
351 -mapp-regs -mbroken-saverestore -mcypress @gol
352 -mfaster-structs -mflat @gol
353 -mfpu -mhard-float -mhard-quad-float @gol
354 -mimpure-text -mlittle-endian -mlive-g0 -mno-app-regs @gol
355 -mno-faster-structs -mno-flat -mno-fpu @gol
356 -mno-impure-text -mno-stack-bias -mno-unaligned-doubles @gol
357 -msoft-float -msoft-quad-float -msparclite -mstack-bias @gol
358 -msupersparc -munaligned-doubles -mv8}
361 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
362 -mapcs-26 -mapcs-32 @gol
363 -mapcs-stack-check -mno-apcs-stack-check @gol
364 -mapcs-float -mno-apcs-float @gol
365 -mapcs-reentrant -mno-apcs-reentrant @gol
366 -msched-prolog -mno-sched-prolog @gol
367 -mlittle-endian -mbig-endian -mwords-little-endian @gol
368 -malignment-traps -mno-alignment-traps @gol
369 -msoft-float -mhard-float -mfpe @gol
370 -mthumb-interwork -mno-thumb-interwork @gol
371 -mcpu=@var{name} -march=@var{name} -mfpe=@var{name} @gol
372 -mstructure-size-boundary=@var{n} @gol
373 -mabort-on-noreturn @gol
374 -mlong-calls -mno-long-calls @gol
375 -msingle-pic-base -mno-single-pic-base @gol
376 -mpic-register=@var{reg} @gol
377 -mnop-fun-dllimport @gol
378 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
379 -mpoke-function-name @gol
381 -mtpcs-frame -mtpcs-leaf-frame @gol
382 -mcaller-super-interworking -mcallee-super-interworking}
384 @emph{MN10200 Options}
387 @emph{MN10300 Options}
388 @gccoptlist{-mmult-bug -mno-mult-bug @gol
389 -mam33 -mno-am33 @gol
390 -mam33-2 -mno-am33-2 @gol
393 @emph{M32R/D Options}
394 @gccoptlist{-m32rx -m32r -mcode-model=@var{model-type} @gol
395 -msdata=@var{sdata-type} -G @var{num}}
398 @gccoptlist{-m88000 -m88100 -m88110 -mbig-pic @gol
399 -mcheck-zero-division -mhandle-large-shift @gol
400 -midentify-revision -mno-check-zero-division @gol
401 -mno-ocs-debug-info -mno-ocs-frame-position @gol
402 -mno-optimize-arg-area -mno-serialize-volatile @gol
403 -mno-underscores -mocs-debug-info @gol
404 -mocs-frame-position -moptimize-arg-area @gol
405 -mserialize-volatile -mshort-data-@var{num} -msvr3 @gol
406 -msvr4 -mtrap-large-shift -muse-div-instruction @gol
407 -mversion-03.00 -mwarn-passed-structs}
409 @emph{RS/6000 and PowerPC Options}
410 @gccoptlist{-mcpu=@var{cpu-type} @gol
411 -mtune=@var{cpu-type} @gol
412 -mpower -mno-power -mpower2 -mno-power2 @gol
413 -mpowerpc -mpowerpc64 -mno-powerpc @gol
414 -maltivec -mno-altivec @gol
415 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
416 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
417 -mnew-mnemonics -mold-mnemonics @gol
418 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
419 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
420 -malign-power -malign-natural @gol
421 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
422 -mstring -mno-string -mupdate -mno-update @gol
423 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
424 -mstrict-align -mno-strict-align -mrelocatable @gol
425 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
426 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
427 -mdynamic-no-pic @gol
428 -mcall-sysv -mcall-netbsd @gol
429 -maix-struct-return -msvr4-struct-return @gol
430 -mabi=altivec -mabi=no-altivec @gol
431 -mabi=spe -mabi=no-spe @gol
432 -misel=yes -misel=no @gol
433 -mspe=yes -mspe=no @gol
434 -mfloat-gprs=yes -mfloat-gprs=no @gol
435 -mprototype -mno-prototype @gol
436 -msim -mmvme -mads -myellowknife -memb -msdata @gol
437 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
439 @emph{Darwin Options}
440 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
441 -arch_only -bind_at_load -bundle -bundle_loader @gol
442 -client_name -compatibility_version -current_version @gol
443 -dependency-file -dylib_file -dylinker_install_name @gol
444 -dynamic -dynamiclib -exported_symbols_list @gol
445 -filelist -flat_namespace -force_cpusubtype_ALL @gol
446 -force_flat_namespace -headerpad_max_install_names @gol
447 -image_base -init -install_name -keep_private_externs @gol
448 -multi_module -multiply_defined -multiply_defined_unused @gol
449 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
450 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
451 -private_bundle -read_only_relocs -sectalign @gol
452 -sectobjectsymbols -whyload -seg1addr @gol
453 -sectcreate -sectobjectsymbols -sectorder @gol
454 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
455 -segprot -segs_read_only_addr -segs_read_write_addr @gol
456 -single_module -static -sub_library -sub_umbrella @gol
457 -twolevel_namespace -umbrella -undefined @gol
458 -unexported_symbols_list -weak_reference_mismatches @gol
462 @gccoptlist{-mcall-lib-mul -mfp-arg-in-fpregs -mfp-arg-in-gregs @gol
463 -mfull-fp-blocks -mhc-struct-return -min-line-mul @gol
464 -mminimum-fp-blocks -mnohc-struct-return}
467 @gccoptlist{-mabicalls -march=@var{cpu-type} -mtune=@var{cpu=type} @gol
468 -mcpu=@var{cpu-type} -membedded-data -muninit-const-in-rodata @gol
469 -membedded-pic -mfp32 -mfp64 -mfused-madd -mno-fused-madd @gol
470 -mgas -mgp32 -mgp64 -mhard-float -mint64 -mips1 @gol
471 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
472 -mlong64 -mlong32 -mlong-calls -mmemcpy @gol
473 -mmips-as -mmips-tfile -mno-abicalls -mxgot @gol
474 -mno-embedded-data -mno-uninit-const-in-rodata @gol
475 -mno-embedded-pic -mno-long-calls @gol
476 -mno-memcpy -mno-mips-tfile -mno-rnames @gol
477 -mrnames -msoft-float @gol
478 -m4650 -msingle-float -mmad @gol
479 -EL -EB -G @var{num} -nocpp @gol
480 -mabi=32 -mabi=n32 -mabi=64 -mabi=eabi -mabi-fake-default @gol
481 -mfix7000 -mfix-sb1 -mno-fix-sb1 @gol
482 -mno-crt0 -mflush-func=@var{func} -mno-flush-func @gol
483 -mbranch-likely -mno-branch-likely}
485 @emph{i386 and x86-64 Options}
486 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
487 -mfpmath=@var{unit} @gol
488 -masm=@var{dialect} -mno-fancy-math-387 @gol
489 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
490 -mno-wide-multiply -mrtd -malign-double @gol
491 -mpreferred-stack-boundary=@var{num} @gol
492 -mmmx -msse -msse2 -mpni -m3dnow @gol
493 -mthreads -mno-align-stringops -minline-all-stringops @gol
494 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
495 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
496 -mno-red-zone -mno-tls-direct-seg-refs @gol
497 -mcmodel=@var{code-model} @gol
501 @gccoptlist{-march=@var{architecture-type} @gol
502 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
503 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
504 -mjump-in-delay -mlinker-opt -mlong-calls @gol
505 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
506 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
507 -mno-jump-in-delay -mno-long-load-store @gol
508 -mno-portable-runtime -mno-soft-float @gol
509 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
510 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
511 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
512 -nolibdld -static -threads}
514 @emph{Intel 960 Options}
515 @gccoptlist{-m@var{cpu-type} -masm-compat -mclean-linkage @gol
516 -mcode-align -mcomplex-addr -mleaf-procedures @gol
517 -mic-compat -mic2.0-compat -mic3.0-compat @gol
518 -mintel-asm -mno-clean-linkage -mno-code-align @gol
519 -mno-complex-addr -mno-leaf-procedures @gol
520 -mno-old-align -mno-strict-align -mno-tail-call @gol
521 -mnumerics -mold-align -msoft-float -mstrict-align @gol
524 @emph{DEC Alpha Options}
525 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
526 -mieee -mieee-with-inexact -mieee-conformant @gol
527 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
528 -mtrap-precision=@var{mode} -mbuild-constants @gol
529 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
530 -mbwx -mmax -mfix -mcix @gol
531 -mfloat-vax -mfloat-ieee @gol
532 -mexplicit-relocs -msmall-data -mlarge-data @gol
533 -msmall-text -mlarge-text @gol
534 -mmemory-latency=@var{time}}
536 @emph{DEC Alpha/VMS Options}
537 @gccoptlist{-mvms-return-codes}
539 @emph{H8/300 Options}
540 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
543 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
544 -m4-nofpu -m4-single-only -m4-single -m4 @gol
545 -m5-64media -m5-64media-nofpu @gol
546 -m5-32media -m5-32media-nofpu @gol
547 -m5-compact -m5-compact-nofpu @gol
548 -mb -ml -mdalign -mrelax @gol
549 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
550 -mieee -misize -mpadstruct -mspace @gol
551 -mprefergot -musermode}
553 @emph{System V Options}
554 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
557 @gccoptlist{-EB -EL @gol
558 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
559 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
561 @emph{TMS320C3x/C4x Options}
562 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
563 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
564 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
565 -mparallel-insns -mparallel-mpy -mpreserve-float}
568 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
569 -mprolog-function -mno-prolog-function -mspace @gol
570 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
571 -mapp-regs -mno-app-regs @gol
572 -mdisable-callt -mno-disable-callt @gol
578 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
579 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
580 -mregparam -mnoregparam -msb -mnosb @gol
581 -mbitfield -mnobitfield -mhimem -mnohimem}
584 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
585 -mcall-prologues -mno-tablejump -mtiny-stack}
588 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
589 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
590 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
591 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
592 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
595 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
596 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
597 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
598 -mno-base-addresses -msingle-exit -mno-single-exit}
601 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
602 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
603 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
604 -minline-float-divide-max-throughput @gol
605 -minline-int-divide-min-latency @gol
606 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
607 -mfixed-range=@var{register-range}}
610 @gccoptlist{-mextmem -mextmemory -monchip -mno-asm-optimize @gol
611 -masm-optimize -mbranch-cost=@var{n} -mcond-exec=@var{n}}
613 @emph{S/390 and zSeries Options}
614 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
615 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
616 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
617 -m64 -m31 -mdebug -mno-debug -mesa -mzarch}
620 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
621 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
622 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
623 -mstack-align -mdata-align -mconst-align @gol
624 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
625 -melf -maout -melinux -mlinux -sim -sim2}
627 @emph{PDP-11 Options}
628 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
629 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
630 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
631 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
632 -mbranch-expensive -mbranch-cheap @gol
633 -msplit -mno-split -munix-asm -mdec-asm}
635 @emph{Xstormy16 Options}
638 @emph{Xtensa Options}
639 @gccoptlist{-mconst16 -mno-const16 @gol
640 -mfused-madd -mno-fused-madd @gol
641 -mtext-section-literals -mno-text-section-literals @gol
642 -mtarget-align -mno-target-align @gol
643 -mlongcalls -mno-longcalls}
646 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
647 -mhard-float -msoft-float @gol
648 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
649 -mdouble -mno-double @gol
650 -mmedia -mno-media -mmuladd -mno-muladd @gol
651 -mlibrary-pic -macc-4 -macc-8 @gol
652 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
653 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
654 -mvliw-branch -mno-vliw-branch @gol
655 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
656 -mno-nested-cond-exec -mtomcat-stats @gol
659 @item Code Generation Options
660 @xref{Code Gen Options,,Options for Code Generation Conventions}.
661 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
662 -ffixed-@var{reg} -fexceptions @gol
663 -fnon-call-exceptions -funwind-tables @gol
664 -fasynchronous-unwind-tables @gol
665 -finhibit-size-directive -finstrument-functions @gol
666 -fno-common -fno-ident -fno-gnu-linker @gol
667 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
668 -freg-struct-return -fshared-data -fshort-enums @gol
669 -fshort-double -fshort-wchar @gol
670 -fverbose-asm -fpack-struct -fstack-check @gol
671 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
672 -fargument-alias -fargument-noalias @gol
673 -fargument-noalias-global -fleading-underscore @gol
674 -ftls-model=@var{model} @gol
675 -ftrapv -fwrapv -fbounds-check}
679 * Overall Options:: Controlling the kind of output:
680 an executable, object files, assembler files,
681 or preprocessed source.
682 * C Dialect Options:: Controlling the variant of C language compiled.
683 * C++ Dialect Options:: Variations on C++.
684 * Objective-C Dialect Options:: Variations on Objective-C.
685 * Language Independent Options:: Controlling how diagnostics should be
687 * Warning Options:: How picky should the compiler be?
688 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
689 * Optimize Options:: How much optimization?
690 * Preprocessor Options:: Controlling header files and macro definitions.
691 Also, getting dependency information for Make.
692 * Assembler Options:: Passing options to the assembler.
693 * Link Options:: Specifying libraries and so on.
694 * Directory Options:: Where to find header files and libraries.
695 Where to find the compiler executable files.
696 * Spec Files:: How to pass switches to sub-processes.
697 * Target Options:: Running a cross-compiler, or an old version of GCC.
700 @node Overall Options
701 @section Options Controlling the Kind of Output
703 Compilation can involve up to four stages: preprocessing, compilation
704 proper, assembly and linking, always in that order. GCC is capable of
705 preprocessing and compiling several files either into several
706 assembler input files, or into one assembler input file; then each
707 assembler input file produces an object file, and linking combines all
708 the object files (those newly compiled, and those specified as input)
709 into an executable file.
711 @cindex file name suffix
712 For any given input file, the file name suffix determines what kind of
717 C source code which must be preprocessed.
720 C source code which should not be preprocessed.
723 C++ source code which should not be preprocessed.
726 Objective-C source code. Note that you must link with the library
727 @file{libobjc.a} to make an Objective-C program work.
730 Objective-C source code which should not be preprocessed.
733 C or C++ header file to be turned into a precompiled header.
737 @itemx @var{file}.cxx
738 @itemx @var{file}.cpp
739 @itemx @var{file}.CPP
740 @itemx @var{file}.c++
742 C++ source code which must be preprocessed. Note that in @samp{.cxx},
743 the last two letters must both be literally @samp{x}. Likewise,
744 @samp{.C} refers to a literal capital C@.
748 C++ header file to be turned into a precompiled header.
751 @itemx @var{file}.for
752 @itemx @var{file}.FOR
753 Fortran source code which should not be preprocessed.
756 @itemx @var{file}.fpp
757 @itemx @var{file}.FPP
758 Fortran source code which must be preprocessed (with the traditional
762 Fortran source code which must be preprocessed with a RATFOR
763 preprocessor (not included with GCC)@.
765 @xref{Overall Options,,Options Controlling the Kind of Output, g77,
766 Using and Porting GNU Fortran}, for more details of the handling of
769 @c FIXME: Descriptions of Java file types.
776 Ada source code file which contains a library unit declaration (a
777 declaration of a package, subprogram, or generic, or a generic
778 instantiation), or a library unit renaming declaration (a package,
779 generic, or subprogram renaming declaration). Such files are also
782 @itemx @var{file}.adb
783 Ada source code file containing a library unit body (a subprogram or
784 package body). Such files are also called @dfn{bodies}.
786 @c GCC also knows about some suffixes for languages not yet included:
795 Assembler code which must be preprocessed.
798 An object file to be fed straight into linking.
799 Any file name with no recognized suffix is treated this way.
803 You can specify the input language explicitly with the @option{-x} option:
806 @item -x @var{language}
807 Specify explicitly the @var{language} for the following input files
808 (rather than letting the compiler choose a default based on the file
809 name suffix). This option applies to all following input files until
810 the next @option{-x} option. Possible values for @var{language} are:
812 c c-header cpp-output
813 c++ c++-header c++-cpp-output
814 objective-c objective-c-header objc-cpp-output
815 assembler assembler-with-cpp
817 f77 f77-cpp-input ratfor
823 Turn off any specification of a language, so that subsequent files are
824 handled according to their file name suffixes (as they are if @option{-x}
825 has not been used at all).
827 @item -pass-exit-codes
828 @opindex pass-exit-codes
829 Normally the @command{gcc} program will exit with the code of 1 if any
830 phase of the compiler returns a non-success return code. If you specify
831 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
832 numerically highest error produced by any phase that returned an error
836 If you only want some of the stages of compilation, you can use
837 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
838 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
839 @command{gcc} is to stop. Note that some combinations (for example,
840 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
845 Compile or assemble the source files, but do not link. The linking
846 stage simply is not done. The ultimate output is in the form of an
847 object file for each source file.
849 By default, the object file name for a source file is made by replacing
850 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
852 Unrecognized input files, not requiring compilation or assembly, are
857 Stop after the stage of compilation proper; do not assemble. The output
858 is in the form of an assembler code file for each non-assembler input
861 By default, the assembler file name for a source file is made by
862 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
864 Input files that don't require compilation are ignored.
868 Stop after the preprocessing stage; do not run the compiler proper. The
869 output is in the form of preprocessed source code, which is sent to the
872 Input files which don't require preprocessing are ignored.
874 @cindex output file option
877 Place output in file @var{file}. This applies regardless to whatever
878 sort of output is being produced, whether it be an executable file,
879 an object file, an assembler file or preprocessed C code.
881 If you specify @option{-o} when compiling more than one input file, or
882 you are producing an executable file as output, all the source files
883 on the command line will be compiled at once.
885 If @option{-o} is not specified, the default is to put an executable file
886 in @file{a.out}, the object file for @file{@var{source}.@var{suffix}} in
887 @file{@var{source}.o}, its assembler file in @file{@var{source}.s}, and
888 all preprocessed C source on standard output.
892 Print (on standard error output) the commands executed to run the stages
893 of compilation. Also print the version number of the compiler driver
894 program and of the preprocessor and the compiler proper.
898 Like @option{-v} except the commands are not executed and all command
899 arguments are quoted. This is useful for shell scripts to capture the
900 driver-generated command lines.
904 Use pipes rather than temporary files for communication between the
905 various stages of compilation. This fails to work on some systems where
906 the assembler is unable to read from a pipe; but the GNU assembler has
911 Print (on the standard output) a description of the command line options
912 understood by @command{gcc}. If the @option{-v} option is also specified
913 then @option{--help} will also be passed on to the various processes
914 invoked by @command{gcc}, so that they can display the command line options
915 they accept. If the @option{-Wextra} option is also specified then command
916 line options which have no documentation associated with them will also
921 Print (on the standard output) a description of target specific command
922 line options for each tool.
926 Display the version number and copyrights of the invoked GCC.
930 @section Compiling C++ Programs
932 @cindex suffixes for C++ source
933 @cindex C++ source file suffixes
934 C++ source files conventionally use one of the suffixes @samp{.C},
935 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
936 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
937 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
938 files with these names and compiles them as C++ programs even if you
939 call the compiler the same way as for compiling C programs (usually
940 with the name @command{gcc}).
944 However, C++ programs often require class libraries as well as a
945 compiler that understands the C++ language---and under some
946 circumstances, you might want to compile programs or header files from
947 standard input, or otherwise without a suffix that flags them as C++
948 programs. You might also like to precompile a C header file with a
949 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
950 program that calls GCC with the default language set to C++, and
951 automatically specifies linking against the C++ library. On many
952 systems, @command{g++} is also installed with the name @command{c++}.
954 @cindex invoking @command{g++}
955 When you compile C++ programs, you may specify many of the same
956 command-line options that you use for compiling programs in any
957 language; or command-line options meaningful for C and related
958 languages; or options that are meaningful only for C++ programs.
959 @xref{C Dialect Options,,Options Controlling C Dialect}, for
960 explanations of options for languages related to C@.
961 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
962 explanations of options that are meaningful only for C++ programs.
964 @node C Dialect Options
965 @section Options Controlling C Dialect
966 @cindex dialect options
967 @cindex language dialect options
968 @cindex options, dialect
970 The following options control the dialect of C (or languages derived
971 from C, such as C++ and Objective-C) that the compiler accepts:
978 In C mode, support all ISO C90 programs. In C++ mode,
979 remove GNU extensions that conflict with ISO C++.
981 This turns off certain features of GCC that are incompatible with ISO
982 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
983 such as the @code{asm} and @code{typeof} keywords, and
984 predefined macros such as @code{unix} and @code{vax} that identify the
985 type of system you are using. It also enables the undesirable and
986 rarely used ISO trigraph feature. For the C compiler,
987 it disables recognition of C++ style @samp{//} comments as well as
988 the @code{inline} keyword.
990 The alternate keywords @code{__asm__}, @code{__extension__},
991 @code{__inline__} and @code{__typeof__} continue to work despite
992 @option{-ansi}. You would not want to use them in an ISO C program, of
993 course, but it is useful to put them in header files that might be included
994 in compilations done with @option{-ansi}. Alternate predefined macros
995 such as @code{__unix__} and @code{__vax__} are also available, with or
996 without @option{-ansi}.
998 The @option{-ansi} option does not cause non-ISO programs to be
999 rejected gratuitously. For that, @option{-pedantic} is required in
1000 addition to @option{-ansi}. @xref{Warning Options}.
1002 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1003 option is used. Some header files may notice this macro and refrain
1004 from declaring certain functions or defining certain macros that the
1005 ISO standard doesn't call for; this is to avoid interfering with any
1006 programs that might use these names for other things.
1008 Functions which would normally be built in but do not have semantics
1009 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1010 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1011 built-in functions provided by GCC}, for details of the functions
1016 Determine the language standard. This option is currently only
1017 supported when compiling C or C++. A value for this option must be
1018 provided; possible values are
1023 ISO C90 (same as @option{-ansi}).
1025 @item iso9899:199409
1026 ISO C90 as modified in amendment 1.
1032 ISO C99. Note that this standard is not yet fully supported; see
1033 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1034 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1037 Default, ISO C90 plus GNU extensions (including some C99 features).
1041 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1042 this will become the default. The name @samp{gnu9x} is deprecated.
1045 The 1998 ISO C++ standard plus amendments.
1048 The same as @option{-std=c++98} plus GNU extensions. This is the
1049 default for C++ code.
1052 Even when this option is not specified, you can still use some of the
1053 features of newer standards in so far as they do not conflict with
1054 previous C standards. For example, you may use @code{__restrict__} even
1055 when @option{-std=c99} is not specified.
1057 The @option{-std} options specifying some version of ISO C have the same
1058 effects as @option{-ansi}, except that features that were not in ISO C90
1059 but are in the specified version (for example, @samp{//} comments and
1060 the @code{inline} keyword in ISO C99) are not disabled.
1062 @xref{Standards,,Language Standards Supported by GCC}, for details of
1063 these standard versions.
1065 @item -aux-info @var{filename}
1067 Output to the given filename prototyped declarations for all functions
1068 declared and/or defined in a translation unit, including those in header
1069 files. This option is silently ignored in any language other than C@.
1071 Besides declarations, the file indicates, in comments, the origin of
1072 each declaration (source file and line), whether the declaration was
1073 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1074 @samp{O} for old, respectively, in the first character after the line
1075 number and the colon), and whether it came from a declaration or a
1076 definition (@samp{C} or @samp{F}, respectively, in the following
1077 character). In the case of function definitions, a K&R-style list of
1078 arguments followed by their declarations is also provided, inside
1079 comments, after the declaration.
1083 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1084 keyword, so that code can use these words as identifiers. You can use
1085 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1086 instead. @option{-ansi} implies @option{-fno-asm}.
1088 In C++, this switch only affects the @code{typeof} keyword, since
1089 @code{asm} and @code{inline} are standard keywords. You may want to
1090 use the @option{-fno-gnu-keywords} flag instead, which has the same
1091 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1092 switch only affects the @code{asm} and @code{typeof} keywords, since
1093 @code{inline} is a standard keyword in ISO C99.
1096 @itemx -fno-builtin-@var{function}
1097 @opindex fno-builtin
1098 @cindex built-in functions
1099 Don't recognize built-in functions that do not begin with
1100 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1101 functions provided by GCC}, for details of the functions affected,
1102 including those which are not built-in functions when @option{-ansi} or
1103 @option{-std} options for strict ISO C conformance are used because they
1104 do not have an ISO standard meaning.
1106 GCC normally generates special code to handle certain built-in functions
1107 more efficiently; for instance, calls to @code{alloca} may become single
1108 instructions that adjust the stack directly, and calls to @code{memcpy}
1109 may become inline copy loops. The resulting code is often both smaller
1110 and faster, but since the function calls no longer appear as such, you
1111 cannot set a breakpoint on those calls, nor can you change the behavior
1112 of the functions by linking with a different library.
1114 With the @option{-fno-builtin-@var{function}} option
1115 only the built-in function @var{function} is
1116 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1117 function is named this is not built-in in this version of GCC, this
1118 option is ignored. There is no corresponding
1119 @option{-fbuiltin-@var{function}} option; if you wish to enable
1120 built-in functions selectively when using @option{-fno-builtin} or
1121 @option{-ffreestanding}, you may define macros such as:
1124 #define abs(n) __builtin_abs ((n))
1125 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1130 @cindex hosted environment
1132 Assert that compilation takes place in a hosted environment. This implies
1133 @option{-fbuiltin}. A hosted environment is one in which the
1134 entire standard library is available, and in which @code{main} has a return
1135 type of @code{int}. Examples are nearly everything except a kernel.
1136 This is equivalent to @option{-fno-freestanding}.
1138 @item -ffreestanding
1139 @opindex ffreestanding
1140 @cindex hosted environment
1142 Assert that compilation takes place in a freestanding environment. This
1143 implies @option{-fno-builtin}. A freestanding environment
1144 is one in which the standard library may not exist, and program startup may
1145 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1146 This is equivalent to @option{-fno-hosted}.
1148 @xref{Standards,,Language Standards Supported by GCC}, for details of
1149 freestanding and hosted environments.
1151 @item -fms-extensions
1152 @opindex fms-extensions
1153 Accept some non-standard constructs used in Microsoft header files.
1157 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1158 options for strict ISO C conformance) implies @option{-trigraphs}.
1160 @item -no-integrated-cpp
1161 @opindex no-integrated-cpp
1162 Performs a compilation in two passes: preprocessing and compiling. This
1163 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1164 @option{-B} option. The user supplied compilation step can then add in
1165 an additional preprocessing step after normal preprocessing but before
1166 compiling. The default is to use the integrated cpp (internal cpp)
1168 The semantics of this option will change if "cc1", "cc1plus", and
1169 "cc1obj" are merged.
1171 @cindex traditional C language
1172 @cindex C language, traditional
1174 @itemx -traditional-cpp
1175 @opindex traditional-cpp
1176 @opindex traditional
1177 Formerly, these options caused GCC to attempt to emulate a pre-standard
1178 C compiler. They are now only supported with the @option{-E} switch.
1179 The preprocessor continues to support a pre-standard mode. See the GNU
1180 CPP manual for details.
1182 @item -fcond-mismatch
1183 @opindex fcond-mismatch
1184 Allow conditional expressions with mismatched types in the second and
1185 third arguments. The value of such an expression is void. This option
1186 is not supported for C++.
1188 @item -funsigned-char
1189 @opindex funsigned-char
1190 Let the type @code{char} be unsigned, like @code{unsigned char}.
1192 Each kind of machine has a default for what @code{char} should
1193 be. It is either like @code{unsigned char} by default or like
1194 @code{signed char} by default.
1196 Ideally, a portable program should always use @code{signed char} or
1197 @code{unsigned char} when it depends on the signedness of an object.
1198 But many programs have been written to use plain @code{char} and
1199 expect it to be signed, or expect it to be unsigned, depending on the
1200 machines they were written for. This option, and its inverse, let you
1201 make such a program work with the opposite default.
1203 The type @code{char} is always a distinct type from each of
1204 @code{signed char} or @code{unsigned char}, even though its behavior
1205 is always just like one of those two.
1208 @opindex fsigned-char
1209 Let the type @code{char} be signed, like @code{signed char}.
1211 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1212 the negative form of @option{-funsigned-char}. Likewise, the option
1213 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1215 @item -fsigned-bitfields
1216 @itemx -funsigned-bitfields
1217 @itemx -fno-signed-bitfields
1218 @itemx -fno-unsigned-bitfields
1219 @opindex fsigned-bitfields
1220 @opindex funsigned-bitfields
1221 @opindex fno-signed-bitfields
1222 @opindex fno-unsigned-bitfields
1223 These options control whether a bit-field is signed or unsigned, when the
1224 declaration does not use either @code{signed} or @code{unsigned}. By
1225 default, such a bit-field is signed, because this is consistent: the
1226 basic integer types such as @code{int} are signed types.
1228 @item -fwritable-strings
1229 @opindex fwritable-strings
1230 Store string constants in the writable data segment and don't uniquize
1231 them. This is for compatibility with old programs which assume they can
1232 write into string constants.
1234 Writing into string constants is a very bad idea; ``constants'' should
1238 @node C++ Dialect Options
1239 @section Options Controlling C++ Dialect
1241 @cindex compiler options, C++
1242 @cindex C++ options, command line
1243 @cindex options, C++
1244 This section describes the command-line options that are only meaningful
1245 for C++ programs; but you can also use most of the GNU compiler options
1246 regardless of what language your program is in. For example, you
1247 might compile a file @code{firstClass.C} like this:
1250 g++ -g -frepo -O -c firstClass.C
1254 In this example, only @option{-frepo} is an option meant
1255 only for C++ programs; you can use the other options with any
1256 language supported by GCC@.
1258 Here is a list of options that are @emph{only} for compiling C++ programs:
1262 @item -fabi-version=@var{n}
1263 @opindex fabi-version
1264 Use version @var{n} of the C++ ABI. Version 1 is the version of the C++
1265 ABI that first appeared in G++ 3.2. Version 0 will always be the
1266 version that conforms most closely to the C++ ABI specification.
1267 Therefore, the ABI obtained using version 0 will change as ABI bugs are
1270 The default is version 1.
1272 @item -fno-access-control
1273 @opindex fno-access-control
1274 Turn off all access checking. This switch is mainly useful for working
1275 around bugs in the access control code.
1279 Check that the pointer returned by @code{operator new} is non-null
1280 before attempting to modify the storage allocated. This check is
1281 normally unnecessary because the C++ standard specifies that
1282 @code{operator new} will only return @code{0} if it is declared
1283 @samp{throw()}, in which case the compiler will always check the
1284 return value even without this option. In all other cases, when
1285 @code{operator new} has a non-empty exception specification, memory
1286 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1287 @samp{new (nothrow)}.
1289 @item -fconserve-space
1290 @opindex fconserve-space
1291 Put uninitialized or runtime-initialized global variables into the
1292 common segment, as C does. This saves space in the executable at the
1293 cost of not diagnosing duplicate definitions. If you compile with this
1294 flag and your program mysteriously crashes after @code{main()} has
1295 completed, you may have an object that is being destroyed twice because
1296 two definitions were merged.
1298 This option is no longer useful on most targets, now that support has
1299 been added for putting variables into BSS without making them common.
1301 @item -fno-const-strings
1302 @opindex fno-const-strings
1303 Give string constants type @code{char *} instead of type @code{const
1304 char *}. By default, G++ uses type @code{const char *} as required by
1305 the standard. Even if you use @option{-fno-const-strings}, you cannot
1306 actually modify the value of a string constant, unless you also use
1307 @option{-fwritable-strings}.
1309 This option might be removed in a future release of G++. For maximum
1310 portability, you should structure your code so that it works with
1311 string constants that have type @code{const char *}.
1313 @item -fno-elide-constructors
1314 @opindex fno-elide-constructors
1315 The C++ standard allows an implementation to omit creating a temporary
1316 which is only used to initialize another object of the same type.
1317 Specifying this option disables that optimization, and forces G++ to
1318 call the copy constructor in all cases.
1320 @item -fno-enforce-eh-specs
1321 @opindex fno-enforce-eh-specs
1322 Don't check for violation of exception specifications at runtime. This
1323 option violates the C++ standard, but may be useful for reducing code
1324 size in production builds, much like defining @samp{NDEBUG}. The compiler
1325 will still optimize based on the exception specifications.
1327 @item -fexternal-templates
1328 @opindex fexternal-templates
1330 Cause @samp{#pragma interface} and @samp{implementation} to apply to
1331 template instantiation; template instances are emitted or not according
1332 to the location of the template definition. @xref{Template
1333 Instantiation}, for more information.
1335 This option is deprecated.
1337 @item -falt-external-templates
1338 @opindex falt-external-templates
1339 Similar to @option{-fexternal-templates}, but template instances are
1340 emitted or not according to the place where they are first instantiated.
1341 @xref{Template Instantiation}, for more information.
1343 This option is deprecated.
1346 @itemx -fno-for-scope
1348 @opindex fno-for-scope
1349 If @option{-ffor-scope} is specified, the scope of variables declared in
1350 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1351 as specified by the C++ standard.
1352 If @option{-fno-for-scope} is specified, the scope of variables declared in
1353 a @i{for-init-statement} extends to the end of the enclosing scope,
1354 as was the case in old versions of G++, and other (traditional)
1355 implementations of C++.
1357 The default if neither flag is given to follow the standard,
1358 but to allow and give a warning for old-style code that would
1359 otherwise be invalid, or have different behavior.
1361 @item -fno-gnu-keywords
1362 @opindex fno-gnu-keywords
1363 Do not recognize @code{typeof} as a keyword, so that code can use this
1364 word as an identifier. You can use the keyword @code{__typeof__} instead.
1365 @option{-ansi} implies @option{-fno-gnu-keywords}.
1367 @item -fno-implicit-templates
1368 @opindex fno-implicit-templates
1369 Never emit code for non-inline templates which are instantiated
1370 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1371 @xref{Template Instantiation}, for more information.
1373 @item -fno-implicit-inline-templates
1374 @opindex fno-implicit-inline-templates
1375 Don't emit code for implicit instantiations of inline templates, either.
1376 The default is to handle inlines differently so that compiles with and
1377 without optimization will need the same set of explicit instantiations.
1379 @item -fno-implement-inlines
1380 @opindex fno-implement-inlines
1381 To save space, do not emit out-of-line copies of inline functions
1382 controlled by @samp{#pragma implementation}. This will cause linker
1383 errors if these functions are not inlined everywhere they are called.
1385 @item -fms-extensions
1386 @opindex fms-extensions
1387 Disable pedantic warnings about constructs used in MFC, such as implicit
1388 int and getting a pointer to member function via non-standard syntax.
1390 @item -fno-nonansi-builtins
1391 @opindex fno-nonansi-builtins
1392 Disable built-in declarations of functions that are not mandated by
1393 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1394 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1396 @item -fno-operator-names
1397 @opindex fno-operator-names
1398 Do not treat the operator name keywords @code{and}, @code{bitand},
1399 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1400 synonyms as keywords.
1402 @item -fno-optional-diags
1403 @opindex fno-optional-diags
1404 Disable diagnostics that the standard says a compiler does not need to
1405 issue. Currently, the only such diagnostic issued by G++ is the one for
1406 a name having multiple meanings within a class.
1409 @opindex fpermissive
1410 Downgrade some diagnostics about nonconformant code from errors to
1411 warnings. Thus, using @option{-fpermissive} will allow some
1412 nonconforming code to compile.
1416 Enable automatic template instantiation at link time. This option also
1417 implies @option{-fno-implicit-templates}. @xref{Template
1418 Instantiation}, for more information.
1422 Disable generation of information about every class with virtual
1423 functions for use by the C++ runtime type identification features
1424 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1425 of the language, you can save some space by using this flag. Note that
1426 exception handling uses the same information, but it will generate it as
1431 Emit statistics about front-end processing at the end of the compilation.
1432 This information is generally only useful to the G++ development team.
1434 @item -ftemplate-depth-@var{n}
1435 @opindex ftemplate-depth
1436 Set the maximum instantiation depth for template classes to @var{n}.
1437 A limit on the template instantiation depth is needed to detect
1438 endless recursions during template class instantiation. ANSI/ISO C++
1439 conforming programs must not rely on a maximum depth greater than 17.
1441 @item -fuse-cxa-atexit
1442 @opindex fuse-cxa-atexit
1443 Register destructors for objects with static storage duration with the
1444 @code{__cxa_atexit} function rather than the @code{atexit} function.
1445 This option is required for fully standards-compliant handling of static
1446 destructors, but will only work if your C library supports
1447 @code{__cxa_atexit}.
1451 Do not use weak symbol support, even if it is provided by the linker.
1452 By default, G++ will use weak symbols if they are available. This
1453 option exists only for testing, and should not be used by end-users;
1454 it will result in inferior code and has no benefits. This option may
1455 be removed in a future release of G++.
1459 Do not search for header files in the standard directories specific to
1460 C++, but do still search the other standard directories. (This option
1461 is used when building the C++ library.)
1464 In addition, these optimization, warning, and code generation options
1465 have meanings only for C++ programs:
1468 @item -fno-default-inline
1469 @opindex fno-default-inline
1470 Do not assume @samp{inline} for functions defined inside a class scope.
1471 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1472 functions will have linkage like inline functions; they just won't be
1475 @item -Wabi @r{(C++ only)}
1477 Warn when G++ generates code that is probably not compatible with the
1478 vendor-neutral C++ ABI. Although an effort has been made to warn about
1479 all such cases, there are probably some cases that are not warned about,
1480 even though G++ is generating incompatible code. There may also be
1481 cases where warnings are emitted even though the code that is generated
1484 You should rewrite your code to avoid these warnings if you are
1485 concerned about the fact that code generated by G++ may not be binary
1486 compatible with code generated by other compilers.
1488 The known incompatibilities at this point include:
1493 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1494 pack data into the same byte as a base class. For example:
1497 struct A @{ virtual void f(); int f1 : 1; @};
1498 struct B : public A @{ int f2 : 1; @};
1502 In this case, G++ will place @code{B::f2} into the same byte
1503 as@code{A::f1}; other compilers will not. You can avoid this problem
1504 by explicitly padding @code{A} so that its size is a multiple of the
1505 byte size on your platform; that will cause G++ and other compilers to
1506 layout @code{B} identically.
1509 Incorrect handling of tail-padding for virtual bases. G++ does not use
1510 tail padding when laying out virtual bases. For example:
1513 struct A @{ virtual void f(); char c1; @};
1514 struct B @{ B(); char c2; @};
1515 struct C : public A, public virtual B @{@};
1519 In this case, G++ will not place @code{B} into the tail-padding for
1520 @code{A}; other compilers will. You can avoid this problem by
1521 explicitly padding @code{A} so that its size is a multiple of its
1522 alignment (ignoring virtual base classes); that will cause G++ and other
1523 compilers to layout @code{C} identically.
1526 Incorrect handling of bit-fields with declared widths greater than that
1527 of their underlying types, when the bit-fields appear in a union. For
1531 union U @{ int i : 4096; @};
1535 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1536 union too small by the number of bits in an @code{int}.
1539 Empty classes can be placed at incorrect offsets. For example:
1549 struct C : public B, public A @{@};
1553 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1554 it should be placed at offset zero. G++ mistakenly believes that the
1555 @code{A} data member of @code{B} is already at offset zero.
1558 Names of template functions whose types involve @code{typename} or
1559 template template parameters can be mangled incorrectly.
1562 template <typename Q>
1563 void f(typename Q::X) @{@}
1565 template <template <typename> class Q>
1566 void f(typename Q<int>::X) @{@}
1570 Instantiations of these templates may be mangled incorrectly.
1574 @item -Wctor-dtor-privacy @r{(C++ only)}
1575 @opindex Wctor-dtor-privacy
1576 Warn when a class seems unusable because all the constructors or
1577 destructors in that class are private, and it has neither friends nor
1578 public static member functions.
1580 @item -Wnon-virtual-dtor @r{(C++ only)}
1581 @opindex Wnon-virtual-dtor
1582 Warn when a class appears to be polymorphic, thereby requiring a virtual
1583 destructor, yet it declares a non-virtual one.
1584 This warning is enabled by @option{-Wall}.
1586 @item -Wreorder @r{(C++ only)}
1588 @cindex reordering, warning
1589 @cindex warning for reordering of member initializers
1590 Warn when the order of member initializers given in the code does not
1591 match the order in which they must be executed. For instance:
1597 A(): j (0), i (1) @{ @}
1601 The compiler will rearrange the member initializers for @samp{i}
1602 and @samp{j} to match the declaration order of the members, emitting
1603 a warning to that effect. This warning is enabled by @option{-Wall}.
1606 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1609 @item -Weffc++ @r{(C++ only)}
1611 Warn about violations of the following style guidelines from Scott Meyers'
1612 @cite{Effective C++} book:
1616 Item 11: Define a copy constructor and an assignment operator for classes
1617 with dynamically allocated memory.
1620 Item 12: Prefer initialization to assignment in constructors.
1623 Item 14: Make destructors virtual in base classes.
1626 Item 15: Have @code{operator=} return a reference to @code{*this}.
1629 Item 23: Don't try to return a reference when you must return an object.
1633 Also warn about violations of the following style guidelines from
1634 Scott Meyers' @cite{More Effective C++} book:
1638 Item 6: Distinguish between prefix and postfix forms of increment and
1639 decrement operators.
1642 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1646 When selecting this option, be aware that the standard library
1647 headers do not obey all of these guidelines; use @samp{grep -v}
1648 to filter out those warnings.
1650 @item -Wno-deprecated @r{(C++ only)}
1651 @opindex Wno-deprecated
1652 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1654 @item -Wno-non-template-friend @r{(C++ only)}
1655 @opindex Wno-non-template-friend
1656 Disable warnings when non-templatized friend functions are declared
1657 within a template. Since the advent of explicit template specification
1658 support in G++, if the name of the friend is an unqualified-id (i.e.,
1659 @samp{friend foo(int)}), the C++ language specification demands that the
1660 friend declare or define an ordinary, nontemplate function. (Section
1661 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1662 could be interpreted as a particular specialization of a templatized
1663 function. Because this non-conforming behavior is no longer the default
1664 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1665 check existing code for potential trouble spots and is on by default.
1666 This new compiler behavior can be turned off with
1667 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1668 but disables the helpful warning.
1670 @item -Wold-style-cast @r{(C++ only)}
1671 @opindex Wold-style-cast
1672 Warn if an old-style (C-style) cast to a non-void type is used within
1673 a C++ program. The new-style casts (@samp{static_cast},
1674 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1675 unintended effects and much easier to search for.
1677 @item -Woverloaded-virtual @r{(C++ only)}
1678 @opindex Woverloaded-virtual
1679 @cindex overloaded virtual fn, warning
1680 @cindex warning for overloaded virtual fn
1681 Warn when a function declaration hides virtual functions from a
1682 base class. For example, in:
1689 struct B: public A @{
1694 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1702 will fail to compile.
1704 @item -Wno-pmf-conversions @r{(C++ only)}
1705 @opindex Wno-pmf-conversions
1706 Disable the diagnostic for converting a bound pointer to member function
1709 @item -Wsign-promo @r{(C++ only)}
1710 @opindex Wsign-promo
1711 Warn when overload resolution chooses a promotion from unsigned or
1712 enumeral type to a signed type, over a conversion to an unsigned type of
1713 the same size. Previous versions of G++ would try to preserve
1714 unsignedness, but the standard mandates the current behavior.
1716 @item -Wsynth @r{(C++ only)}
1718 @cindex warning for synthesized methods
1719 @cindex synthesized methods, warning
1720 Warn when G++'s synthesis behavior does not match that of cfront. For
1726 A& operator = (int);
1736 In this example, G++ will synthesize a default @samp{A& operator =
1737 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1740 @node Objective-C Dialect Options
1741 @section Options Controlling Objective-C Dialect
1743 @cindex compiler options, Objective-C
1744 @cindex Objective-C options, command line
1745 @cindex options, Objective-C
1746 This section describes the command-line options that are only meaningful
1747 for Objective-C programs, but you can also use most of the GNU compiler
1748 options regardless of what language your program is in. For example,
1749 you might compile a file @code{some_class.m} like this:
1752 gcc -g -fgnu-runtime -O -c some_class.m
1756 In this example, @option{-fgnu-runtime} is an option meant only for
1757 Objective-C programs; you can use the other options with any language
1760 Here is a list of options that are @emph{only} for compiling Objective-C
1764 @item -fconstant-string-class=@var{class-name}
1765 @opindex fconstant-string-class
1766 Use @var{class-name} as the name of the class to instantiate for each
1767 literal string specified with the syntax @code{@@"@dots{}"}. The default
1768 class name is @code{NXConstantString}.
1771 @opindex fgnu-runtime
1772 Generate object code compatible with the standard GNU Objective-C
1773 runtime. This is the default for most types of systems.
1775 @item -fnext-runtime
1776 @opindex fnext-runtime
1777 Generate output compatible with the NeXT runtime. This is the default
1778 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1779 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1784 Dump interface declarations for all classes seen in the source file to a
1785 file named @file{@var{sourcename}.decl}.
1788 @opindex Wno-protocol
1789 If a class is declared to implement a protocol, a warning is issued for
1790 every method in the protocol that is not implemented by the class. The
1791 default behavior is to issue a warning for every method not explicitly
1792 implemented in the class, even if a method implementation is inherited
1793 from the superclass. If you use the @code{-Wno-protocol} option, then
1794 methods inherited from the superclass are considered to be implemented,
1795 and no warning is issued for them.
1799 Warn if multiple methods of different types for the same selector are
1800 found during compilation. The check is performed on the list of methods
1801 in the final stage of compilation. Additionally, a check is performed
1802 for each selector appearing in a @code{@@selector(@dots{})}
1803 expression, and a corresponding method for that selector has been found
1804 during compilation. Because these checks scan the method table only at
1805 the end of compilation, these warnings are not produced if the final
1806 stage of compilation is not reached, for example because an error is
1807 found during compilation, or because the @code{-fsyntax-only} option is
1810 @item -Wundeclared-selector
1811 @opindex Wundeclared-selector
1812 Warn if a @code{@@selector(@dots{})} expression referring to an
1813 undeclared selector is found. A selector is considered undeclared if no
1814 method with that name has been declared before the
1815 @code{@@selector(@dots{})} expression, either explicitly in an
1816 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1817 an @code{@@implementation} section. This option always performs its
1818 checks as soon as a @code{@@selector(@dots{})} expression is found,
1819 while @code{-Wselector} only performs its checks in the final stage of
1820 compilation. This also enforces the coding style convention
1821 that methods and selectors must be declared before being used.
1823 @c not documented because only avail via -Wp
1824 @c @item -print-objc-runtime-info
1828 @node Language Independent Options
1829 @section Options to Control Diagnostic Messages Formatting
1830 @cindex options to control diagnostics formatting
1831 @cindex diagnostic messages
1832 @cindex message formatting
1834 Traditionally, diagnostic messages have been formatted irrespective of
1835 the output device's aspect (e.g.@: its width, @dots{}). The options described
1836 below can be used to control the diagnostic messages formatting
1837 algorithm, e.g.@: how many characters per line, how often source location
1838 information should be reported. Right now, only the C++ front end can
1839 honor these options. However it is expected, in the near future, that
1840 the remaining front ends would be able to digest them correctly.
1843 @item -fmessage-length=@var{n}
1844 @opindex fmessage-length
1845 Try to format error messages so that they fit on lines of about @var{n}
1846 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1847 the front ends supported by GCC@. If @var{n} is zero, then no
1848 line-wrapping will be done; each error message will appear on a single
1851 @opindex fdiagnostics-show-location
1852 @item -fdiagnostics-show-location=once
1853 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1854 reporter to emit @emph{once} source location information; that is, in
1855 case the message is too long to fit on a single physical line and has to
1856 be wrapped, the source location won't be emitted (as prefix) again,
1857 over and over, in subsequent continuation lines. This is the default
1860 @item -fdiagnostics-show-location=every-line
1861 Only meaningful in line-wrapping mode. Instructs the diagnostic
1862 messages reporter to emit the same source location information (as
1863 prefix) for physical lines that result from the process of breaking
1864 a message which is too long to fit on a single line.
1868 @node Warning Options
1869 @section Options to Request or Suppress Warnings
1870 @cindex options to control warnings
1871 @cindex warning messages
1872 @cindex messages, warning
1873 @cindex suppressing warnings
1875 Warnings are diagnostic messages that report constructions which
1876 are not inherently erroneous but which are risky or suggest there
1877 may have been an error.
1879 You can request many specific warnings with options beginning @samp{-W},
1880 for example @option{-Wimplicit} to request warnings on implicit
1881 declarations. Each of these specific warning options also has a
1882 negative form beginning @samp{-Wno-} to turn off warnings;
1883 for example, @option{-Wno-implicit}. This manual lists only one of the
1884 two forms, whichever is not the default.
1886 The following options control the amount and kinds of warnings produced
1887 by GCC; for further, language-specific options also refer to
1888 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
1891 @cindex syntax checking
1893 @opindex fsyntax-only
1894 Check the code for syntax errors, but don't do anything beyond that.
1898 Issue all the warnings demanded by strict ISO C and ISO C++;
1899 reject all programs that use forbidden extensions, and some other
1900 programs that do not follow ISO C and ISO C++. For ISO C, follows the
1901 version of the ISO C standard specified by any @option{-std} option used.
1903 Valid ISO C and ISO C++ programs should compile properly with or without
1904 this option (though a rare few will require @option{-ansi} or a
1905 @option{-std} option specifying the required version of ISO C)@. However,
1906 without this option, certain GNU extensions and traditional C and C++
1907 features are supported as well. With this option, they are rejected.
1909 @option{-pedantic} does not cause warning messages for use of the
1910 alternate keywords whose names begin and end with @samp{__}. Pedantic
1911 warnings are also disabled in the expression that follows
1912 @code{__extension__}. However, only system header files should use
1913 these escape routes; application programs should avoid them.
1914 @xref{Alternate Keywords}.
1916 Some users try to use @option{-pedantic} to check programs for strict ISO
1917 C conformance. They soon find that it does not do quite what they want:
1918 it finds some non-ISO practices, but not all---only those for which
1919 ISO C @emph{requires} a diagnostic, and some others for which
1920 diagnostics have been added.
1922 A feature to report any failure to conform to ISO C might be useful in
1923 some instances, but would require considerable additional work and would
1924 be quite different from @option{-pedantic}. We don't have plans to
1925 support such a feature in the near future.
1927 Where the standard specified with @option{-std} represents a GNU
1928 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
1929 corresponding @dfn{base standard}, the version of ISO C on which the GNU
1930 extended dialect is based. Warnings from @option{-pedantic} are given
1931 where they are required by the base standard. (It would not make sense
1932 for such warnings to be given only for features not in the specified GNU
1933 C dialect, since by definition the GNU dialects of C include all
1934 features the compiler supports with the given option, and there would be
1935 nothing to warn about.)
1937 @item -pedantic-errors
1938 @opindex pedantic-errors
1939 Like @option{-pedantic}, except that errors are produced rather than
1944 Inhibit all warning messages.
1948 Inhibit warning messages about the use of @samp{#import}.
1950 @item -Wchar-subscripts
1951 @opindex Wchar-subscripts
1952 Warn if an array subscript has type @code{char}. This is a common cause
1953 of error, as programmers often forget that this type is signed on some
1958 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
1959 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
1963 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
1964 the arguments supplied have types appropriate to the format string
1965 specified, and that the conversions specified in the format string make
1966 sense. This includes standard functions, and others specified by format
1967 attributes (@pxref{Function Attributes}), in the @code{printf},
1968 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
1969 not in the C standard) families.
1971 The formats are checked against the format features supported by GNU
1972 libc version 2.2. These include all ISO C90 and C99 features, as well
1973 as features from the Single Unix Specification and some BSD and GNU
1974 extensions. Other library implementations may not support all these
1975 features; GCC does not support warning about features that go beyond a
1976 particular library's limitations. However, if @option{-pedantic} is used
1977 with @option{-Wformat}, warnings will be given about format features not
1978 in the selected standard version (but not for @code{strfmon} formats,
1979 since those are not in any version of the C standard). @xref{C Dialect
1980 Options,,Options Controlling C Dialect}.
1982 Since @option{-Wformat} also checks for null format arguments for
1983 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
1985 @option{-Wformat} is included in @option{-Wall}. For more control over some
1986 aspects of format checking, the options @option{-Wno-format-y2k},
1987 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
1988 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
1989 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
1991 @item -Wno-format-y2k
1992 @opindex Wno-format-y2k
1993 If @option{-Wformat} is specified, do not warn about @code{strftime}
1994 formats which may yield only a two-digit year.
1996 @item -Wno-format-extra-args
1997 @opindex Wno-format-extra-args
1998 If @option{-Wformat} is specified, do not warn about excess arguments to a
1999 @code{printf} or @code{scanf} format function. The C standard specifies
2000 that such arguments are ignored.
2002 Where the unused arguments lie between used arguments that are
2003 specified with @samp{$} operand number specifications, normally
2004 warnings are still given, since the implementation could not know what
2005 type to pass to @code{va_arg} to skip the unused arguments. However,
2006 in the case of @code{scanf} formats, this option will suppress the
2007 warning if the unused arguments are all pointers, since the Single
2008 Unix Specification says that such unused arguments are allowed.
2010 @item -Wno-format-zero-length
2011 @opindex Wno-format-zero-length
2012 If @option{-Wformat} is specified, do not warn about zero-length formats.
2013 The C standard specifies that zero-length formats are allowed.
2015 @item -Wformat-nonliteral
2016 @opindex Wformat-nonliteral
2017 If @option{-Wformat} is specified, also warn if the format string is not a
2018 string literal and so cannot be checked, unless the format function
2019 takes its format arguments as a @code{va_list}.
2021 @item -Wformat-security
2022 @opindex Wformat-security
2023 If @option{-Wformat} is specified, also warn about uses of format
2024 functions that represent possible security problems. At present, this
2025 warns about calls to @code{printf} and @code{scanf} functions where the
2026 format string is not a string literal and there are no format arguments,
2027 as in @code{printf (foo);}. This may be a security hole if the format
2028 string came from untrusted input and contains @samp{%n}. (This is
2029 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2030 in future warnings may be added to @option{-Wformat-security} that are not
2031 included in @option{-Wformat-nonliteral}.)
2035 Enable @option{-Wformat} plus format checks not included in
2036 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2037 -Wformat-nonliteral -Wformat-security}.
2041 Enable warning about passing a null pointer for arguments marked as
2042 requiring a non-null value by the @code{nonnull} function attribute.
2044 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2045 can be disabled with the @option{-Wno-nonnull} option.
2047 @item -Winit-self @r{(C, C++, and Objective-C only)}
2049 Enable warning about uninitialized variables which are initalized with themselves.
2050 Note this option can only be used with the @option{-Wuninitialized} option and
2051 that only works with @option{-O}.
2053 For an example, the following code will not warn about i being uninitialized
2054 without this option:
2065 @item -Wimplicit-int
2066 @opindex Wimplicit-int
2067 Warn when a declaration does not specify a type.
2069 @item -Wimplicit-function-declaration
2070 @itemx -Werror-implicit-function-declaration
2071 @opindex Wimplicit-function-declaration
2072 @opindex Werror-implicit-function-declaration
2073 Give a warning (or error) whenever a function is used before being
2078 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2082 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2083 function with external linkage, returning int, taking either zero
2084 arguments, two, or three arguments of appropriate types.
2086 @item -Wmissing-braces
2087 @opindex Wmissing-braces
2088 Warn if an aggregate or union initializer is not fully bracketed. In
2089 the following example, the initializer for @samp{a} is not fully
2090 bracketed, but that for @samp{b} is fully bracketed.
2093 int a[2][2] = @{ 0, 1, 2, 3 @};
2094 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2098 @opindex Wparentheses
2099 Warn if parentheses are omitted in certain contexts, such
2100 as when there is an assignment in a context where a truth value
2101 is expected, or when operators are nested whose precedence people
2102 often get confused about.
2104 Also warn about constructions where there may be confusion to which
2105 @code{if} statement an @code{else} branch belongs. Here is an example of
2120 In C, every @code{else} branch belongs to the innermost possible @code{if}
2121 statement, which in this example is @code{if (b)}. This is often not
2122 what the programmer expected, as illustrated in the above example by
2123 indentation the programmer chose. When there is the potential for this
2124 confusion, GCC will issue a warning when this flag is specified.
2125 To eliminate the warning, add explicit braces around the innermost
2126 @code{if} statement so there is no way the @code{else} could belong to
2127 the enclosing @code{if}. The resulting code would look like this:
2143 @item -Wsequence-point
2144 @opindex Wsequence-point
2145 Warn about code that may have undefined semantics because of violations
2146 of sequence point rules in the C standard.
2148 The C standard defines the order in which expressions in a C program are
2149 evaluated in terms of @dfn{sequence points}, which represent a partial
2150 ordering between the execution of parts of the program: those executed
2151 before the sequence point, and those executed after it. These occur
2152 after the evaluation of a full expression (one which is not part of a
2153 larger expression), after the evaluation of the first operand of a
2154 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2155 function is called (but after the evaluation of its arguments and the
2156 expression denoting the called function), and in certain other places.
2157 Other than as expressed by the sequence point rules, the order of
2158 evaluation of subexpressions of an expression is not specified. All
2159 these rules describe only a partial order rather than a total order,
2160 since, for example, if two functions are called within one expression
2161 with no sequence point between them, the order in which the functions
2162 are called is not specified. However, the standards committee have
2163 ruled that function calls do not overlap.
2165 It is not specified when between sequence points modifications to the
2166 values of objects take effect. Programs whose behavior depends on this
2167 have undefined behavior; the C standard specifies that ``Between the
2168 previous and next sequence point an object shall have its stored value
2169 modified at most once by the evaluation of an expression. Furthermore,
2170 the prior value shall be read only to determine the value to be
2171 stored.''. If a program breaks these rules, the results on any
2172 particular implementation are entirely unpredictable.
2174 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2175 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2176 diagnosed by this option, and it may give an occasional false positive
2177 result, but in general it has been found fairly effective at detecting
2178 this sort of problem in programs.
2180 The present implementation of this option only works for C programs. A
2181 future implementation may also work for C++ programs.
2183 The C standard is worded confusingly, therefore there is some debate
2184 over the precise meaning of the sequence point rules in subtle cases.
2185 Links to discussions of the problem, including proposed formal
2186 definitions, may be found on our readings page, at
2187 @w{@uref{http://gcc.gnu.org/readings.html}}.
2190 @opindex Wreturn-type
2191 Warn whenever a function is defined with a return-type that defaults to
2192 @code{int}. Also warn about any @code{return} statement with no
2193 return-value in a function whose return-type is not @code{void}.
2195 For C++, a function without return type always produces a diagnostic
2196 message, even when @option{-Wno-return-type} is specified. The only
2197 exceptions are @samp{main} and functions defined in system headers.
2201 Warn whenever a @code{switch} statement has an index of enumeral type
2202 and lacks a @code{case} for one or more of the named codes of that
2203 enumeration. (The presence of a @code{default} label prevents this
2204 warning.) @code{case} labels outside the enumeration range also
2205 provoke warnings when this option is used.
2207 @item -Wswitch-default
2208 @opindex Wswitch-switch
2209 Warn whenever a @code{switch} statement does not have a @code{default}
2213 @opindex Wswitch-enum
2214 Warn whenever a @code{switch} statement has an index of enumeral type
2215 and lacks a @code{case} for one or more of the named codes of that
2216 enumeration. @code{case} labels outside the enumeration range also
2217 provoke warnings when this option is used.
2221 Warn if any trigraphs are encountered that might change the meaning of
2222 the program (trigraphs within comments are not warned about).
2224 @item -Wunused-function
2225 @opindex Wunused-function
2226 Warn whenever a static function is declared but not defined or a
2227 non\-inline static function is unused.
2229 @item -Wunused-label
2230 @opindex Wunused-label
2231 Warn whenever a label is declared but not used.
2233 To suppress this warning use the @samp{unused} attribute
2234 (@pxref{Variable Attributes}).
2236 @item -Wunused-parameter
2237 @opindex Wunused-parameter
2238 Warn whenever a function parameter is unused aside from its declaration.
2240 To suppress this warning use the @samp{unused} attribute
2241 (@pxref{Variable Attributes}).
2243 @item -Wunused-variable
2244 @opindex Wunused-variable
2245 Warn whenever a local variable or non-constant static variable is unused
2246 aside from its declaration
2248 To suppress this warning use the @samp{unused} attribute
2249 (@pxref{Variable Attributes}).
2251 @item -Wunused-value
2252 @opindex Wunused-value
2253 Warn whenever a statement computes a result that is explicitly not used.
2255 To suppress this warning cast the expression to @samp{void}.
2259 All the above @option{-Wunused} options combined.
2261 In order to get a warning about an unused function parameter, you must
2262 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2263 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2265 @item -Wuninitialized
2266 @opindex Wuninitialized
2267 Warn if an automatic variable is used without first being initialized or
2268 if a variable may be clobbered by a @code{setjmp} call.
2270 These warnings are possible only in optimizing compilation,
2271 because they require data flow information that is computed only
2272 when optimizing. If you don't specify @option{-O}, you simply won't
2275 If you want to warn about code which uses the uninitialized value of the
2276 variable in its own initializer, use the @option{-Winit-self} option.
2278 These warnings occur only for variables that are candidates for
2279 register allocation. Therefore, they do not occur for a variable that
2280 is declared @code{volatile}, or whose address is taken, or whose size
2281 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2282 structures, unions or arrays, even when they are in registers.
2284 Note that there may be no warning about a variable that is used only
2285 to compute a value that itself is never used, because such
2286 computations may be deleted by data flow analysis before the warnings
2289 These warnings are made optional because GCC is not smart
2290 enough to see all the reasons why the code might be correct
2291 despite appearing to have an error. Here is one example of how
2312 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2313 always initialized, but GCC doesn't know this. Here is
2314 another common case:
2319 if (change_y) save_y = y, y = new_y;
2321 if (change_y) y = save_y;
2326 This has no bug because @code{save_y} is used only if it is set.
2328 @cindex @code{longjmp} warnings
2329 This option also warns when a non-volatile automatic variable might be
2330 changed by a call to @code{longjmp}. These warnings as well are possible
2331 only in optimizing compilation.
2333 The compiler sees only the calls to @code{setjmp}. It cannot know
2334 where @code{longjmp} will be called; in fact, a signal handler could
2335 call it at any point in the code. As a result, you may get a warning
2336 even when there is in fact no problem because @code{longjmp} cannot
2337 in fact be called at the place which would cause a problem.
2339 Some spurious warnings can be avoided if you declare all the functions
2340 you use that never return as @code{noreturn}. @xref{Function
2343 @item -Wunknown-pragmas
2344 @opindex Wunknown-pragmas
2345 @cindex warning for unknown pragmas
2346 @cindex unknown pragmas, warning
2347 @cindex pragmas, warning of unknown
2348 Warn when a #pragma directive is encountered which is not understood by
2349 GCC@. If this command line option is used, warnings will even be issued
2350 for unknown pragmas in system header files. This is not the case if
2351 the warnings were only enabled by the @option{-Wall} command line option.
2353 @item -Wstrict-aliasing
2354 @opindex Wstrict-aliasing
2355 This option is only active when @option{-fstrict-aliasing} is active.
2356 It warns about code which might break the strict aliasing rules that the
2357 compiler is using for optimization. The warning does not catch all
2358 cases, but does attempt to catch the more common pitfalls. It is
2359 included in @option{-Wall}.
2363 All of the above @samp{-W} options combined. This enables all the
2364 warnings about constructions that some users consider questionable, and
2365 that are easy to avoid (or modify to prevent the warning), even in
2366 conjunction with macros. This also enables some language-specific
2367 warnings described in @ref{C++ Dialect Options} and
2368 @ref{Objective-C Dialect Options}.
2371 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2372 Some of them warn about constructions that users generally do not
2373 consider questionable, but which occasionally you might wish to check
2374 for; others warn about constructions that are necessary or hard to avoid
2375 in some cases, and there is no simple way to modify the code to suppress
2382 (This option used to be called @option{-W}. The older name is still
2383 supported, but the newer name is more descriptive.) Print extra warning
2384 messages for these events:
2388 A function can return either with or without a value. (Falling
2389 off the end of the function body is considered returning without
2390 a value.) For example, this function would evoke such a
2404 An expression-statement or the left-hand side of a comma expression
2405 contains no side effects.
2406 To suppress the warning, cast the unused expression to void.
2407 For example, an expression such as @samp{x[i,j]} will cause a warning,
2408 but @samp{x[(void)i,j]} will not.
2411 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2414 A comparison like @samp{x<=y<=z} appears; this is equivalent to
2415 @samp{(x<=y ? 1 : 0) <= z}, which is a different interpretation from
2416 that of ordinary mathematical notation.
2419 Storage-class specifiers like @code{static} are not the first things in
2420 a declaration. According to the C Standard, this usage is obsolescent.
2423 The return type of a function has a type qualifier such as @code{const}.
2424 Such a type qualifier has no effect, since the value returned by a
2425 function is not an lvalue. (But don't warn about the GNU extension of
2426 @code{volatile void} return types. That extension will be warned about
2427 if @option{-pedantic} is specified.)
2430 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2434 A comparison between signed and unsigned values could produce an
2435 incorrect result when the signed value is converted to unsigned.
2436 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2439 An aggregate has an initializer which does not initialize all members.
2440 For example, the following code would cause such a warning, because
2441 @code{x.h} would be implicitly initialized to zero:
2444 struct s @{ int f, g, h; @};
2445 struct s x = @{ 3, 4 @};
2449 A function parameter is declared without a type specifier in K&R-style
2457 An empty body occurs in an @samp{if} or @samp{else} statement.
2460 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2461 @samp{>}, or @samp{>=}.
2464 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2467 Any of several floating-point events that often indicate errors, such as
2468 overflow, underflow, loss of precision, etc.
2470 @item @r{(C++ only)}
2471 An enumerator and a non-enumerator both appear in a conditional expression.
2473 @item @r{(C++ only)}
2474 A non-static reference or non-static @samp{const} member appears in a
2475 class without constructors.
2477 @item @r{(C++ only)}
2478 Ambiguous virtual bases.
2480 @item @r{(C++ only)}
2481 Subscripting an array which has been declared @samp{register}.
2483 @item @r{(C++ only)}
2484 Taking the address of a variable which has been declared @samp{register}.
2486 @item @r{(C++ only)}
2487 A base class is not initialized in a derived class' copy constructor.
2490 @item -Wno-div-by-zero
2491 @opindex Wno-div-by-zero
2492 @opindex Wdiv-by-zero
2493 Do not warn about compile-time integer division by zero. Floating point
2494 division by zero is not warned about, as it can be a legitimate way of
2495 obtaining infinities and NaNs.
2497 @item -Wsystem-headers
2498 @opindex Wsystem-headers
2499 @cindex warnings from system headers
2500 @cindex system headers, warnings from
2501 Print warning messages for constructs found in system header files.
2502 Warnings from system headers are normally suppressed, on the assumption
2503 that they usually do not indicate real problems and would only make the
2504 compiler output harder to read. Using this command line option tells
2505 GCC to emit warnings from system headers as if they occurred in user
2506 code. However, note that using @option{-Wall} in conjunction with this
2507 option will @emph{not} warn about unknown pragmas in system
2508 headers---for that, @option{-Wunknown-pragmas} must also be used.
2511 @opindex Wfloat-equal
2512 Warn if floating point values are used in equality comparisons.
2514 The idea behind this is that sometimes it is convenient (for the
2515 programmer) to consider floating-point values as approximations to
2516 infinitely precise real numbers. If you are doing this, then you need
2517 to compute (by analyzing the code, or in some other way) the maximum or
2518 likely maximum error that the computation introduces, and allow for it
2519 when performing comparisons (and when producing output, but that's a
2520 different problem). In particular, instead of testing for equality, you
2521 would check to see whether the two values have ranges that overlap; and
2522 this is done with the relational operators, so equality comparisons are
2525 @item -Wtraditional @r{(C only)}
2526 @opindex Wtraditional
2527 Warn about certain constructs that behave differently in traditional and
2528 ISO C@. Also warn about ISO C constructs that have no traditional C
2529 equivalent, and/or problematic constructs which should be avoided.
2533 Macro parameters that appear within string literals in the macro body.
2534 In traditional C macro replacement takes place within string literals,
2535 but does not in ISO C@.
2538 In traditional C, some preprocessor directives did not exist.
2539 Traditional preprocessors would only consider a line to be a directive
2540 if the @samp{#} appeared in column 1 on the line. Therefore
2541 @option{-Wtraditional} warns about directives that traditional C
2542 understands but would ignore because the @samp{#} does not appear as the
2543 first character on the line. It also suggests you hide directives like
2544 @samp{#pragma} not understood by traditional C by indenting them. Some
2545 traditional implementations would not recognize @samp{#elif}, so it
2546 suggests avoiding it altogether.
2549 A function-like macro that appears without arguments.
2552 The unary plus operator.
2555 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2556 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2557 constants.) Note, these suffixes appear in macros defined in the system
2558 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2559 Use of these macros in user code might normally lead to spurious
2560 warnings, however gcc's integrated preprocessor has enough context to
2561 avoid warning in these cases.
2564 A function declared external in one block and then used after the end of
2568 A @code{switch} statement has an operand of type @code{long}.
2571 A non-@code{static} function declaration follows a @code{static} one.
2572 This construct is not accepted by some traditional C compilers.
2575 The ISO type of an integer constant has a different width or
2576 signedness from its traditional type. This warning is only issued if
2577 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2578 typically represent bit patterns, are not warned about.
2581 Usage of ISO string concatenation is detected.
2584 Initialization of automatic aggregates.
2587 Identifier conflicts with labels. Traditional C lacks a separate
2588 namespace for labels.
2591 Initialization of unions. If the initializer is zero, the warning is
2592 omitted. This is done under the assumption that the zero initializer in
2593 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2594 initializer warnings and relies on default initialization to zero in the
2598 Conversions by prototypes between fixed/floating point values and vice
2599 versa. The absence of these prototypes when compiling with traditional
2600 C would cause serious problems. This is a subset of the possible
2601 conversion warnings, for the full set use @option{-Wconversion}.
2604 Use of ISO C style function definitions. This warning intentionally is
2605 @emph{not} issued for prototype declarations or variadic functions
2606 because these ISO C features will appear in your code when using
2607 libiberty's traditional C compatibility macros, @code{PARAMS} and
2608 @code{VPARAMS}. This warning is also bypassed for nested functions
2609 because that feature is already a gcc extension and thus not relevant to
2610 traditional C compatibility.
2613 @item -Wdeclaration-after-statement @r{(C only)}
2614 @opindex Wdeclaration-after-statement
2615 Warn when a declaration is found after a statement in a block. This
2616 construct, known from C++, was introduced with ISO C99 and is by default
2617 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2618 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2622 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2624 @item -Wendif-labels
2625 @opindex Wendif-labels
2626 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2630 Warn whenever a local variable shadows another local variable, parameter or
2631 global variable or whenever a built-in function is shadowed.
2633 @item -Wlarger-than-@var{len}
2634 @opindex Wlarger-than
2635 Warn whenever an object of larger than @var{len} bytes is defined.
2637 @item -Wpointer-arith
2638 @opindex Wpointer-arith
2639 Warn about anything that depends on the ``size of'' a function type or
2640 of @code{void}. GNU C assigns these types a size of 1, for
2641 convenience in calculations with @code{void *} pointers and pointers
2644 @item -Wbad-function-cast @r{(C only)}
2645 @opindex Wbad-function-cast
2646 Warn whenever a function call is cast to a non-matching type.
2647 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2651 Warn whenever a pointer is cast so as to remove a type qualifier from
2652 the target type. For example, warn if a @code{const char *} is cast
2653 to an ordinary @code{char *}.
2656 @opindex Wcast-align
2657 Warn whenever a pointer is cast such that the required alignment of the
2658 target is increased. For example, warn if a @code{char *} is cast to
2659 an @code{int *} on machines where integers can only be accessed at
2660 two- or four-byte boundaries.
2662 @item -Wwrite-strings
2663 @opindex Wwrite-strings
2664 When compiling C, give string constants the type @code{const
2665 char[@var{length}]} so that
2666 copying the address of one into a non-@code{const} @code{char *}
2667 pointer will get a warning; when compiling C++, warn about the
2668 deprecated conversion from string constants to @code{char *}.
2669 These warnings will help you find at
2670 compile time code that can try to write into a string constant, but
2671 only if you have been very careful about using @code{const} in
2672 declarations and prototypes. Otherwise, it will just be a nuisance;
2673 this is why we did not make @option{-Wall} request these warnings.
2676 @opindex Wconversion
2677 Warn if a prototype causes a type conversion that is different from what
2678 would happen to the same argument in the absence of a prototype. This
2679 includes conversions of fixed point to floating and vice versa, and
2680 conversions changing the width or signedness of a fixed point argument
2681 except when the same as the default promotion.
2683 Also, warn if a negative integer constant expression is implicitly
2684 converted to an unsigned type. For example, warn about the assignment
2685 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2686 casts like @code{(unsigned) -1}.
2688 @item -Wsign-compare
2689 @opindex Wsign-compare
2690 @cindex warning for comparison of signed and unsigned values
2691 @cindex comparison of signed and unsigned values, warning
2692 @cindex signed and unsigned values, comparison warning
2693 Warn when a comparison between signed and unsigned values could produce
2694 an incorrect result when the signed value is converted to unsigned.
2695 This warning is also enabled by @option{-Wextra}; to get the other warnings
2696 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2698 @item -Waggregate-return
2699 @opindex Waggregate-return
2700 Warn if any functions that return structures or unions are defined or
2701 called. (In languages where you can return an array, this also elicits
2704 @item -Wstrict-prototypes @r{(C only)}
2705 @opindex Wstrict-prototypes
2706 Warn if a function is declared or defined without specifying the
2707 argument types. (An old-style function definition is permitted without
2708 a warning if preceded by a declaration which specifies the argument
2711 @item -Wold-style-definition @r{(C only)}
2712 @opindex Wold-style-definition
2713 Warn if an old-style function definition is used. A warning is given
2714 even if there is a previous prototype.
2716 @item -Wmissing-prototypes @r{(C only)}
2717 @opindex Wmissing-prototypes
2718 Warn if a global function is defined without a previous prototype
2719 declaration. This warning is issued even if the definition itself
2720 provides a prototype. The aim is to detect global functions that fail
2721 to be declared in header files.
2723 @item -Wmissing-declarations @r{(C only)}
2724 @opindex Wmissing-declarations
2725 Warn if a global function is defined without a previous declaration.
2726 Do so even if the definition itself provides a prototype.
2727 Use this option to detect global functions that are not declared in
2730 @item -Wmissing-noreturn
2731 @opindex Wmissing-noreturn
2732 Warn about functions which might be candidates for attribute @code{noreturn}.
2733 Note these are only possible candidates, not absolute ones. Care should
2734 be taken to manually verify functions actually do not ever return before
2735 adding the @code{noreturn} attribute, otherwise subtle code generation
2736 bugs could be introduced. You will not get a warning for @code{main} in
2737 hosted C environments.
2739 @item -Wmissing-format-attribute
2740 @opindex Wmissing-format-attribute
2742 If @option{-Wformat} is enabled, also warn about functions which might be
2743 candidates for @code{format} attributes. Note these are only possible
2744 candidates, not absolute ones. GCC will guess that @code{format}
2745 attributes might be appropriate for any function that calls a function
2746 like @code{vprintf} or @code{vscanf}, but this might not always be the
2747 case, and some functions for which @code{format} attributes are
2748 appropriate may not be detected. This option has no effect unless
2749 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2751 @item -Wno-multichar
2752 @opindex Wno-multichar
2754 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2755 Usually they indicate a typo in the user's code, as they have
2756 implementation-defined values, and should not be used in portable code.
2758 @item -Wno-deprecated-declarations
2759 @opindex Wno-deprecated-declarations
2760 Do not warn about uses of functions, variables, and types marked as
2761 deprecated by using the @code{deprecated} attribute.
2762 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2763 @pxref{Type Attributes}.)
2767 Warn if a structure is given the packed attribute, but the packed
2768 attribute has no effect on the layout or size of the structure.
2769 Such structures may be mis-aligned for little benefit. For
2770 instance, in this code, the variable @code{f.x} in @code{struct bar}
2771 will be misaligned even though @code{struct bar} does not itself
2772 have the packed attribute:
2779 @} __attribute__((packed));
2789 Warn if padding is included in a structure, either to align an element
2790 of the structure or to align the whole structure. Sometimes when this
2791 happens it is possible to rearrange the fields of the structure to
2792 reduce the padding and so make the structure smaller.
2794 @item -Wredundant-decls
2795 @opindex Wredundant-decls
2796 Warn if anything is declared more than once in the same scope, even in
2797 cases where multiple declaration is valid and changes nothing.
2799 @item -Wnested-externs @r{(C only)}
2800 @opindex Wnested-externs
2801 Warn if an @code{extern} declaration is encountered within a function.
2803 @item -Wunreachable-code
2804 @opindex Wunreachable-code
2805 Warn if the compiler detects that code will never be executed.
2807 This option is intended to warn when the compiler detects that at
2808 least a whole line of source code will never be executed, because
2809 some condition is never satisfied or because it is after a
2810 procedure that never returns.
2812 It is possible for this option to produce a warning even though there
2813 are circumstances under which part of the affected line can be executed,
2814 so care should be taken when removing apparently-unreachable code.
2816 For instance, when a function is inlined, a warning may mean that the
2817 line is unreachable in only one inlined copy of the function.
2819 This option is not made part of @option{-Wall} because in a debugging
2820 version of a program there is often substantial code which checks
2821 correct functioning of the program and is, hopefully, unreachable
2822 because the program does work. Another common use of unreachable
2823 code is to provide behavior which is selectable at compile-time.
2827 Warn if a function can not be inlined and it was declared as inline.
2828 Even with this option, the compiler will not warn about failures to
2829 inline functions declared in system headers.
2831 The compiler uses a variety of heuristics to determine whether or not
2832 to inline a function. For example, the compiler takes into account
2833 the size of the function being inlined and the the amount of inlining
2834 that has already been done in the current function. Therefore,
2835 seemingly insignificant changes in the source program can cause the
2836 warnings produced by @option{-Winline} to appear or disappear.
2838 @item -Wno-invalid-offsetof @r{(C++ only)}
2839 @opindex Wno-invalid-offsetof
2840 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
2841 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
2842 to a non-POD type is undefined. In existing C++ implementations,
2843 however, @samp{offsetof} typically gives meaningful results even when
2844 applied to certain kinds of non-POD types. (Such as a simple
2845 @samp{struct} that fails to be a POD type only by virtue of having a
2846 constructor.) This flag is for users who are aware that they are
2847 writing nonportable code and who have deliberately chosen to ignore the
2850 The restrictions on @samp{offsetof} may be relaxed in a future version
2851 of the C++ standard.
2854 @opindex Winvalid-pch
2855 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
2856 the search path but can't be used.
2860 @opindex Wno-long-long
2861 Warn if @samp{long long} type is used. This is default. To inhibit
2862 the warning messages, use @option{-Wno-long-long}. Flags
2863 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
2864 only when @option{-pedantic} flag is used.
2866 @item -Wdisabled-optimization
2867 @opindex Wdisabled-optimization
2868 Warn if a requested optimization pass is disabled. This warning does
2869 not generally indicate that there is anything wrong with your code; it
2870 merely indicates that GCC's optimizers were unable to handle the code
2871 effectively. Often, the problem is that your code is too big or too
2872 complex; GCC will refuse to optimize programs when the optimization
2873 itself is likely to take inordinate amounts of time.
2877 Make all warnings into errors.
2880 @node Debugging Options
2881 @section Options for Debugging Your Program or GCC
2882 @cindex options, debugging
2883 @cindex debugging information options
2885 GCC has various special options that are used for debugging
2886 either your program or GCC:
2891 Produce debugging information in the operating system's native format
2892 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
2895 On most systems that use stabs format, @option{-g} enables use of extra
2896 debugging information that only GDB can use; this extra information
2897 makes debugging work better in GDB but will probably make other debuggers
2899 refuse to read the program. If you want to control for certain whether
2900 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
2901 @option{-gxcoff+}, @option{-gxcoff}, @option{-gdwarf-1+}, @option{-gdwarf-1},
2902 or @option{-gvms} (see below).
2904 Unlike most other C compilers, GCC allows you to use @option{-g} with
2905 @option{-O}. The shortcuts taken by optimized code may occasionally
2906 produce surprising results: some variables you declared may not exist
2907 at all; flow of control may briefly move where you did not expect it;
2908 some statements may not be executed because they compute constant
2909 results or their values were already at hand; some statements may
2910 execute in different places because they were moved out of loops.
2912 Nevertheless it proves possible to debug optimized output. This makes
2913 it reasonable to use the optimizer for programs that might have bugs.
2915 The following options are useful when GCC is generated with the
2916 capability for more than one debugging format.
2920 Produce debugging information for use by GDB@. This means to use the
2921 most expressive format available (DWARF 2, stabs, or the native format
2922 if neither of those are supported), including GDB extensions if at all
2927 Produce debugging information in stabs format (if that is supported),
2928 without GDB extensions. This is the format used by DBX on most BSD
2929 systems. On MIPS, Alpha and System V Release 4 systems this option
2930 produces stabs debugging output which is not understood by DBX or SDB@.
2931 On System V Release 4 systems this option requires the GNU assembler.
2933 @item -feliminate-unused-debug-symbols
2934 @opindex feliminate-unused-debug-symbols
2935 Produce debugging information in stabs format (if that is supported),
2936 for only symbols that are actually used.
2940 Produce debugging information in stabs format (if that is supported),
2941 using GNU extensions understood only by the GNU debugger (GDB)@. The
2942 use of these extensions is likely to make other debuggers crash or
2943 refuse to read the program.
2947 Produce debugging information in COFF format (if that is supported).
2948 This is the format used by SDB on most System V systems prior to
2953 Produce debugging information in XCOFF format (if that is supported).
2954 This is the format used by the DBX debugger on IBM RS/6000 systems.
2958 Produce debugging information in XCOFF format (if that is supported),
2959 using GNU extensions understood only by the GNU debugger (GDB)@. The
2960 use of these extensions is likely to make other debuggers crash or
2961 refuse to read the program, and may cause assemblers other than the GNU
2962 assembler (GAS) to fail with an error.
2966 Produce debugging information in DWARF version 1 format (if that is
2967 supported). This is the format used by SDB on most System V Release 4
2970 This option is deprecated.
2974 Produce debugging information in DWARF version 1 format (if that is
2975 supported), using GNU extensions understood only by the GNU debugger
2976 (GDB)@. The use of these extensions is likely to make other debuggers
2977 crash or refuse to read the program.
2979 This option is deprecated.
2983 Produce debugging information in DWARF version 2 format (if that is
2984 supported). This is the format used by DBX on IRIX 6.
2988 Produce debugging information in VMS debug format (if that is
2989 supported). This is the format used by DEBUG on VMS systems.
2992 @itemx -ggdb@var{level}
2993 @itemx -gstabs@var{level}
2994 @itemx -gcoff@var{level}
2995 @itemx -gxcoff@var{level}
2996 @itemx -gvms@var{level}
2997 Request debugging information and also use @var{level} to specify how
2998 much information. The default level is 2.
3000 Level 1 produces minimal information, enough for making backtraces in
3001 parts of the program that you don't plan to debug. This includes
3002 descriptions of functions and external variables, but no information
3003 about local variables and no line numbers.
3005 Level 3 includes extra information, such as all the macro definitions
3006 present in the program. Some debuggers support macro expansion when
3007 you use @option{-g3}.
3009 Note that in order to avoid confusion between DWARF1 debug level 2,
3010 and DWARF2, neither @option{-gdwarf} nor @option{-gdwarf-2} accept
3011 a concatenated debug level. Instead use an additional @option{-g@var{level}}
3012 option to change the debug level for DWARF1 or DWARF2.
3014 @item -feliminate-dwarf2-dups
3015 @opindex feliminate-dwarf2-dups
3016 Compress DWARF2 debugging information by eliminating duplicated
3017 information about each symbol. This option only makes sense when
3018 generating DWARF2 debugging information with @option{-gdwarf-2}.
3020 @cindex @command{prof}
3023 Generate extra code to write profile information suitable for the
3024 analysis program @command{prof}. You must use this option when compiling
3025 the source files you want data about, and you must also use it when
3028 @cindex @command{gprof}
3031 Generate extra code to write profile information suitable for the
3032 analysis program @command{gprof}. You must use this option when compiling
3033 the source files you want data about, and you must also use it when
3038 Makes the compiler print out each function name as it is compiled, and
3039 print some statistics about each pass when it finishes.
3042 @opindex ftime-report
3043 Makes the compiler print some statistics about the time consumed by each
3044 pass when it finishes.
3047 @opindex fmem-report
3048 Makes the compiler print some statistics about permanent memory
3049 allocation when it finishes.
3051 @item -fprofile-arcs
3052 @opindex fprofile-arcs
3053 Add code so that program flow @dfn{arcs} are instrumented. During
3054 execution the program records how many times each branch and call is
3055 executed and how many times it is taken or returns. When the compiled
3056 program exits it saves this data to a file called
3057 @file{@var{auxname}.gcda} for each source file. The data may be used for
3058 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3059 test coverage analysis (@option{-ftest-coverage}). Each object file's
3060 @var{auxname} is generated from the name of the output file, if
3061 explicitly specified and it is not the final executable, otherwise it is
3062 the basename of the source file. In both cases any suffix is removed
3063 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3064 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3069 Compile the source files with @option{-fprofile-arcs} plus optimization
3070 and code generation options. For test coverage analysis, use the
3071 additional @option{-ftest-coverage} option. You do not need to profile
3072 every source file in a program.
3075 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3076 (the latter implies the former).
3079 Run the program on a representative workload to generate the arc profile
3080 information. This may be repeated any number of times. You can run
3081 concurrent instances of your program, and provided that the file system
3082 supports locking, the data files will be correctly updated. Also
3083 @code{fork} calls are detected and correctly handled (double counting
3087 For profile-directed optimizations, compile the source files again with
3088 the same optimization and code generation options plus
3089 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3090 Control Optimization}).
3093 For test coverage analysis, use @command{gcov} to produce human readable
3094 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3095 @command{gcov} documentation for further information.
3099 With @option{-fprofile-arcs}, for each function of your program GCC
3100 creates a program flow graph, then finds a spanning tree for the graph.
3101 Only arcs that are not on the spanning tree have to be instrumented: the
3102 compiler adds code to count the number of times that these arcs are
3103 executed. When an arc is the only exit or only entrance to a block, the
3104 instrumentation code can be added to the block; otherwise, a new basic
3105 block must be created to hold the instrumentation code.
3108 @item -ftest-coverage
3109 @opindex ftest-coverage
3110 Produce a notes file that the @command{gcov} code-coverage utility
3111 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3112 show program coverage. Each source file's note file is called
3113 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3114 above for a description of @var{auxname} and instructions on how to
3115 generate test coverage data. Coverage data will match the source files
3116 more closely, if you do not optimize.
3118 @item -d@var{letters}
3120 Says to make debugging dumps during compilation at times specified by
3121 @var{letters}. This is used for debugging the compiler. The file names
3122 for most of the dumps are made by appending a pass number and a word to
3123 the @var{dumpname}. @var{dumpname} is generated from the name of the
3124 output file, if explicitly specified and it is not an executable,
3125 otherwise it is the basename of the source file. In both cases any
3126 suffix is removed (e.g. @file{foo.00.rtl} or @file{foo.01.sibling}).
3127 Here are the possible letters for use in @var{letters}, and their
3133 Annotate the assembler output with miscellaneous debugging information.
3136 Dump after computing branch probabilities, to @file{@var{file}.16.bp}.
3139 Dump after block reordering, to @file{@var{file}.32.bbro}.
3142 Dump after instruction combination, to the file @file{@var{file}.22.combine}.
3145 Dump after the first if conversion, to the file @file{@var{file}.17.ce1}.
3146 Also dump after the second if conversion, to the file @file{@var{file}.23.ce2}.
3149 Dump after branch target load optimization, to to @file{@var{file}.34.btl}.
3150 Also dump after delayed branch scheduling, to @file{@var{file}.37.dbr}.
3153 Dump all macro definitions, at the end of preprocessing, in addition to
3157 Dump after SSA optimizations, to @file{@var{file}.05.ssa} and
3158 @file{@var{file}.010.ussa}.
3161 Dump after the second if conversion, to @file{@var{file}.33.ce3}.
3164 Dump after control and data flow analysis, to @file{@var{file}.15.cfg}.
3165 Also dump after life analysis, to @file{@var{file}.21.life}.
3168 Dump after purging @code{ADDRESSOF} codes, to @file{@var{file}.11.addressof}.
3171 Dump after global register allocation, to @file{@var{file}.27.greg}.
3174 Dump after GCSE, to @file{@var{file}.12.gcse}.
3175 Also dump after jump bypassing and control flow optimizations, to
3176 @file{@var{file}.14.bypass}.
3179 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3182 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3185 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3188 Dump after conversion from registers to stack, to @file{@var{file}.36.stack}.
3191 Dump after local register allocation, to @file{@var{file}.26.lreg}.
3194 Dump after loop optimization passes, to @file{@var{file}.13.loop} and
3195 @file{@var{file}.19.loop2}.
3198 Dump after performing the machine dependent reorganization pass, to
3199 @file{@var{file}.37.mach}.
3202 Dump after register renumbering, to @file{@var{file}.31.rnreg}.
3205 Dump after the register move pass, to @file{@var{file}.24.regmove}.
3208 Dump after post-reload optimizations, to @file{@var{file}.28.postreload}.
3211 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3214 Dump after the second scheduling pass, to @file{@var{file}.35.sched2}.
3217 Dump after CSE (including the jump optimization that sometimes follows
3218 CSE), to @file{@var{file}.019.cse}.
3221 Dump after the first scheduling pass, to @file{@var{file}.25.sched}.
3224 Dump after the second CSE pass (including the jump optimization that
3225 sometimes follows CSE), to @file{@var{file}.20.cse2}.
3228 Dump after running tracer, to @file{@var{file}.18.tracer}.
3231 Dump after null pointer elimination pass to @file{@var{file}.018.null}.
3234 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3237 Dump after the second flow pass, to @file{@var{file}.29.flow2}.
3240 Dump after SSA conditional constant propagation, to
3241 @file{@var{file}.06.ssaccp}.
3244 Dump after SSA dead code elimination, to @file{@var{file}.07.ssadce}.
3247 Dump after the peephole pass, to @file{@var{file}.30.peephole2}.
3250 Produce all the dumps listed above.
3253 Produce a core dump whenever an error occurs.
3256 Print statistics on memory usage, at the end of the run, to
3260 Annotate the assembler output with a comment indicating which
3261 pattern and alternative was used. The length of each instruction is
3265 Dump the RTL in the assembler output as a comment before each instruction.
3266 Also turns on @option{-dp} annotation.
3269 For each of the other indicated dump files (except for
3270 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3271 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3274 Just generate RTL for a function instead of compiling it. Usually used
3278 Dump debugging information during parsing, to standard error.
3281 @item -fdump-unnumbered
3282 @opindex fdump-unnumbered
3283 When doing debugging dumps (see @option{-d} option above), suppress instruction
3284 numbers and line number note output. This makes it more feasible to
3285 use diff on debugging dumps for compiler invocations with different
3286 options, in particular with and without @option{-g}.
3288 @item -fdump-translation-unit @r{(C and C++ only)}
3289 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3290 @opindex fdump-translation-unit
3291 Dump a representation of the tree structure for the entire translation
3292 unit to a file. The file name is made by appending @file{.tu} to the
3293 source file name. If the @samp{-@var{options}} form is used, @var{options}
3294 controls the details of the dump as described for the
3295 @option{-fdump-tree} options.
3297 @item -fdump-class-hierarchy @r{(C++ only)}
3298 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3299 @opindex fdump-class-hierarchy
3300 Dump a representation of each class's hierarchy and virtual function
3301 table layout to a file. The file name is made by appending @file{.class}
3302 to the source file name. If the @samp{-@var{options}} form is used,
3303 @var{options} controls the details of the dump as described for the
3304 @option{-fdump-tree} options.
3306 @item -fdump-tree-@var{switch} @r{(C++ only)}
3307 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C++ only)}
3309 Control the dumping at various stages of processing the intermediate
3310 language tree to a file. The file name is generated by appending a switch
3311 specific suffix to the source file name. If the @samp{-@var{options}}
3312 form is used, @var{options} is a list of @samp{-} separated options that
3313 control the details of the dump. Not all options are applicable to all
3314 dumps, those which are not meaningful will be ignored. The following
3315 options are available
3319 Print the address of each node. Usually this is not meaningful as it
3320 changes according to the environment and source file. Its primary use
3321 is for tying up a dump file with a debug environment.
3323 Inhibit dumping of members of a scope or body of a function merely
3324 because that scope has been reached. Only dump such items when they
3325 are directly reachable by some other path.
3327 Turn on all options.
3330 The following tree dumps are possible:
3333 Dump before any tree based optimization, to @file{@var{file}.original}.
3335 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3337 Dump after function inlining, to @file{@var{file}.inlined}.
3340 @item -frandom-seed=@var{string}
3341 @opindex frandom-string
3342 This option provides a seed that GCC uses when it would otherwise use
3343 random numbers. It is used to generate certain symbol names
3344 that have to be different in every compiled file. It is also used to
3345 place unique stamps in coverage data files and the object files that
3346 produce them. You can use the @option{-frandom-seed} option to produce
3347 reproducibly identical object files.
3349 The @var{string} should be different for every file you compile.
3351 @item -fsched-verbose=@var{n}
3352 @opindex fsched-verbose
3353 On targets that use instruction scheduling, this option controls the
3354 amount of debugging output the scheduler prints. This information is
3355 written to standard error, unless @option{-dS} or @option{-dR} is
3356 specified, in which case it is output to the usual dump
3357 listing file, @file{.sched} or @file{.sched2} respectively. However
3358 for @var{n} greater than nine, the output is always printed to standard
3361 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3362 same information as @option{-dRS}. For @var{n} greater than one, it
3363 also output basic block probabilities, detailed ready list information
3364 and unit/insn info. For @var{n} greater than two, it includes RTL
3365 at abort point, control-flow and regions info. And for @var{n} over
3366 four, @option{-fsched-verbose} also includes dependence info.
3370 Store the usual ``temporary'' intermediate files permanently; place them
3371 in the current directory and name them based on the source file. Thus,
3372 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3373 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3374 preprocessed @file{foo.i} output file even though the compiler now
3375 normally uses an integrated preprocessor.
3379 Report the CPU time taken by each subprocess in the compilation
3380 sequence. For C source files, this is the compiler proper and assembler
3381 (plus the linker if linking is done). The output looks like this:
3388 The first number on each line is the ``user time,'' that is time spent
3389 executing the program itself. The second number is ``system time,''
3390 time spent executing operating system routines on behalf of the program.
3391 Both numbers are in seconds.
3393 @item -print-file-name=@var{library}
3394 @opindex print-file-name
3395 Print the full absolute name of the library file @var{library} that
3396 would be used when linking---and don't do anything else. With this
3397 option, GCC does not compile or link anything; it just prints the
3400 @item -print-multi-directory
3401 @opindex print-multi-directory
3402 Print the directory name corresponding to the multilib selected by any
3403 other switches present in the command line. This directory is supposed
3404 to exist in @env{GCC_EXEC_PREFIX}.
3406 @item -print-multi-lib
3407 @opindex print-multi-lib
3408 Print the mapping from multilib directory names to compiler switches
3409 that enable them. The directory name is separated from the switches by
3410 @samp{;}, and each switch starts with an @samp{@@} instead of the
3411 @samp{-}, without spaces between multiple switches. This is supposed to
3412 ease shell-processing.
3414 @item -print-prog-name=@var{program}
3415 @opindex print-prog-name
3416 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3418 @item -print-libgcc-file-name
3419 @opindex print-libgcc-file-name
3420 Same as @option{-print-file-name=libgcc.a}.
3422 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3423 but you do want to link with @file{libgcc.a}. You can do
3426 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3429 @item -print-search-dirs
3430 @opindex print-search-dirs
3431 Print the name of the configured installation directory and a list of
3432 program and library directories gcc will search---and don't do anything else.
3434 This is useful when gcc prints the error message
3435 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3436 To resolve this you either need to put @file{cpp0} and the other compiler
3437 components where gcc expects to find them, or you can set the environment
3438 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3439 Don't forget the trailing '/'.
3440 @xref{Environment Variables}.
3443 @opindex dumpmachine
3444 Print the compiler's target machine (for example,
3445 @samp{i686-pc-linux-gnu})---and don't do anything else.
3448 @opindex dumpversion
3449 Print the compiler version (for example, @samp{3.0})---and don't do
3454 Print the compiler's built-in specs---and don't do anything else. (This
3455 is used when GCC itself is being built.) @xref{Spec Files}.
3457 @item -feliminate-unused-debug-types
3458 @opindex feliminate-unused-debug-types
3459 Normally, when producing DWARF2 output, GCC will emit debugging
3460 information for all types declared in a compilation
3461 unit, regardless of whether or not they are actually used
3462 in that compilation unit. Sometimes this is useful, such as
3463 if, in the debugger, you want to cast a value to a type that is
3464 not actually used in your program (but is declared). More often,
3465 however, this results in a significant amount of wasted space.
3466 With this option, GCC will avoid producing debug symbol output
3467 for types that are nowhere used in the source file being compiled.
3470 @node Optimize Options
3471 @section Options That Control Optimization
3472 @cindex optimize options
3473 @cindex options, optimization
3475 These options control various sorts of optimizations.
3477 Without any optimization option, the compiler's goal is to reduce the
3478 cost of compilation and to make debugging produce the expected
3479 results. Statements are independent: if you stop the program with a
3480 breakpoint between statements, you can then assign a new value to any
3481 variable or change the program counter to any other statement in the
3482 function and get exactly the results you would expect from the source
3485 Turning on optimization flags makes the compiler attempt to improve
3486 the performance and/or code size at the expense of compilation time
3487 and possibly the ability to debug the program.
3489 The compiler performs optimisation based on the knowledge it has of
3490 the program. Using the @option{-funit-at-a-time} flag will allow the
3491 compiler to consider information gained from later functions in the
3492 file when compiling a function. Compiling multiple files at once to a
3493 single output file (and using @option{-funit-at-a-time}) will allow
3494 the compiler to use information gained from all of the files when
3495 compiling each of them.
3497 Not all optimizations are controlled directly by a flag. Only
3498 optimizations that have a flag are listed.
3505 Optimize. Optimizing compilation takes somewhat more time, and a lot
3506 more memory for a large function.
3508 With @option{-O}, the compiler tries to reduce code size and execution
3509 time, without performing any optimizations that take a great deal of
3512 @option{-O} turns on the following optimization flags:
3513 @gccoptlist{-fdefer-pop @gol
3514 -fmerge-constants @gol
3516 -floop-optimize @gol
3518 -fif-conversion @gol
3519 -fif-conversion2 @gol
3520 -fdelayed-branch @gol
3521 -fguess-branch-probability @gol
3524 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3525 where doing so does not interfere with debugging.
3529 Optimize even more. GCC performs nearly all supported optimizations
3530 that do not involve a space-speed tradeoff. The compiler does not
3531 perform loop unrolling or function inlining when you specify @option{-O2}.
3532 As compared to @option{-O}, this option increases both compilation time
3533 and the performance of the generated code.
3535 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3536 also turns on the following optimization flags:
3537 @gccoptlist{-fforce-mem @gol
3538 -foptimize-sibling-calls @gol
3539 -fstrength-reduce @gol
3540 -fcse-follow-jumps -fcse-skip-blocks @gol
3541 -frerun-cse-after-loop -frerun-loop-opt @gol
3542 -fgcse -fgcse-lm -fgcse-sm @gol
3543 -fdelete-null-pointer-checks @gol
3544 -fexpensive-optimizations @gol
3546 -fschedule-insns -fschedule-insns2 @gol
3547 -fsched-interblock -fsched-spec @gol
3550 -freorder-blocks -freorder-functions @gol
3551 -fstrict-aliasing @gol
3552 -falign-functions -falign-jumps @gol
3553 -falign-loops -falign-labels}
3555 Please note the warning under @option{-fgcse} about
3556 invoking @option{-O2} on programs that use computed gotos.
3560 Optimize yet more. @option{-O3} turns on all optimizations specified by
3561 @option{-O2} and also turns on the @option{-finline-functions},
3562 @option{-funit-at-a-time} and @option{-frename-registers} options.
3566 Do not optimize. This is the default.
3570 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3571 do not typically increase code size. It also performs further
3572 optimizations designed to reduce code size.
3574 @option{-Os} disables the following optimization flags:
3575 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3576 -falign-labels -freorder-blocks -fprefetch-loop-arrays}
3578 If you use multiple @option{-O} options, with or without level numbers,
3579 the last such option is the one that is effective.
3582 Options of the form @option{-f@var{flag}} specify machine-independent
3583 flags. Most flags have both positive and negative forms; the negative
3584 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3585 below, only one of the forms is listed---the one you typically will
3586 use. You can figure out the other form by either removing @samp{no-}
3589 The following options control specific optimizations. They are either
3590 activated by @option{-O} options or are related to ones that are. You
3591 can use the following flags in the rare cases when ``fine-tuning'' of
3592 optimizations to be performed is desired.
3595 @item -fno-default-inline
3596 @opindex fno-default-inline
3597 Do not make member functions inline by default merely because they are
3598 defined inside the class scope (C++ only). Otherwise, when you specify
3599 @w{@option{-O}}, member functions defined inside class scope are compiled
3600 inline by default; i.e., you don't need to add @samp{inline} in front of
3601 the member function name.
3603 @item -fno-defer-pop
3604 @opindex fno-defer-pop
3605 Always pop the arguments to each function call as soon as that function
3606 returns. For machines which must pop arguments after a function call,
3607 the compiler normally lets arguments accumulate on the stack for several
3608 function calls and pops them all at once.
3610 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3614 Force memory operands to be copied into registers before doing
3615 arithmetic on them. This produces better code by making all memory
3616 references potential common subexpressions. When they are not common
3617 subexpressions, instruction combination should eliminate the separate
3620 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3623 @opindex fforce-addr
3624 Force memory address constants to be copied into registers before
3625 doing arithmetic on them. This may produce better code just as
3626 @option{-fforce-mem} may.
3628 @item -fomit-frame-pointer
3629 @opindex fomit-frame-pointer
3630 Don't keep the frame pointer in a register for functions that
3631 don't need one. This avoids the instructions to save, set up and
3632 restore frame pointers; it also makes an extra register available
3633 in many functions. @strong{It also makes debugging impossible on
3636 On some machines, such as the VAX, this flag has no effect, because
3637 the standard calling sequence automatically handles the frame pointer
3638 and nothing is saved by pretending it doesn't exist. The
3639 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3640 whether a target machine supports this flag. @xref{Registers,,Register
3641 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3643 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3645 @item -foptimize-sibling-calls
3646 @opindex foptimize-sibling-calls
3647 Optimize sibling and tail recursive calls.
3649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3653 Don't pay attention to the @code{inline} keyword. Normally this option
3654 is used to keep the compiler from expanding any functions inline.
3655 Note that if you are not optimizing, no functions can be expanded inline.
3657 @item -finline-functions
3658 @opindex finline-functions
3659 Integrate all simple functions into their callers. The compiler
3660 heuristically decides which functions are simple enough to be worth
3661 integrating in this way.
3663 If all calls to a given function are integrated, and the function is
3664 declared @code{static}, then the function is normally not output as
3665 assembler code in its own right.
3667 Enabled at level @option{-O3}.
3669 @item -finline-limit=@var{n}
3670 @opindex finline-limit
3671 By default, gcc limits the size of functions that can be inlined. This flag
3672 allows the control of this limit for functions that are explicitly marked as
3673 inline (i.e., marked with the inline keyword or defined within the class
3674 definition in c++). @var{n} is the size of functions that can be inlined in
3675 number of pseudo instructions (not counting parameter handling). The default
3676 value of @var{n} is 600.
3677 Increasing this value can result in more inlined code at
3678 the cost of compilation time and memory consumption. Decreasing usually makes
3679 the compilation faster and less code will be inlined (which presumably
3680 means slower programs). This option is particularly useful for programs that
3681 use inlining heavily such as those based on recursive templates with C++.
3683 Inlining is actually controlled by a number of parameters, which may be
3684 specified individually by using @option{--param @var{name}=@var{value}}.
3685 The @option{-finline-limit=@var{n}} option sets some of these parameters
3689 @item max-inline-insns
3691 @item max-inline-insns-single
3692 is set to @var{n}/2.
3693 @item max-inline-insns-auto
3694 is set to @var{n}/2.
3695 @item min-inline-insns
3696 is set to 130 or @var{n}/4, whichever is smaller.
3697 @item max-inline-insns-rtl
3701 Using @option{-finline-limit=600} thus results in the default settings
3702 for these parameters. See below for a documentation of the individual
3703 parameters controlling inlining.
3705 @emph{Note:} pseudo instruction represents, in this particular context, an
3706 abstract measurement of function's size. In no way, it represents a count
3707 of assembly instructions and as such its exact meaning might change from one
3708 release to an another.
3710 @item -fkeep-inline-functions
3711 @opindex fkeep-inline-functions
3712 Even if all calls to a given function are integrated, and the function
3713 is declared @code{static}, nevertheless output a separate run-time
3714 callable version of the function. This switch does not affect
3715 @code{extern inline} functions.
3717 @item -fkeep-static-consts
3718 @opindex fkeep-static-consts
3719 Emit variables declared @code{static const} when optimization isn't turned
3720 on, even if the variables aren't referenced.
3722 GCC enables this option by default. If you want to force the compiler to
3723 check if the variable was referenced, regardless of whether or not
3724 optimization is turned on, use the @option{-fno-keep-static-consts} option.
3726 @item -fmerge-constants
3727 Attempt to merge identical constants (string constants and floating point
3728 constants) across compilation units.
3730 This option is the default for optimized compilation if the assembler and
3731 linker support it. Use @option{-fno-merge-constants} to inhibit this
3734 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3736 @item -fmerge-all-constants
3737 Attempt to merge identical constants and identical variables.
3739 This option implies @option{-fmerge-constants}. In addition to
3740 @option{-fmerge-constants} this considers e.g. even constant initialized
3741 arrays or initialized constant variables with integral or floating point
3742 types. Languages like C or C++ require each non-automatic variable to
3743 have distinct location, so using this option will result in non-conforming
3748 Use a graph coloring register allocator. Currently this option is meant
3749 for testing, so we are interested to hear about miscompilations with
3752 @item -fno-branch-count-reg
3753 @opindex fno-branch-count-reg
3754 Do not use ``decrement and branch'' instructions on a count register,
3755 but instead generate a sequence of instructions that decrement a
3756 register, compare it against zero, then branch based upon the result.
3757 This option is only meaningful on architectures that support such
3758 instructions, which include x86, PowerPC, IA-64 and S/390.
3760 The default is @option{-fbranch-count-reg}, enabled when
3761 @option{-fstrength-reduce} is enabled.
3763 @item -fno-function-cse
3764 @opindex fno-function-cse
3765 Do not put function addresses in registers; make each instruction that
3766 calls a constant function contain the function's address explicitly.
3768 This option results in less efficient code, but some strange hacks
3769 that alter the assembler output may be confused by the optimizations
3770 performed when this option is not used.
3772 The default is @option{-ffunction-cse}
3774 @item -fno-zero-initialized-in-bss
3775 @opindex fno-zero-initialized-in-bss
3776 If the target supports a BSS section, GCC by default puts variables that
3777 are initialized to zero into BSS@. This can save space in the resulting
3780 This option turns off this behavior because some programs explicitly
3781 rely on variables going to the data section. E.g., so that the
3782 resulting executable can find the beginning of that section and/or make
3783 assumptions based on that.
3785 The default is @option{-fzero-initialized-in-bss}.
3787 @item -fstrength-reduce
3788 @opindex fstrength-reduce
3789 Perform the optimizations of loop strength reduction and
3790 elimination of iteration variables.
3792 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3794 @item -fthread-jumps
3795 @opindex fthread-jumps
3796 Perform optimizations where we check to see if a jump branches to a
3797 location where another comparison subsumed by the first is found. If
3798 so, the first branch is redirected to either the destination of the
3799 second branch or a point immediately following it, depending on whether
3800 the condition is known to be true or false.
3802 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3804 @item -fcse-follow-jumps
3805 @opindex fcse-follow-jumps
3806 In common subexpression elimination, scan through jump instructions
3807 when the target of the jump is not reached by any other path. For
3808 example, when CSE encounters an @code{if} statement with an
3809 @code{else} clause, CSE will follow the jump when the condition
3812 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3814 @item -fcse-skip-blocks
3815 @opindex fcse-skip-blocks
3816 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
3817 follow jumps which conditionally skip over blocks. When CSE
3818 encounters a simple @code{if} statement with no else clause,
3819 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
3820 body of the @code{if}.
3822 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3824 @item -frerun-cse-after-loop
3825 @opindex frerun-cse-after-loop
3826 Re-run common subexpression elimination after loop optimizations has been
3829 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3831 @item -frerun-loop-opt
3832 @opindex frerun-loop-opt
3833 Run the loop optimizer twice.
3835 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3839 Perform a global common subexpression elimination pass.
3840 This pass also performs global constant and copy propagation.
3842 @emph{Note:} When compiling a program using computed gotos, a GCC
3843 extension, you may get better runtime performance if you disable
3844 the global common subexpression elimination pass by adding
3845 @option{-fno-gcse} to the command line.
3847 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3851 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
3852 attempt to move loads which are only killed by stores into themselves. This
3853 allows a loop containing a load/store sequence to be changed to a load outside
3854 the loop, and a copy/store within the loop.
3856 Enabled by default when gcse is enabled.
3860 When @option{-fgcse-sm} is enabled, A store motion pass is run after global common
3861 subexpression elimination. This pass will attempt to move stores out of loops.
3862 When used in conjunction with @option{-fgcse-lm}, loops containing a load/store sequence
3863 can be changed to a load before the loop and a store after the loop.
3865 Enabled by default when gcse is enabled.
3867 @item -floop-optimize
3868 @opindex floop-optimize
3869 Perform loop optimizations: move constant expressions out of loops, simplify
3870 exit test conditions and optionally do strength-reduction and loop unrolling as
3873 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3875 @item -fcrossjumping
3876 @opindex crossjumping
3877 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
3878 resulting code may or may not perform better than without cross-jumping.
3880 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3882 @item -fif-conversion
3883 @opindex if-conversion
3884 Attempt to transform conditional jumps into branch-less equivalents. This
3885 include use of conditional moves, min, max, set flags and abs instructions, and
3886 some tricks doable by standard arithmetics. The use of conditional execution
3887 on chips where it is available is controlled by @code{if-conversion2}.
3889 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3891 @item -fif-conversion2
3892 @opindex if-conversion2
3893 Use conditional execution (where available) to transform conditional jumps into
3894 branch-less equivalents.
3896 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3898 @item -fdelete-null-pointer-checks
3899 @opindex fdelete-null-pointer-checks
3900 Use global dataflow analysis to identify and eliminate useless checks
3901 for null pointers. The compiler assumes that dereferencing a null
3902 pointer would have halted the program. If a pointer is checked after
3903 it has already been dereferenced, it cannot be null.
3905 In some environments, this assumption is not true, and programs can
3906 safely dereference null pointers. Use
3907 @option{-fno-delete-null-pointer-checks} to disable this optimization
3908 for programs which depend on that behavior.
3910 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3912 @item -fexpensive-optimizations
3913 @opindex fexpensive-optimizations
3914 Perform a number of minor optimizations that are relatively expensive.
3916 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3918 @item -foptimize-register-move
3920 @opindex foptimize-register-move
3922 Attempt to reassign register numbers in move instructions and as
3923 operands of other simple instructions in order to maximize the amount of
3924 register tying. This is especially helpful on machines with two-operand
3927 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
3930 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3932 @item -fdelayed-branch
3933 @opindex fdelayed-branch
3934 If supported for the target machine, attempt to reorder instructions
3935 to exploit instruction slots available after delayed branch
3938 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3940 @item -fschedule-insns
3941 @opindex fschedule-insns
3942 If supported for the target machine, attempt to reorder instructions to
3943 eliminate execution stalls due to required data being unavailable. This
3944 helps machines that have slow floating point or memory load instructions
3945 by allowing other instructions to be issued until the result of the load
3946 or floating point instruction is required.
3948 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3950 @item -fschedule-insns2
3951 @opindex fschedule-insns2
3952 Similar to @option{-fschedule-insns}, but requests an additional pass of
3953 instruction scheduling after register allocation has been done. This is
3954 especially useful on machines with a relatively small number of
3955 registers and where memory load instructions take more than one cycle.
3957 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3959 @item -fno-sched-interblock
3960 @opindex fno-sched-interblock
3961 Don't schedule instructions across basic blocks. This is normally
3962 enabled by default when scheduling before register allocation, i.e.@:
3963 with @option{-fschedule-insns} or at @option{-O2} or higher.
3965 @item -fno-sched-spec
3966 @opindex fno-sched-spec
3967 Don't allow speculative motion of non-load instructions. This is normally
3968 enabled by default when scheduling before register allocation, i.e.@:
3969 with @option{-fschedule-insns} or at @option{-O2} or higher.
3971 @item -fsched-spec-load
3972 @opindex fsched-spec-load
3973 Allow speculative motion of some load instructions. This only makes
3974 sense when scheduling before register allocation, i.e.@: with
3975 @option{-fschedule-insns} or at @option{-O2} or higher.
3977 @item -fsched-spec-load-dangerous
3978 @opindex fsched-spec-load-dangerous
3979 Allow speculative motion of more load instructions. This only makes
3980 sense when scheduling before register allocation, i.e.@: with
3981 @option{-fschedule-insns} or at @option{-O2} or higher.
3983 @item -fsched2-use-superblocks
3984 @opindex fsched2-use-superblocks
3985 When scheduling after register allocation, do use superblock scheduling
3986 algorithm. Superblock scheduling allows motion across basic block boundaries
3987 resulting on faster schedules. This option is experimental, as not all machine
3988 descriptions used by GCC model the CPU closely enough to avoid unreliable
3989 results from the algorithm.
3991 This only makes sense when scheduling after register allocation, i.e.@: with
3992 @option{-fschedule-insns2} or at @option{-O2} or higher.
3994 @item -fsched2-use-traces
3995 @opindex fsched2-use-traces
3996 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
3997 allocation and additionally perform code duplication in order to increase the
3998 size of superblocks using tracer pass. See @option{-ftracer} for details on
4001 This mode should produce faster but significantly longer programs. Also
4002 without @code{-fbranch-probabilities} the traces constructed may not match the
4003 reality and hurt the performance. This only makes
4004 sense when scheduling after register allocation, i.e.@: with
4005 @option{-fschedule-insns2} or at @option{-O2} or higher.
4007 @item -fcaller-saves
4008 @opindex fcaller-saves
4009 Enable values to be allocated in registers that will be clobbered by
4010 function calls, by emitting extra instructions to save and restore the
4011 registers around such calls. Such allocation is done only when it
4012 seems to result in better code than would otherwise be produced.
4014 This option is always enabled by default on certain machines, usually
4015 those which have no call-preserved registers to use instead.
4017 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4019 @item -fmove-all-movables
4020 @opindex fmove-all-movables
4021 Forces all invariant computations in loops to be moved
4024 @item -freduce-all-givs
4025 @opindex freduce-all-givs
4026 Forces all general-induction variables in loops to be
4029 @emph{Note:} When compiling programs written in Fortran,
4030 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4031 by default when you use the optimizer.
4033 These options may generate better or worse code; results are highly
4034 dependent on the structure of loops within the source code.
4036 These two options are intended to be removed someday, once
4037 they have helped determine the efficacy of various
4038 approaches to improving loop optimizations.
4040 Please let us (@w{@email{gcc@@gcc.gnu.org}} and @w{@email{fortran@@gnu.org}})
4041 know how use of these options affects
4042 the performance of your production code.
4043 We're very interested in code that runs @emph{slower}
4044 when these options are @emph{enabled}.
4047 @itemx -fno-peephole2
4048 @opindex fno-peephole
4049 @opindex fno-peephole2
4050 Disable any machine-specific peephole optimizations. The difference
4051 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4052 are implemented in the compiler; some targets use one, some use the
4053 other, a few use both.
4055 @option{-fpeephole} is enabled by default.
4056 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4058 @item -fno-guess-branch-probability
4059 @opindex fno-guess-branch-probability
4060 Do not guess branch probabilities using a randomized model.
4062 Sometimes gcc will opt to use a randomized model to guess branch
4063 probabilities, when none are available from either profiling feedback
4064 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4065 different runs of the compiler on the same program may produce different
4068 In a hard real-time system, people don't want different runs of the
4069 compiler to produce code that has different behavior; minimizing
4070 non-determinism is of paramount import. This switch allows users to
4071 reduce non-determinism, possibly at the expense of inferior
4074 The default is @option{-fguess-branch-probability} at levels
4075 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4077 @item -freorder-blocks
4078 @opindex freorder-blocks
4079 Reorder basic blocks in the compiled function in order to reduce number of
4080 taken branches and improve code locality.
4082 Enabled at levels @option{-O2}, @option{-O3}.
4084 @item -freorder-functions
4085 @opindex freorder-functions
4086 Reorder basic blocks in the compiled function in order to reduce number of
4087 taken branches and improve code locality. This is implemented by using special
4088 subsections @code{text.hot} for most frequently executed functions and
4089 @code{text.unlikely} for unlikely executed functions. Reordering is done by
4090 the linker so object file format must support named sections and linker must
4091 place them in a reasonable way.
4093 Also profile feedback must be available in to make this option effective. See
4094 @option{-fprofile-arcs} for details.
4096 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4098 @item -fstrict-aliasing
4099 @opindex fstrict-aliasing
4100 Allows the compiler to assume the strictest aliasing rules applicable to
4101 the language being compiled. For C (and C++), this activates
4102 optimizations based on the type of expressions. In particular, an
4103 object of one type is assumed never to reside at the same address as an
4104 object of a different type, unless the types are almost the same. For
4105 example, an @code{unsigned int} can alias an @code{int}, but not a
4106 @code{void*} or a @code{double}. A character type may alias any other
4109 Pay special attention to code like this:
4122 The practice of reading from a different union member than the one most
4123 recently written to (called ``type-punning'') is common. Even with
4124 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4125 is accessed through the union type. So, the code above will work as
4126 expected. However, this code might not:
4137 Every language that wishes to perform language-specific alias analysis
4138 should define a function that computes, given an @code{tree}
4139 node, an alias set for the node. Nodes in different alias sets are not
4140 allowed to alias. For an example, see the C front-end function
4141 @code{c_get_alias_set}.
4143 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4145 @item -falign-functions
4146 @itemx -falign-functions=@var{n}
4147 @opindex falign-functions
4148 Align the start of functions to the next power-of-two greater than
4149 @var{n}, skipping up to @var{n} bytes. For instance,
4150 @option{-falign-functions=32} aligns functions to the next 32-byte
4151 boundary, but @option{-falign-functions=24} would align to the next
4152 32-byte boundary only if this can be done by skipping 23 bytes or less.
4154 @option{-fno-align-functions} and @option{-falign-functions=1} are
4155 equivalent and mean that functions will not be aligned.
4157 Some assemblers only support this flag when @var{n} is a power of two;
4158 in that case, it is rounded up.
4160 If @var{n} is not specified or is zero, use a machine-dependent default.
4162 Enabled at levels @option{-O2}, @option{-O3}.
4164 @item -falign-labels
4165 @itemx -falign-labels=@var{n}
4166 @opindex falign-labels
4167 Align all branch targets to a power-of-two boundary, skipping up to
4168 @var{n} bytes like @option{-falign-functions}. This option can easily
4169 make code slower, because it must insert dummy operations for when the
4170 branch target is reached in the usual flow of the code.
4172 @option{-fno-align-labels} and @option{-falign-labels=1} are
4173 equivalent and mean that labels will not be aligned.
4175 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4176 are greater than this value, then their values are used instead.
4178 If @var{n} is not specified or is zero, use a machine-dependent default
4179 which is very likely to be @samp{1}, meaning no alignment.
4181 Enabled at levels @option{-O2}, @option{-O3}.
4184 @itemx -falign-loops=@var{n}
4185 @opindex falign-loops
4186 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4187 like @option{-falign-functions}. The hope is that the loop will be
4188 executed many times, which will make up for any execution of the dummy
4191 @option{-fno-align-loops} and @option{-falign-loops=1} are
4192 equivalent and mean that loops will not be aligned.
4194 If @var{n} is not specified or is zero, use a machine-dependent default.
4196 Enabled at levels @option{-O2}, @option{-O3}.
4199 @itemx -falign-jumps=@var{n}
4200 @opindex falign-jumps
4201 Align branch targets to a power-of-two boundary, for branch targets
4202 where the targets can only be reached by jumping, skipping up to @var{n}
4203 bytes like @option{-falign-functions}. In this case, no dummy operations
4206 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4207 equivalent and mean that loops will not be aligned.
4209 If @var{n} is not specified or is zero, use a machine-dependent default.
4211 Enabled at levels @option{-O2}, @option{-O3}.
4213 @item -frename-registers
4214 @opindex frename-registers
4215 Attempt to avoid false dependencies in scheduled code by making use
4216 of registers left over after register allocation. This optimization
4217 will most benefit processors with lots of registers. It can, however,
4218 make debugging impossible, since variables will no longer stay in
4219 a ``home register''.
4221 Enabled at levels @option{-O3}.
4223 @item -fno-cprop-registers
4224 @opindex fno-cprop-registers
4225 After register allocation and post-register allocation instruction splitting,
4226 we perform a copy-propagation pass to try to reduce scheduling dependencies
4227 and occasionally eliminate the copy.
4229 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4233 The following options control compiler behavior regarding floating
4234 point arithmetic. These options trade off between speed and
4235 correctness. All must be specifically enabled.
4239 @opindex ffloat-store
4240 Do not store floating point variables in registers, and inhibit other
4241 options that might change whether a floating point value is taken from a
4244 @cindex floating point precision
4245 This option prevents undesirable excess precision on machines such as
4246 the 68000 where the floating registers (of the 68881) keep more
4247 precision than a @code{double} is supposed to have. Similarly for the
4248 x86 architecture. For most programs, the excess precision does only
4249 good, but a few programs rely on the precise definition of IEEE floating
4250 point. Use @option{-ffloat-store} for such programs, after modifying
4251 them to store all pertinent intermediate computations into variables.
4255 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4256 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4257 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4259 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4261 This option should never be turned on by any @option{-O} option since
4262 it can result in incorrect output for programs which depend on
4263 an exact implementation of IEEE or ISO rules/specifications for
4266 @item -fno-math-errno
4267 @opindex fno-math-errno
4268 Do not set ERRNO after calling math functions that are executed
4269 with a single instruction, e.g., sqrt. A program that relies on
4270 IEEE exceptions for math error handling may want to use this flag
4271 for speed while maintaining IEEE arithmetic compatibility.
4273 This option should never be turned on by any @option{-O} option since
4274 it can result in incorrect output for programs which depend on
4275 an exact implementation of IEEE or ISO rules/specifications for
4278 The default is @option{-fmath-errno}.
4280 @item -funsafe-math-optimizations
4281 @opindex funsafe-math-optimizations
4282 Allow optimizations for floating-point arithmetic that (a) assume
4283 that arguments and results are valid and (b) may violate IEEE or
4284 ANSI standards. When used at link-time, it may include libraries
4285 or startup files that change the default FPU control word or other
4286 similar optimizations.
4288 This option should never be turned on by any @option{-O} option since
4289 it can result in incorrect output for programs which depend on
4290 an exact implementation of IEEE or ISO rules/specifications for
4293 The default is @option{-fno-unsafe-math-optimizations}.
4295 @item -ffinite-math-only
4296 @opindex ffinite-math-only
4297 Allow optimizations for floating-point arithmetic that assume
4298 that arguments and results are not NaNs or +-Infs.
4300 This option should never be turned on by any @option{-O} option since
4301 it can result in incorrect output for programs which depend on
4302 an exact implementation of IEEE or ISO rules/specifications.
4304 The default is @option{-fno-finite-math-only}.
4306 @item -fno-trapping-math
4307 @opindex fno-trapping-math
4308 Compile code assuming that floating-point operations cannot generate
4309 user-visible traps. These traps include division by zero, overflow,
4310 underflow, inexact result and invalid operation. This option implies
4311 @option{-fno-signaling-nans}. Setting this option may allow faster
4312 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4314 This option should never be turned on by any @option{-O} option since
4315 it can result in incorrect output for programs which depend on
4316 an exact implementation of IEEE or ISO rules/specifications for
4319 The default is @option{-ftrapping-math}.
4321 @item -frounding-math
4322 @opindex frounding-math
4323 Disable transformations and optimizations that assume default floating
4324 point rounding behavior. This is round-to-zero for all floating point
4325 to integer conversions, and round-to-nearest for all other arithmetic
4326 truncations. This option should be specified for programs that change
4327 the FP rounding mode dynamically, or that may be executed with a
4328 non-default rounding mode. This option disables constant folding of
4329 floating point expressions at compile-time (which may be affected by
4330 rounding mode) and arithmetic transformations that are unsafe in the
4331 presence of sign-dependent rounding modes.
4333 The default is @option{-fno-rounding-math}.
4335 This option is experimental and does not currently guarantee to
4336 disable all GCC optimizations that are affected by rounding mode.
4337 Future versions of gcc may provide finer control of this setting
4338 using C99's @code{FENV_ACCESS} pragma. This command line option
4339 will be used to specify the default state for @code{FENV_ACCESS}.
4341 @item -fsignaling-nans
4342 @opindex fsignaling-nans
4343 Compile code assuming that IEEE signaling NaNs may generate user-visible
4344 traps during floating-point operations. Setting this option disables
4345 optimizations that may change the number of exceptions visible with
4346 signaling NaNs. This option implies @option{-ftrapping-math}.
4348 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4351 The default is @option{-fno-signaling-nans}.
4353 This option is experimental and does not currently guarantee to
4354 disable all GCC optimizations that affect signaling NaN behavior.
4356 @item -fsingle-precision-constant
4357 @opindex fsingle-precision-constant
4358 Treat floating point constant as single precision constant instead of
4359 implicitly converting it to double precision constant.
4364 The following options control optimizations that may improve
4365 performance, but are not enabled by any @option{-O} options. This
4366 section includes experimental options that may produce broken code.
4369 @item -fbranch-probabilities
4370 @opindex fbranch-probabilities
4371 After running a program compiled with @option{-fprofile-arcs}
4372 (@pxref{Debugging Options,, Options for Debugging Your Program or
4373 @command{gcc}}), you can compile it a second time using
4374 @option{-fbranch-probabilities}, to improve optimizations based on
4375 the number of times each branch was taken. When the program
4376 compiled with @option{-fprofile-arcs} exits it saves arc execution
4377 counts to a file called @file{@var{sourcename}.gcda} for each source
4378 file The information in this data file is very dependent on the
4379 structure of the generated code, so you must use the same source code
4380 and the same optimization options for both compilations.
4382 With @option{-fbranch-probabilities}, GCC puts a
4383 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4384 These can be used to improve optimization. Currently, they are only
4385 used in one place: in @file{reorg.c}, instead of guessing which path a
4386 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4387 exactly determine which path is taken more often.
4389 @item -fprofile-values
4390 @opindex fprofile-values
4391 If combined with @option{-fprofile-arcs}, it adds code so that some
4392 data about values of expressions in the program is gathered.
4394 With @option{-fbranch-probabilities}, it reads back the data gathered
4395 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4396 notes to instructions for their later usage in optimizations.
4400 Use a graph coloring register allocator. Currently this option is meant
4401 for testing, so we are interested to hear about miscompilations with
4406 Perform tail duplication to enlarge superblock size. This transformation
4407 simplifies the control flow of the function allowing other optimizations to do
4410 @item -funit-at-a-time
4411 @opindex funit-at-a-time
4412 Parse the whole compilation unit before starting to produce code.
4413 This allows some extra optimizations to take place but consumes more
4416 @item -funroll-loops
4417 @opindex funroll-loops
4418 Unroll loops whose number of iterations can be determined at compile time or
4419 upon entry to the loop. @option{-funroll-loops} implies
4420 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4421 (i.e. complete removal of loops with small constant number of iterations).
4422 This option makes code larger, and may or may not make it run faster.
4424 @item -funroll-all-loops
4425 @opindex funroll-all-loops
4426 Unroll all loops, even if their number of iterations is uncertain when
4427 the loop is entered. This usually makes programs run more slowly.
4428 @option{-funroll-all-loops} implies the same options as
4429 @option{-funroll-loops}.
4432 @opindex fpeel-loops
4433 Peels the loops for that there is enough information that they do not
4434 roll much (from profile feedback). It also turns on complete loop peeling
4435 (i.e. complete removal of loops with small constant number of iterations).
4437 @item -funswitch-loops
4438 @opindex funswitch-loops
4439 Move branches with loop invariant conditions out of the loop, with duplicates
4440 of the loop on both branches (modified according to result of the condition).
4442 @item -fold-unroll-loops
4443 @opindex fold-unroll-loops
4444 Unroll loops whose number of iterations can be determined at compile
4445 time or upon entry to the loop, using the old loop unroller whose loop
4446 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4447 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4448 option makes code larger, and may or may not make it run faster.
4450 @item -fold-unroll-all-loops
4451 @opindex fold-unroll-all-loops
4452 Unroll all loops, even if their number of iterations is uncertain when
4453 the loop is entered. This is done using the old loop unroller whose loop
4454 recognition is based on notes from frontend. This usually makes programs run more slowly.
4455 @option{-fold-unroll-all-loops} implies the same options as
4456 @option{-fold-unroll-loops}.
4458 @item -funswitch-loops
4459 @opindex funswitch-loops
4460 Move branches with loop invariant conditions out of the loop, with duplicates
4461 of the loop on both branches (modified according to result of the condition).
4463 @item -funswitch-loops
4464 @opindex funswitch-loops
4465 Move branches with loop invariant conditions out of the loop, with duplicates
4466 of the loop on both branches (modified according to result of the condition).
4468 @item -fprefetch-loop-arrays
4469 @opindex fprefetch-loop-arrays
4470 If supported by the target machine, generate instructions to prefetch
4471 memory to improve the performance of loops that access large arrays.
4473 Disabled at level @option{-Os}.
4475 @item -ffunction-sections
4476 @itemx -fdata-sections
4477 @opindex ffunction-sections
4478 @opindex fdata-sections
4479 Place each function or data item into its own section in the output
4480 file if the target supports arbitrary sections. The name of the
4481 function or the name of the data item determines the section's name
4484 Use these options on systems where the linker can perform optimizations
4485 to improve locality of reference in the instruction space. Most systems
4486 using the ELF object format and SPARC processors running Solaris 2 have
4487 linkers with such optimizations. AIX may have these optimizations in
4490 Only use these options when there are significant benefits from doing
4491 so. When you specify these options, the assembler and linker will
4492 create larger object and executable files and will also be slower.
4493 You will not be able to use @code{gprof} on all systems if you
4494 specify this option and you may have problems with debugging if
4495 you specify both this option and @option{-g}.
4499 Perform optimizations in static single assignment form. Each function's
4500 flow graph is translated into SSA form, optimizations are performed, and
4501 the flow graph is translated back from SSA form. Users should not
4502 specify this option, since it is not yet ready for production use.
4506 Perform Sparse Conditional Constant Propagation in SSA form. Requires
4507 @option{-fssa}. Like @option{-fssa}, this is an experimental feature.
4511 Perform aggressive dead-code elimination in SSA form. Requires @option{-fssa}.
4512 Like @option{-fssa}, this is an experimental feature.
4514 @item -fbranch-target-load-optimize
4515 @opindex fbranch-target-load-optimize
4516 Perform branch target register load optimization before prologue / epilogue
4518 The use of target registers can typically be exposed only during reload,
4519 thus hoisting loads out of loops and doing inter-block scheduling needs
4520 a separate optimization pass.
4522 @item -fbranch-target-load-optimize2
4523 @opindex fbranch-target-load-optimize2
4524 Perform branch target register load optimization after prologue / epilogue
4530 @item --param @var{name}=@var{value}
4532 In some places, GCC uses various constants to control the amount of
4533 optimization that is done. For example, GCC will not inline functions
4534 that contain more that a certain number of instructions. You can
4535 control some of these constants on the command-line using the
4536 @option{--param} option.
4538 In each case, the @var{value} is an integer. The allowable choices for
4539 @var{name} are given in the following table:
4542 @item max-crossjump-edges
4543 The maximum number of incoming edges to consider for crossjumping.
4544 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
4545 the number of edges incoming to each block. Increasing values mean
4546 more aggressive optimization, making the compile time increase with
4547 probably small improvement in executable size.
4549 @item max-delay-slot-insn-search
4550 The maximum number of instructions to consider when looking for an
4551 instruction to fill a delay slot. If more than this arbitrary number of
4552 instructions is searched, the time savings from filling the delay slot
4553 will be minimal so stop searching. Increasing values mean more
4554 aggressive optimization, making the compile time increase with probably
4555 small improvement in executable run time.
4557 @item max-delay-slot-live-search
4558 When trying to fill delay slots, the maximum number of instructions to
4559 consider when searching for a block with valid live register
4560 information. Increasing this arbitrarily chosen value means more
4561 aggressive optimization, increasing the compile time. This parameter
4562 should be removed when the delay slot code is rewritten to maintain the
4565 @item max-gcse-memory
4566 The approximate maximum amount of memory that will be allocated in
4567 order to perform the global common subexpression elimination
4568 optimization. If more memory than specified is required, the
4569 optimization will not be done.
4571 @item max-gcse-passes
4572 The maximum number of passes of GCSE to run.
4574 @item max-pending-list-length
4575 The maximum number of pending dependencies scheduling will allow
4576 before flushing the current state and starting over. Large functions
4577 with few branches or calls can create excessively large lists which
4578 needlessly consume memory and resources.
4580 @item max-inline-insns-single
4581 Several parameters control the tree inliner used in gcc.
4582 This number sets the maximum number of instructions (counted in gcc's
4583 internal representation) in a single function that the tree inliner
4584 will consider for inlining. This only affects functions declared
4585 inline and methods implemented in a class declaration (C++).
4586 The default value is 500.
4588 @item max-inline-insns-auto
4589 When you use @option{-finline-functions} (included in @option{-O3}),
4590 a lot of functions that would otherwise not be considered for inlining
4591 by the compiler will be investigated. To those functions, a different
4592 (more restrictive) limit compared to functions declared inline can
4594 The default value is 150.
4596 @item max-inline-insns
4597 The tree inliner does decrease the allowable size for single functions
4598 to be inlined after we already inlined the number of instructions
4599 given here by repeated inlining. This number should be a factor of
4600 two or more larger than the single function limit.
4601 Higher numbers result in better runtime performance, but incur higher
4602 compile-time resource (CPU time, memory) requirements and result in
4603 larger binaries. Very high values are not advisable, as too large
4604 binaries may adversely affect runtime performance.
4605 The default value is 200.
4607 @item max-inline-slope
4608 After exceeding the maximum number of inlined instructions by repeated
4609 inlining, a linear function is used to decrease the allowable size
4610 for single functions. The slope of that function is the negative
4611 reciprocal of the number specified here.
4612 This parameter is ignored when @option{-funit-at-a-time} is used.
4613 The default value is 32.
4615 @item min-inline-insns
4616 The repeated inlining is throttled more and more by the linear function
4617 after exceeding the limit. To avoid too much throttling, a minimum for
4618 this function is specified here to allow repeated inlining for very small
4619 functions even when a lot of repeated inlining already has been done.
4620 This parameter is ignored when @option{-funit-at-a-time} is used.
4621 The default value is 10.
4623 @item large-function-insns
4624 The limit specifying really large functions. For functions greater than this
4625 limit inlining is constrained by @option{--param large-function-growth}.
4626 This parameter is usefull primarily to avoid extreme compilation time caused by non-linear
4627 algorithms used by the backend.
4628 This parameter is ignored when @option{-funit-at-a-time} is not used.
4629 The default value is 30000.
4631 @item large-function-growth
4632 Specifies maximal growth of large functtion caused by inlining in percents.
4633 This parameter is ignored when @option{-funit-at-a-time} is not used.
4634 The default value is 200.
4636 @item inline-unit-growth
4637 Specifies maximal overall growth of the compilation unit caused by inlining.
4638 This parameter is ignored when @option{-funit-at-a-time} is not used.
4639 The default value is 150.
4641 @item max-inline-insns-rtl
4642 For languages that use the RTL inliner (this happens at a later stage
4643 than tree inlining), you can set the maximum allowable size (counted
4644 in RTL instructions) for the RTL inliner with this parameter.
4645 The default value is 600.
4648 @item max-unrolled-insns
4649 The maximum number of instructions that a loop should have if that loop
4650 is unrolled, and if the loop is unrolled, it determines how many times
4651 the loop code is unrolled.
4653 @item max-average-unrolled-insns
4654 The maximum number of instructions biased by probabilities of their execution
4655 that a loop should have if that loop is unrolled, and if the loop is unrolled,
4656 it determines how many times the loop code is unrolled.
4658 @item max-unroll-times
4659 The maximum number of unrollings of a single loop.
4661 @item max-peeled-insns
4662 The maximum number of instructions that a loop should have if that loop
4663 is peeled, and if the loop is peeled, it determines how many times
4664 the loop code is peeled.
4666 @item max-peel-times
4667 The maximum number of peelings of a single loop.
4669 @item max-completely-peeled-insns
4670 The maximum number of insns of a completely peeled loop.
4672 @item max-completely-peel-times
4673 The maximum number of iterations of a loop to be suitable for complete peeling.
4675 @item max-unswitch-insns
4676 The maximum number of insns of an unswitched loop.
4678 @item max-unswitch-level
4679 The maximum number of branches unswitched in a single loop.
4681 @item hot-bb-count-fraction
4682 Select fraction of the maximal count of repetitions of basic block in program
4683 given basic block needs to have to be considered hot.
4685 @item hot-bb-frequency-fraction
4686 Select fraction of the maximal frequency of executions of basic block in
4687 function given basic block needs to have to be considered hot
4689 @item tracer-dynamic-coverage
4690 @itemx tracer-dynamic-coverage-feedback
4692 This value is used to limit superblock formation once the given percentage of
4693 executed instructions is covered. This limits unnecessary code size
4696 The @option{tracer-dynamic-coverage-feedback} is used only when profile
4697 feedback is available. The real profiles (as opposed to statically estimated
4698 ones) are much less balanced allowing the threshold to be larger value.
4700 @item tracer-max-code-growth
4701 Stop tail duplication once code growth has reached given percentage. This is
4702 rather hokey argument, as most of the duplicates will be eliminated later in
4703 cross jumping, so it may be set to much higher values than is the desired code
4706 @item tracer-min-branch-ratio
4708 Stop reverse growth when the reverse probability of best edge is less than this
4709 threshold (in percent).
4711 @item tracer-min-branch-ratio
4712 @itemx tracer-min-branch-ratio-feedback
4714 Stop forward growth if the best edge do have probability lower than this
4717 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
4718 compilation for profile feedback and one for compilation without. The value
4719 for compilation with profile feedback needs to be more conservative (higher) in
4720 order to make tracer effective.
4722 @item max-cse-path-length
4724 Maximum number of basic blocks on path that cse considers.
4726 @item ggc-min-expand
4728 GCC uses a garbage collector to manage its own memory allocation. This
4729 parameter specifies the minimum percentage by which the garbage
4730 collector's heap should be allowed to expand between collections.
4731 Tuning this may improve compilation speed; it has no effect on code
4734 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
4735 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
4736 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
4737 GCC is not able to calculate RAM on a particular platform, the lower
4738 bound of 30% is used. Setting this parameter and
4739 @option{ggc-min-heapsize} to zero causes a full collection to occur at
4740 every opportunity. This is extremely slow, but can be useful for
4743 @item ggc-min-heapsize
4745 Minimum size of the garbage collector's heap before it begins bothering
4746 to collect garbage. The first collection occurs after the heap expands
4747 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
4748 tuning this may improve compilation speed, and has no effect on code
4751 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
4752 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
4753 available, the notion of "RAM" is the smallest of actual RAM,
4754 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
4755 RAM on a particular platform, the lower bound is used. Setting this
4756 parameter very large effectively disables garbage collection. Setting
4757 this parameter and @option{ggc-min-expand} to zero causes a full
4758 collection to occur at every opportunity.
4760 @item reorder-blocks-duplicate
4761 @itemx reorder-blocks-duplicate-feedback
4763 Used by basic block reordering pass to decide whether to use unconditional
4764 branch or duplicate the code on its destination. Code is duplicated when its
4765 estimated size is smaller than this value multiplied by the estimated size of
4766 unconditional jump in the hot spots of the program.
4768 The @option{reorder-block-duplicate-feedback} is used only when profile
4769 feedback is available and may be set to higher values than
4770 @option{reorder-block-duplicate} since information about the hot spots is more
4775 @node Preprocessor Options
4776 @section Options Controlling the Preprocessor
4777 @cindex preprocessor options
4778 @cindex options, preprocessor
4780 These options control the C preprocessor, which is run on each C source
4781 file before actual compilation.
4783 If you use the @option{-E} option, nothing is done except preprocessing.
4784 Some of these options make sense only together with @option{-E} because
4785 they cause the preprocessor output to be unsuitable for actual
4790 You can use @option{-Wp,@var{option}} to bypass the compiler driver
4791 and pass @var{option} directly through to the preprocessor. If
4792 @var{option} contains commas, it is split into multiple options at the
4793 commas. However, many options are modified, translated or interpreted
4794 by the compiler driver before being passed to the preprocessor, and
4795 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
4796 interface is undocumented and subject to change, so whenever possible
4797 you should avoid using @option{-Wp} and let the driver handle the
4800 @item -Xpreprocessor @var{option}
4801 @opindex preprocessor
4802 Pass @var{option} as an option to the preprocessor. You can use this to
4803 supply system-specific preprocessor options which GCC does not know how to
4806 If you want to pass an option that takes an argument, you must use
4807 @option{-Xpreprocessor} twice, once for the option and once for the argument.
4810 @include cppopts.texi
4812 @node Assembler Options
4813 @section Passing Options to the Assembler
4815 @c prevent bad page break with this line
4816 You can pass options to the assembler.
4819 @item -Wa,@var{option}
4821 Pass @var{option} as an option to the assembler. If @var{option}
4822 contains commas, it is split into multiple options at the commas.
4824 @item -Xassembler @var{option}
4826 Pass @var{option} as an option to the assembler. You can use this to
4827 supply system-specific assembler options which GCC does not know how to
4830 If you want to pass an option that takes an argument, you must use
4831 @option{-Xassembler} twice, once for the option and once for the argument.
4836 @section Options for Linking
4837 @cindex link options
4838 @cindex options, linking
4840 These options come into play when the compiler links object files into
4841 an executable output file. They are meaningless if the compiler is
4842 not doing a link step.
4846 @item @var{object-file-name}
4847 A file name that does not end in a special recognized suffix is
4848 considered to name an object file or library. (Object files are
4849 distinguished from libraries by the linker according to the file
4850 contents.) If linking is done, these object files are used as input
4859 If any of these options is used, then the linker is not run, and
4860 object file names should not be used as arguments. @xref{Overall
4864 @item -l@var{library}
4865 @itemx -l @var{library}
4867 Search the library named @var{library} when linking. (The second
4868 alternative with the library as a separate argument is only for
4869 POSIX compliance and is not recommended.)
4871 It makes a difference where in the command you write this option; the
4872 linker searches and processes libraries and object files in the order they
4873 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
4874 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
4875 to functions in @samp{z}, those functions may not be loaded.
4877 The linker searches a standard list of directories for the library,
4878 which is actually a file named @file{lib@var{library}.a}. The linker
4879 then uses this file as if it had been specified precisely by name.
4881 The directories searched include several standard system directories
4882 plus any that you specify with @option{-L}.
4884 Normally the files found this way are library files---archive files
4885 whose members are object files. The linker handles an archive file by
4886 scanning through it for members which define symbols that have so far
4887 been referenced but not defined. But if the file that is found is an
4888 ordinary object file, it is linked in the usual fashion. The only
4889 difference between using an @option{-l} option and specifying a file name
4890 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
4891 and searches several directories.
4895 You need this special case of the @option{-l} option in order to
4896 link an Objective-C program.
4899 @opindex nostartfiles
4900 Do not use the standard system startup files when linking.
4901 The standard system libraries are used normally, unless @option{-nostdlib}
4902 or @option{-nodefaultlibs} is used.
4904 @item -nodefaultlibs
4905 @opindex nodefaultlibs
4906 Do not use the standard system libraries when linking.
4907 Only the libraries you specify will be passed to the linker.
4908 The standard startup files are used normally, unless @option{-nostartfiles}
4909 is used. The compiler may generate calls to memcmp, memset, and memcpy
4910 for System V (and ISO C) environments or to bcopy and bzero for
4911 BSD environments. These entries are usually resolved by entries in
4912 libc. These entry points should be supplied through some other
4913 mechanism when this option is specified.
4917 Do not use the standard system startup files or libraries when linking.
4918 No startup files and only the libraries you specify will be passed to
4919 the linker. The compiler may generate calls to memcmp, memset, and memcpy
4920 for System V (and ISO C) environments or to bcopy and bzero for
4921 BSD environments. These entries are usually resolved by entries in
4922 libc. These entry points should be supplied through some other
4923 mechanism when this option is specified.
4925 @cindex @option{-lgcc}, use with @option{-nostdlib}
4926 @cindex @option{-nostdlib} and unresolved references
4927 @cindex unresolved references and @option{-nostdlib}
4928 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
4929 @cindex @option{-nodefaultlibs} and unresolved references
4930 @cindex unresolved references and @option{-nodefaultlibs}
4931 One of the standard libraries bypassed by @option{-nostdlib} and
4932 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
4933 that GCC uses to overcome shortcomings of particular machines, or special
4934 needs for some languages.
4935 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
4936 Collection (GCC) Internals},
4937 for more discussion of @file{libgcc.a}.)
4938 In most cases, you need @file{libgcc.a} even when you want to avoid
4939 other standard libraries. In other words, when you specify @option{-nostdlib}
4940 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
4941 This ensures that you have no unresolved references to internal GCC
4942 library subroutines. (For example, @samp{__main}, used to ensure C++
4943 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
4944 GNU Compiler Collection (GCC) Internals}.)
4948 Produce a position independent executable on targets which support it.
4949 For predictable results, you must also specify the same set of options
4950 that were used to generate code (@option{-fpie}, @option{-fPIE},
4951 or model suboptions) when you specify this option.
4955 Remove all symbol table and relocation information from the executable.
4959 On systems that support dynamic linking, this prevents linking with the shared
4960 libraries. On other systems, this option has no effect.
4964 Produce a shared object which can then be linked with other objects to
4965 form an executable. Not all systems support this option. For predictable
4966 results, you must also specify the same set of options that were used to
4967 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
4968 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
4969 needs to build supplementary stub code for constructors to work. On
4970 multi-libbed systems, @samp{gcc -shared} must select the correct support
4971 libraries to link against. Failing to supply the correct flags may lead
4972 to subtle defects. Supplying them in cases where they are not necessary
4975 @item -shared-libgcc
4976 @itemx -static-libgcc
4977 @opindex shared-libgcc
4978 @opindex static-libgcc
4979 On systems that provide @file{libgcc} as a shared library, these options
4980 force the use of either the shared or static version respectively.
4981 If no shared version of @file{libgcc} was built when the compiler was
4982 configured, these options have no effect.
4984 There are several situations in which an application should use the
4985 shared @file{libgcc} instead of the static version. The most common
4986 of these is when the application wishes to throw and catch exceptions
4987 across different shared libraries. In that case, each of the libraries
4988 as well as the application itself should use the shared @file{libgcc}.
4990 Therefore, the G++ and GCJ drivers automatically add
4991 @option{-shared-libgcc} whenever you build a shared library or a main
4992 executable, because C++ and Java programs typically use exceptions, so
4993 this is the right thing to do.
4995 If, instead, you use the GCC driver to create shared libraries, you may
4996 find that they will not always be linked with the shared @file{libgcc}.
4997 If GCC finds, at its configuration time, that you have a GNU linker that
4998 does not support option @option{--eh-frame-hdr}, it will link the shared
4999 version of @file{libgcc} into shared libraries by default. Otherwise,
5000 it will take advantage of the linker and optimize away the linking with
5001 the shared version of @file{libgcc}, linking with the static version of
5002 libgcc by default. This allows exceptions to propagate through such
5003 shared libraries, without incurring relocation costs at library load
5006 However, if a library or main executable is supposed to throw or catch
5007 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5008 for the languages used in the program, or using the option
5009 @option{-shared-libgcc}, such that it is linked with the shared
5014 Bind references to global symbols when building a shared object. Warn
5015 about any unresolved references (unless overridden by the link editor
5016 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5019 @item -Xlinker @var{option}
5021 Pass @var{option} as an option to the linker. You can use this to
5022 supply system-specific linker options which GCC does not know how to
5025 If you want to pass an option that takes an argument, you must use
5026 @option{-Xlinker} twice, once for the option and once for the argument.
5027 For example, to pass @option{-assert definitions}, you must write
5028 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5029 @option{-Xlinker "-assert definitions"}, because this passes the entire
5030 string as a single argument, which is not what the linker expects.
5032 @item -Wl,@var{option}
5034 Pass @var{option} as an option to the linker. If @var{option} contains
5035 commas, it is split into multiple options at the commas.
5037 @item -u @var{symbol}
5039 Pretend the symbol @var{symbol} is undefined, to force linking of
5040 library modules to define it. You can use @option{-u} multiple times with
5041 different symbols to force loading of additional library modules.
5044 @node Directory Options
5045 @section Options for Directory Search
5046 @cindex directory options
5047 @cindex options, directory search
5050 These options specify directories to search for header files, for
5051 libraries and for parts of the compiler:
5056 Add the directory @var{dir} to the head of the list of directories to be
5057 searched for header files. This can be used to override a system header
5058 file, substituting your own version, since these directories are
5059 searched before the system header file directories. However, you should
5060 not use this option to add directories that contain vendor-supplied
5061 system header files (use @option{-isystem} for that). If you use more than
5062 one @option{-I} option, the directories are scanned in left-to-right
5063 order; the standard system directories come after.
5065 If a standard system include directory, or a directory specified with
5066 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5067 option will be ignored. The directory will still be searched but as a
5068 system directory at its normal position in the system include chain.
5069 This is to ensure that GCC's procedure to fix buggy system headers and
5070 the ordering for the include_next directive are not inadvertently changed.
5071 If you really need to change the search order for system directories,
5072 use the @option{-nostdinc} and/or @option{-isystem} options.
5076 Any directories you specify with @option{-I} options before the @option{-I-}
5077 option are searched only for the case of @samp{#include "@var{file}"};
5078 they are not searched for @samp{#include <@var{file}>}.
5080 If additional directories are specified with @option{-I} options after
5081 the @option{-I-}, these directories are searched for all @samp{#include}
5082 directives. (Ordinarily @emph{all} @option{-I} directories are used
5085 In addition, the @option{-I-} option inhibits the use of the current
5086 directory (where the current input file came from) as the first search
5087 directory for @samp{#include "@var{file}"}. There is no way to
5088 override this effect of @option{-I-}. With @option{-I.} you can specify
5089 searching the directory which was current when the compiler was
5090 invoked. That is not exactly the same as what the preprocessor does
5091 by default, but it is often satisfactory.
5093 @option{-I-} does not inhibit the use of the standard system directories
5094 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5099 Add directory @var{dir} to the list of directories to be searched
5102 @item -B@var{prefix}
5104 This option specifies where to find the executables, libraries,
5105 include files, and data files of the compiler itself.
5107 The compiler driver program runs one or more of the subprograms
5108 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5109 @var{prefix} as a prefix for each program it tries to run, both with and
5110 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5112 For each subprogram to be run, the compiler driver first tries the
5113 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5114 was not specified, the driver tries two standard prefixes, which are
5115 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5116 those results in a file name that is found, the unmodified program
5117 name is searched for using the directories specified in your
5118 @env{PATH} environment variable.
5120 The compiler will check to see if the path provided by the @option{-B}
5121 refers to a directory, and if necessary it will add a directory
5122 separator character at the end of the path.
5124 @option{-B} prefixes that effectively specify directory names also apply
5125 to libraries in the linker, because the compiler translates these
5126 options into @option{-L} options for the linker. They also apply to
5127 includes files in the preprocessor, because the compiler translates these
5128 options into @option{-isystem} options for the preprocessor. In this case,
5129 the compiler appends @samp{include} to the prefix.
5131 The run-time support file @file{libgcc.a} can also be searched for using
5132 the @option{-B} prefix, if needed. If it is not found there, the two
5133 standard prefixes above are tried, and that is all. The file is left
5134 out of the link if it is not found by those means.
5136 Another way to specify a prefix much like the @option{-B} prefix is to use
5137 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5140 As a special kludge, if the path provided by @option{-B} is
5141 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5142 9, then it will be replaced by @file{[dir/]include}. This is to help
5143 with boot-strapping the compiler.
5145 @item -specs=@var{file}
5147 Process @var{file} after the compiler reads in the standard @file{specs}
5148 file, in order to override the defaults that the @file{gcc} driver
5149 program uses when determining what switches to pass to @file{cc1},
5150 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5151 @option{-specs=@var{file}} can be specified on the command line, and they
5152 are processed in order, from left to right.
5158 @section Specifying subprocesses and the switches to pass to them
5161 @command{gcc} is a driver program. It performs its job by invoking a
5162 sequence of other programs to do the work of compiling, assembling and
5163 linking. GCC interprets its command-line parameters and uses these to
5164 deduce which programs it should invoke, and which command-line options
5165 it ought to place on their command lines. This behavior is controlled
5166 by @dfn{spec strings}. In most cases there is one spec string for each
5167 program that GCC can invoke, but a few programs have multiple spec
5168 strings to control their behavior. The spec strings built into GCC can
5169 be overridden by using the @option{-specs=} command-line switch to specify
5172 @dfn{Spec files} are plaintext files that are used to construct spec
5173 strings. They consist of a sequence of directives separated by blank
5174 lines. The type of directive is determined by the first non-whitespace
5175 character on the line and it can be one of the following:
5178 @item %@var{command}
5179 Issues a @var{command} to the spec file processor. The commands that can
5183 @item %include <@var{file}>
5185 Search for @var{file} and insert its text at the current point in the
5188 @item %include_noerr <@var{file}>
5189 @cindex %include_noerr
5190 Just like @samp{%include}, but do not generate an error message if the include
5191 file cannot be found.
5193 @item %rename @var{old_name} @var{new_name}
5195 Rename the spec string @var{old_name} to @var{new_name}.
5199 @item *[@var{spec_name}]:
5200 This tells the compiler to create, override or delete the named spec
5201 string. All lines after this directive up to the next directive or
5202 blank line are considered to be the text for the spec string. If this
5203 results in an empty string then the spec will be deleted. (Or, if the
5204 spec did not exist, then nothing will happened.) Otherwise, if the spec
5205 does not currently exist a new spec will be created. If the spec does
5206 exist then its contents will be overridden by the text of this
5207 directive, unless the first character of that text is the @samp{+}
5208 character, in which case the text will be appended to the spec.
5210 @item [@var{suffix}]:
5211 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5212 and up to the next directive or blank line are considered to make up the
5213 spec string for the indicated suffix. When the compiler encounters an
5214 input file with the named suffix, it will processes the spec string in
5215 order to work out how to compile that file. For example:
5222 This says that any input file whose name ends in @samp{.ZZ} should be
5223 passed to the program @samp{z-compile}, which should be invoked with the
5224 command-line switch @option{-input} and with the result of performing the
5225 @samp{%i} substitution. (See below.)
5227 As an alternative to providing a spec string, the text that follows a
5228 suffix directive can be one of the following:
5231 @item @@@var{language}
5232 This says that the suffix is an alias for a known @var{language}. This is
5233 similar to using the @option{-x} command-line switch to GCC to specify a
5234 language explicitly. For example:
5241 Says that .ZZ files are, in fact, C++ source files.
5244 This causes an error messages saying:
5247 @var{name} compiler not installed on this system.
5251 GCC already has an extensive list of suffixes built into it.
5252 This directive will add an entry to the end of the list of suffixes, but
5253 since the list is searched from the end backwards, it is effectively
5254 possible to override earlier entries using this technique.
5258 GCC has the following spec strings built into it. Spec files can
5259 override these strings or create their own. Note that individual
5260 targets can also add their own spec strings to this list.
5263 asm Options to pass to the assembler
5264 asm_final Options to pass to the assembler post-processor
5265 cpp Options to pass to the C preprocessor
5266 cc1 Options to pass to the C compiler
5267 cc1plus Options to pass to the C++ compiler
5268 endfile Object files to include at the end of the link
5269 link Options to pass to the linker
5270 lib Libraries to include on the command line to the linker
5271 libgcc Decides which GCC support library to pass to the linker
5272 linker Sets the name of the linker
5273 predefines Defines to be passed to the C preprocessor
5274 signed_char Defines to pass to CPP to say whether @code{char} is signed
5276 startfile Object files to include at the start of the link
5279 Here is a small example of a spec file:
5285 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5288 This example renames the spec called @samp{lib} to @samp{old_lib} and
5289 then overrides the previous definition of @samp{lib} with a new one.
5290 The new definition adds in some extra command-line options before
5291 including the text of the old definition.
5293 @dfn{Spec strings} are a list of command-line options to be passed to their
5294 corresponding program. In addition, the spec strings can contain
5295 @samp{%}-prefixed sequences to substitute variable text or to
5296 conditionally insert text into the command line. Using these constructs
5297 it is possible to generate quite complex command lines.
5299 Here is a table of all defined @samp{%}-sequences for spec
5300 strings. Note that spaces are not generated automatically around the
5301 results of expanding these sequences. Therefore you can concatenate them
5302 together or combine them with constant text in a single argument.
5306 Substitute one @samp{%} into the program name or argument.
5309 Substitute the name of the input file being processed.
5312 Substitute the basename of the input file being processed.
5313 This is the substring up to (and not including) the last period
5314 and not including the directory.
5317 This is the same as @samp{%b}, but include the file suffix (text after
5321 Marks the argument containing or following the @samp{%d} as a
5322 temporary file name, so that that file will be deleted if GCC exits
5323 successfully. Unlike @samp{%g}, this contributes no text to the
5326 @item %g@var{suffix}
5327 Substitute a file name that has suffix @var{suffix} and is chosen
5328 once per compilation, and mark the argument in the same way as
5329 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5330 name is now chosen in a way that is hard to predict even when previously
5331 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5332 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5333 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5334 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5335 was simply substituted with a file name chosen once per compilation,
5336 without regard to any appended suffix (which was therefore treated
5337 just like ordinary text), making such attacks more likely to succeed.
5339 @item %u@var{suffix}
5340 Like @samp{%g}, but generates a new temporary file name even if
5341 @samp{%u@var{suffix}} was already seen.
5343 @item %U@var{suffix}
5344 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5345 new one if there is no such last file name. In the absence of any
5346 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5347 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5348 would involve the generation of two distinct file names, one
5349 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5350 simply substituted with a file name chosen for the previous @samp{%u},
5351 without regard to any appended suffix.
5353 @item %j@var{suffix}
5354 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5355 writable, and if save-temps is off; otherwise, substitute the name
5356 of a temporary file, just like @samp{%u}. This temporary file is not
5357 meant for communication between processes, but rather as a junk
5360 @item %|@var{suffix}
5361 @itemx %m@var{suffix}
5362 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5363 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5364 all. These are the two most common ways to instruct a program that it
5365 should read from standard input or write to standard output. If you
5366 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5367 construct: see for example @file{f/lang-specs.h}.
5369 @item %.@var{SUFFIX}
5370 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5371 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5372 terminated by the next space or %.
5375 Marks the argument containing or following the @samp{%w} as the
5376 designated output file of this compilation. This puts the argument
5377 into the sequence of arguments that @samp{%o} will substitute later.
5380 Substitutes the names of all the output files, with spaces
5381 automatically placed around them. You should write spaces
5382 around the @samp{%o} as well or the results are undefined.
5383 @samp{%o} is for use in the specs for running the linker.
5384 Input files whose names have no recognized suffix are not compiled
5385 at all, but they are included among the output files, so they will
5389 Substitutes the suffix for object files. Note that this is
5390 handled specially when it immediately follows @samp{%g, %u, or %U},
5391 because of the need for those to form complete file names. The
5392 handling is such that @samp{%O} is treated exactly as if it had already
5393 been substituted, except that @samp{%g, %u, and %U} do not currently
5394 support additional @var{suffix} characters following @samp{%O} as they would
5395 following, for example, @samp{.o}.
5398 Substitutes the standard macro predefinitions for the
5399 current target machine. Use this when running @code{cpp}.
5402 Like @samp{%p}, but puts @samp{__} before and after the name of each
5403 predefined macro, except for macros that start with @samp{__} or with
5404 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5408 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5409 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5410 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5414 Current argument is the name of a library or startup file of some sort.
5415 Search for that file in a standard list of directories and substitute
5416 the full name found.
5419 Print @var{str} as an error message. @var{str} is terminated by a newline.
5420 Use this when inconsistent options are detected.
5423 Substitute the contents of spec string @var{name} at this point.
5426 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
5428 @item %x@{@var{option}@}
5429 Accumulate an option for @samp{%X}.
5432 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
5436 Output the accumulated assembler options specified by @option{-Wa}.
5439 Output the accumulated preprocessor options specified by @option{-Wp}.
5442 Process the @code{asm} spec. This is used to compute the
5443 switches to be passed to the assembler.
5446 Process the @code{asm_final} spec. This is a spec string for
5447 passing switches to an assembler post-processor, if such a program is
5451 Process the @code{link} spec. This is the spec for computing the
5452 command line passed to the linker. Typically it will make use of the
5453 @samp{%L %G %S %D and %E} sequences.
5456 Dump out a @option{-L} option for each directory that GCC believes might
5457 contain startup files. If the target supports multilibs then the
5458 current multilib directory will be prepended to each of these paths.
5461 Output the multilib directory with directory separators replaced with
5462 @samp{_}. If multilib directories are not set, or the multilib directory is
5463 @file{.} then this option emits nothing.
5466 Process the @code{lib} spec. This is a spec string for deciding which
5467 libraries should be included on the command line to the linker.
5470 Process the @code{libgcc} spec. This is a spec string for deciding
5471 which GCC support library should be included on the command line to the linker.
5474 Process the @code{startfile} spec. This is a spec for deciding which
5475 object files should be the first ones passed to the linker. Typically
5476 this might be a file named @file{crt0.o}.
5479 Process the @code{endfile} spec. This is a spec string that specifies
5480 the last object files that will be passed to the linker.
5483 Process the @code{cpp} spec. This is used to construct the arguments
5484 to be passed to the C preprocessor.
5487 Process the @code{signed_char} spec. This is intended to be used
5488 to tell cpp whether a char is signed. It typically has the definition:
5490 %@{funsigned-char:-D__CHAR_UNSIGNED__@}
5494 Process the @code{cc1} spec. This is used to construct the options to be
5495 passed to the actual C compiler (@samp{cc1}).
5498 Process the @code{cc1plus} spec. This is used to construct the options to be
5499 passed to the actual C++ compiler (@samp{cc1plus}).
5502 Substitute the variable part of a matched option. See below.
5503 Note that each comma in the substituted string is replaced by
5507 Remove all occurrences of @code{-S} from the command line. Note---this
5508 command is position dependent. @samp{%} commands in the spec string
5509 before this one will see @code{-S}, @samp{%} commands in the spec string
5510 after this one will not.
5512 @item %:@var{function}(@var{args})
5513 Call the named function @var{function}, passing it @var{args}.
5514 @var{args} is first processed as a nested spec string, then split
5515 into an argument vector in the usual fashion. The function returns
5516 a string which is processed as if it had appeared literally as part
5517 of the current spec.
5519 The following built-in spec functions are provided:
5522 @item @code{if-exists}
5523 The @code{if-exists} spec function takes one argument, an absolute
5524 pathname to a file. If the file exists, @code{if-exists} returns the
5525 pathname. Here is a small example of its usage:
5529 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
5532 @item @code{if-exists-else}
5533 The @code{if-exists-else} spec function is similar to the @code{if-exists}
5534 spec function, except that it takes two arguments. The first argument is
5535 an absolute pathname to a file. If the file exists, @code{if-exists-else}
5536 returns the pathname. If it does not exist, it returns the second argument.
5537 This way, @code{if-exists-else} can be used to select one file or another,
5538 based on the existence of the first. Here is a small example of its usage:
5542 crt0%O%s %:if-exists(crti%O%s) \
5543 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
5548 Substitutes the @code{-S} switch, if that switch was given to GCC@.
5549 If that switch was not specified, this substitutes nothing. Note that
5550 the leading dash is omitted when specifying this option, and it is
5551 automatically inserted if the substitution is performed. Thus the spec
5552 string @samp{%@{foo@}} would match the command-line option @option{-foo}
5553 and would output the command line option @option{-foo}.
5555 @item %W@{@code{S}@}
5556 Like %@{@code{S}@} but mark last argument supplied within as a file to be
5559 @item %@{@code{S}*@}
5560 Substitutes all the switches specified to GCC whose names start
5561 with @code{-S}, but which also take an argument. This is used for
5562 switches like @option{-o}, @option{-D}, @option{-I}, etc.
5563 GCC considers @option{-o foo} as being
5564 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
5565 text, including the space. Thus two arguments would be generated.
5567 @item %@{@code{S}*&@code{T}*@}
5568 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
5569 (the order of @code{S} and @code{T} in the spec is not significant).
5570 There can be any number of ampersand-separated variables; for each the
5571 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
5573 @item %@{@code{S}:@code{X}@}
5574 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
5576 @item %@{!@code{S}:@code{X}@}
5577 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
5579 @item %@{@code{S}*:@code{X}@}
5580 Substitutes @code{X} if one or more switches whose names start with
5581 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
5582 once, no matter how many such switches appeared. However, if @code{%*}
5583 appears somewhere in @code{X}, then @code{X} will be substituted once
5584 for each matching switch, with the @code{%*} replaced by the part of
5585 that switch that matched the @code{*}.
5587 @item %@{.@code{S}:@code{X}@}
5588 Substitutes @code{X}, if processing a file with suffix @code{S}.
5590 @item %@{!.@code{S}:@code{X}@}
5591 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
5593 @item %@{@code{S}|@code{P}:@code{X}@}
5594 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
5595 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
5596 although they have a stronger binding than the @samp{|}. If @code{%*}
5597 appears in @code{X}, all of the alternatives must be starred, and only
5598 the first matching alternative is substituted.
5600 For example, a spec string like this:
5603 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
5606 will output the following command-line options from the following input
5607 command-line options:
5612 -d fred.c -foo -baz -boggle
5613 -d jim.d -bar -baz -boggle
5616 @item %@{S:X; T:Y; :D@}
5618 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
5619 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
5620 be as many clauses as you need. This may be combined with @code{.},
5621 @code{!}, @code{|}, and @code{*} as needed.
5626 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
5627 construct may contain other nested @samp{%} constructs or spaces, or
5628 even newlines. They are processed as usual, as described above.
5629 Trailing white space in @code{X} is ignored. White space may also
5630 appear anywhere on the left side of the colon in these constructs,
5631 except between @code{.} or @code{*} and the corresponding word.
5633 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
5634 handled specifically in these constructs. If another value of
5635 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
5636 @option{-W} switch is found later in the command line, the earlier
5637 switch value is ignored, except with @{@code{S}*@} where @code{S} is
5638 just one letter, which passes all matching options.
5640 The character @samp{|} at the beginning of the predicate text is used to
5641 indicate that a command should be piped to the following command, but
5642 only if @option{-pipe} is specified.
5644 It is built into GCC which switches take arguments and which do not.
5645 (You might think it would be useful to generalize this to allow each
5646 compiler's spec to say which switches take arguments. But this cannot
5647 be done in a consistent fashion. GCC cannot even decide which input
5648 files have been specified without knowing which switches take arguments,
5649 and it must know which input files to compile in order to tell which
5652 GCC also knows implicitly that arguments starting in @option{-l} are to be
5653 treated as compiler output files, and passed to the linker in their
5654 proper position among the other output files.
5656 @c man begin OPTIONS
5658 @node Target Options
5659 @section Specifying Target Machine and Compiler Version
5660 @cindex target options
5661 @cindex cross compiling
5662 @cindex specifying machine version
5663 @cindex specifying compiler version and target machine
5664 @cindex compiler version, specifying
5665 @cindex target machine, specifying
5667 The usual way to run GCC is to run the executable called @file{gcc}, or
5668 @file{<machine>-gcc} when cross-compiling, or
5669 @file{<machine>-gcc-<version>} to run a version other than the one that
5670 was installed last. Sometimes this is inconvenient, so GCC provides
5671 options that will switch to another cross-compiler or version.
5674 @item -b @var{machine}
5676 The argument @var{machine} specifies the target machine for compilation.
5678 The value to use for @var{machine} is the same as was specified as the
5679 machine type when configuring GCC as a cross-compiler. For
5680 example, if a cross-compiler was configured with @samp{configure
5681 i386v}, meaning to compile for an 80386 running System V, then you
5682 would specify @option{-b i386v} to run that cross compiler.
5684 @item -V @var{version}
5686 The argument @var{version} specifies which version of GCC to run.
5687 This is useful when multiple versions are installed. For example,
5688 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
5691 The @option{-V} and @option{-b} options work by running the
5692 @file{<machine>-gcc-<version>} executable, so there's no real reason to
5693 use them if you can just run that directly.
5695 @node Submodel Options
5696 @section Hardware Models and Configurations
5697 @cindex submodel options
5698 @cindex specifying hardware config
5699 @cindex hardware models and configurations, specifying
5700 @cindex machine dependent options
5702 Earlier we discussed the standard option @option{-b} which chooses among
5703 different installed compilers for completely different target
5704 machines, such as VAX vs.@: 68000 vs.@: 80386.
5706 In addition, each of these target machine types can have its own
5707 special options, starting with @samp{-m}, to choose among various
5708 hardware models or configurations---for example, 68010 vs 68020,
5709 floating coprocessor or none. A single installed version of the
5710 compiler can compile for any model or configuration, according to the
5713 Some configurations of the compiler also support additional special
5714 options, usually for compatibility with other compilers on the same
5717 These options are defined by the macro @code{TARGET_SWITCHES} in the
5718 machine description. The default for the options is also defined by
5719 that macro, which enables you to change the defaults.
5731 * RS/6000 and PowerPC Options::
5735 * i386 and x86-64 Options::
5737 * Intel 960 Options::
5738 * DEC Alpha Options::
5739 * DEC Alpha/VMS Options::
5742 * System V Options::
5743 * TMS320C3x/C4x Options::
5751 * S/390 and zSeries Options::
5755 * Xstormy16 Options::
5760 @node M680x0 Options
5761 @subsection M680x0 Options
5762 @cindex M680x0 options
5764 These are the @samp{-m} options defined for the 68000 series. The default
5765 values for these options depends on which style of 68000 was selected when
5766 the compiler was configured; the defaults for the most common choices are
5774 Generate output for a 68000. This is the default
5775 when the compiler is configured for 68000-based systems.
5777 Use this option for microcontrollers with a 68000 or EC000 core,
5778 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
5784 Generate output for a 68020. This is the default
5785 when the compiler is configured for 68020-based systems.
5789 Generate output containing 68881 instructions for floating point.
5790 This is the default for most 68020 systems unless @option{--nfp} was
5791 specified when the compiler was configured.
5795 Generate output for a 68030. This is the default when the compiler is
5796 configured for 68030-based systems.
5800 Generate output for a 68040. This is the default when the compiler is
5801 configured for 68040-based systems.
5803 This option inhibits the use of 68881/68882 instructions that have to be
5804 emulated by software on the 68040. Use this option if your 68040 does not
5805 have code to emulate those instructions.
5809 Generate output for a 68060. This is the default when the compiler is
5810 configured for 68060-based systems.
5812 This option inhibits the use of 68020 and 68881/68882 instructions that
5813 have to be emulated by software on the 68060. Use this option if your 68060
5814 does not have code to emulate those instructions.
5818 Generate output for a CPU32. This is the default
5819 when the compiler is configured for CPU32-based systems.
5821 Use this option for microcontrollers with a
5822 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
5823 68336, 68340, 68341, 68349 and 68360.
5827 Generate output for a 520X ``coldfire'' family cpu. This is the default
5828 when the compiler is configured for 520X-based systems.
5830 Use this option for microcontroller with a 5200 core, including
5831 the MCF5202, MCF5203, MCF5204 and MCF5202.
5836 Generate output for a 68040, without using any of the new instructions.
5837 This results in code which can run relatively efficiently on either a
5838 68020/68881 or a 68030 or a 68040. The generated code does use the
5839 68881 instructions that are emulated on the 68040.
5843 Generate output for a 68060, without using any of the new instructions.
5844 This results in code which can run relatively efficiently on either a
5845 68020/68881 or a 68030 or a 68040. The generated code does use the
5846 68881 instructions that are emulated on the 68060.
5849 @opindex msoft-float
5850 Generate output containing library calls for floating point.
5851 @strong{Warning:} the requisite libraries are not available for all m68k
5852 targets. Normally the facilities of the machine's usual C compiler are
5853 used, but this can't be done directly in cross-compilation. You must
5854 make your own arrangements to provide suitable library functions for
5855 cross-compilation. The embedded targets @samp{m68k-*-aout} and
5856 @samp{m68k-*-coff} do provide software floating point support.
5860 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5863 @opindex mnobitfield
5864 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
5865 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
5869 Do use the bit-field instructions. The @option{-m68020} option implies
5870 @option{-mbitfield}. This is the default if you use a configuration
5871 designed for a 68020.
5875 Use a different function-calling convention, in which functions
5876 that take a fixed number of arguments return with the @code{rtd}
5877 instruction, which pops their arguments while returning. This
5878 saves one instruction in the caller since there is no need to pop
5879 the arguments there.
5881 This calling convention is incompatible with the one normally
5882 used on Unix, so you cannot use it if you need to call libraries
5883 compiled with the Unix compiler.
5885 Also, you must provide function prototypes for all functions that
5886 take variable numbers of arguments (including @code{printf});
5887 otherwise incorrect code will be generated for calls to those
5890 In addition, seriously incorrect code will result if you call a
5891 function with too many arguments. (Normally, extra arguments are
5892 harmlessly ignored.)
5894 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
5895 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
5898 @itemx -mno-align-int
5900 @opindex mno-align-int
5901 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
5902 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
5903 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
5904 Aligning variables on 32-bit boundaries produces code that runs somewhat
5905 faster on processors with 32-bit busses at the expense of more memory.
5907 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
5908 align structures containing the above types differently than
5909 most published application binary interface specifications for the m68k.
5913 Use the pc-relative addressing mode of the 68000 directly, instead of
5914 using a global offset table. At present, this option implies @option{-fpic},
5915 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
5916 not presently supported with @option{-mpcrel}, though this could be supported for
5917 68020 and higher processors.
5919 @item -mno-strict-align
5920 @itemx -mstrict-align
5921 @opindex mno-strict-align
5922 @opindex mstrict-align
5923 Do not (do) assume that unaligned memory references will be handled by
5928 @node M68hc1x Options
5929 @subsection M68hc1x Options
5930 @cindex M68hc1x options
5932 These are the @samp{-m} options defined for the 68hc11 and 68hc12
5933 microcontrollers. The default values for these options depends on
5934 which style of microcontroller was selected when the compiler was configured;
5935 the defaults for the most common choices are given below.
5942 Generate output for a 68HC11. This is the default
5943 when the compiler is configured for 68HC11-based systems.
5949 Generate output for a 68HC12. This is the default
5950 when the compiler is configured for 68HC12-based systems.
5956 Generate output for a 68HCS12.
5959 @opindex mauto-incdec
5960 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
5967 Enable the use of 68HC12 min and max instructions.
5970 @itemx -mno-long-calls
5971 @opindex mlong-calls
5972 @opindex mno-long-calls
5973 Treat all calls as being far away (near). If calls are assumed to be
5974 far away, the compiler will use the @code{call} instruction to
5975 call a function and the @code{rtc} instruction for returning.
5979 Consider type @code{int} to be 16 bits wide, like @code{short int}.
5981 @item -msoft-reg-count=@var{count}
5982 @opindex msoft-reg-count
5983 Specify the number of pseudo-soft registers which are used for the
5984 code generation. The maximum number is 32. Using more pseudo-soft
5985 register may or may not result in better code depending on the program.
5986 The default is 4 for 68HC11 and 2 for 68HC12.
5991 @subsection VAX Options
5994 These @samp{-m} options are defined for the VAX:
5999 Do not output certain jump instructions (@code{aobleq} and so on)
6000 that the Unix assembler for the VAX cannot handle across long
6005 Do output those jump instructions, on the assumption that you
6006 will assemble with the GNU assembler.
6010 Output code for g-format floating point numbers instead of d-format.
6014 @subsection SPARC Options
6015 @cindex SPARC options
6017 These @samp{-m} switches are supported on the SPARC:
6022 @opindex mno-app-regs
6024 Specify @option{-mapp-regs} to generate output using the global registers
6025 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
6028 To be fully SVR4 ABI compliant at the cost of some performance loss,
6029 specify @option{-mno-app-regs}. You should compile libraries and system
6030 software with this option.
6035 @opindex mhard-float
6036 Generate output containing floating point instructions. This is the
6042 @opindex msoft-float
6043 Generate output containing library calls for floating point.
6044 @strong{Warning:} the requisite libraries are not available for all SPARC
6045 targets. Normally the facilities of the machine's usual C compiler are
6046 used, but this cannot be done directly in cross-compilation. You must make
6047 your own arrangements to provide suitable library functions for
6048 cross-compilation. The embedded targets @samp{sparc-*-aout} and
6049 @samp{sparclite-*-*} do provide software floating point support.
6051 @option{-msoft-float} changes the calling convention in the output file;
6052 therefore, it is only useful if you compile @emph{all} of a program with
6053 this option. In particular, you need to compile @file{libgcc.a}, the
6054 library that comes with GCC, with @option{-msoft-float} in order for
6057 @item -mhard-quad-float
6058 @opindex mhard-quad-float
6059 Generate output containing quad-word (long double) floating point
6062 @item -msoft-quad-float
6063 @opindex msoft-quad-float
6064 Generate output containing library calls for quad-word (long double)
6065 floating point instructions. The functions called are those specified
6066 in the SPARC ABI@. This is the default.
6068 As of this writing, there are no sparc implementations that have hardware
6069 support for the quad-word floating point instructions. They all invoke
6070 a trap handler for one of these instructions, and then the trap handler
6071 emulates the effect of the instruction. Because of the trap handler overhead,
6072 this is much slower than calling the ABI library routines. Thus the
6073 @option{-msoft-quad-float} option is the default.
6079 With @option{-mflat}, the compiler does not generate save/restore instructions
6080 and will use a ``flat'' or single register window calling convention.
6081 This model uses %i7 as the frame pointer and is compatible with the normal
6082 register window model. Code from either may be intermixed.
6083 The local registers and the input registers (0--5) are still treated as
6084 ``call saved'' registers and will be saved on the stack as necessary.
6086 With @option{-mno-flat} (the default), the compiler emits save/restore
6087 instructions (except for leaf functions) and is the normal mode of operation.
6089 @item -mno-unaligned-doubles
6090 @itemx -munaligned-doubles
6091 @opindex mno-unaligned-doubles
6092 @opindex munaligned-doubles
6093 Assume that doubles have 8 byte alignment. This is the default.
6095 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
6096 alignment only if they are contained in another type, or if they have an
6097 absolute address. Otherwise, it assumes they have 4 byte alignment.
6098 Specifying this option avoids some rare compatibility problems with code
6099 generated by other compilers. It is not the default because it results
6100 in a performance loss, especially for floating point code.
6102 @item -mno-faster-structs
6103 @itemx -mfaster-structs
6104 @opindex mno-faster-structs
6105 @opindex mfaster-structs
6106 With @option{-mfaster-structs}, the compiler assumes that structures
6107 should have 8 byte alignment. This enables the use of pairs of
6108 @code{ldd} and @code{std} instructions for copies in structure
6109 assignment, in place of twice as many @code{ld} and @code{st} pairs.
6110 However, the use of this changed alignment directly violates the SPARC
6111 ABI@. Thus, it's intended only for use on targets where the developer
6112 acknowledges that their resulting code will not be directly in line with
6113 the rules of the ABI@.
6116 @opindex mimpure-text
6117 @option{-mimpure-text}, used in addition to @option{-shared}, tells
6118 the compiler to not pass @option{-z text} to the linker when linking a
6119 shared object. Using this option, you can link position-dependent
6120 code into a shared object.
6122 @option{-mimpure-text} suppresses the ``relocations remain against
6123 allocatable but non-writable sections'' linker error message.
6124 However, the necessary relocations will trigger copy-on-write, and the
6125 shared object is not actually shared across processes. Instead of
6126 using @option{-mimpure-text}, you should compile all source code with
6127 @option{-fpic} or @option{-fPIC}.
6129 This option is only available on SunOS and Solaris.
6135 These two options select variations on the SPARC architecture.
6137 By default (unless specifically configured for the Fujitsu SPARClite),
6138 GCC generates code for the v7 variant of the SPARC architecture.
6140 @option{-mv8} will give you SPARC v8 code. The only difference from v7
6141 code is that the compiler emits the integer multiply and integer
6142 divide instructions which exist in SPARC v8 but not in SPARC v7.
6144 @option{-msparclite} will give you SPARClite code. This adds the integer
6145 multiply, integer divide step and scan (@code{ffs}) instructions which
6146 exist in SPARClite but not in SPARC v7.
6148 These options are deprecated and will be deleted in a future GCC release.
6149 They have been replaced with @option{-mcpu=xxx}.
6154 @opindex msupersparc
6155 These two options select the processor for which the code is optimized.
6157 With @option{-mcypress} (the default), the compiler optimizes code for the
6158 Cypress CY7C602 chip, as used in the SPARCStation/SPARCServer 3xx series.
6159 This is also appropriate for the older SPARCStation 1, 2, IPX etc.
6161 With @option{-msupersparc} the compiler optimizes code for the SuperSPARC cpu, as
6162 used in the SPARCStation 10, 1000 and 2000 series. This flag also enables use
6163 of the full SPARC v8 instruction set.
6165 These options are deprecated and will be deleted in a future GCC release.
6166 They have been replaced with @option{-mcpu=xxx}.
6168 @item -mcpu=@var{cpu_type}
6170 Set the instruction set, register set, and instruction scheduling parameters
6171 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
6172 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
6173 @samp{hypersparc}, @samp{sparclite86x}, @samp{f930}, @samp{f934},
6174 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
6177 Default instruction scheduling parameters are used for values that select
6178 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
6179 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
6181 Here is a list of each supported architecture and their supported
6186 v8: supersparc, hypersparc
6187 sparclite: f930, f934, sparclite86x
6189 v9: ultrasparc, ultrasparc3
6192 @item -mtune=@var{cpu_type}
6194 Set the instruction scheduling parameters for machine type
6195 @var{cpu_type}, but do not set the instruction set or register set that the
6196 option @option{-mcpu=@var{cpu_type}} would.
6198 The same values for @option{-mcpu=@var{cpu_type}} can be used for
6199 @option{-mtune=@var{cpu_type}}, but the only useful values are those
6200 that select a particular cpu implementation. Those are @samp{cypress},
6201 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
6202 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
6207 These @samp{-m} switches are supported in addition to the above
6208 on the SPARCLET processor.
6211 @item -mlittle-endian
6212 @opindex mlittle-endian
6213 Generate code for a processor running in little-endian mode.
6217 Treat register @code{%g0} as a normal register.
6218 GCC will continue to clobber it as necessary but will not assume
6219 it always reads as 0.
6221 @item -mbroken-saverestore
6222 @opindex mbroken-saverestore
6223 Generate code that does not use non-trivial forms of the @code{save} and
6224 @code{restore} instructions. Early versions of the SPARCLET processor do
6225 not correctly handle @code{save} and @code{restore} instructions used with
6226 arguments. They correctly handle them used without arguments. A @code{save}
6227 instruction used without arguments increments the current window pointer
6228 but does not allocate a new stack frame. It is assumed that the window
6229 overflow trap handler will properly handle this case as will interrupt
6233 These @samp{-m} switches are supported in addition to the above
6234 on SPARC V9 processors in 64-bit environments.
6237 @item -mlittle-endian
6238 @opindex mlittle-endian
6239 Generate code for a processor running in little-endian mode. It is only
6240 available for a few configurations and most notably not on Solaris.
6246 Generate code for a 32-bit or 64-bit environment.
6247 The 32-bit environment sets int, long and pointer to 32 bits.
6248 The 64-bit environment sets int to 32 bits and long and pointer
6251 @item -mcmodel=medlow
6252 @opindex mcmodel=medlow
6253 Generate code for the Medium/Low code model: the program must be linked
6254 in the low 32 bits of the address space. Pointers are 64 bits.
6255 Programs can be statically or dynamically linked.
6257 @item -mcmodel=medmid
6258 @opindex mcmodel=medmid
6259 Generate code for the Medium/Middle code model: the program must be linked
6260 in the low 44 bits of the address space, the text segment must be less than
6261 2G bytes, and data segment must be within 2G of the text segment.
6262 Pointers are 64 bits.
6264 @item -mcmodel=medany
6265 @opindex mcmodel=medany
6266 Generate code for the Medium/Anywhere code model: the program may be linked
6267 anywhere in the address space, the text segment must be less than
6268 2G bytes, and data segment must be within 2G of the text segment.
6269 Pointers are 64 bits.
6271 @item -mcmodel=embmedany
6272 @opindex mcmodel=embmedany
6273 Generate code for the Medium/Anywhere code model for embedded systems:
6274 assume a 32-bit text and a 32-bit data segment, both starting anywhere
6275 (determined at link time). Register %g4 points to the base of the
6276 data segment. Pointers are still 64 bits.
6277 Programs are statically linked, PIC is not supported.
6280 @itemx -mno-stack-bias
6281 @opindex mstack-bias
6282 @opindex mno-stack-bias
6283 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
6284 frame pointer if present, are offset by @minus{}2047 which must be added back
6285 when making stack frame references.
6286 Otherwise, assume no such offset is present.
6290 @subsection ARM Options
6293 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6298 @opindex mapcs-frame
6299 Generate a stack frame that is compliant with the ARM Procedure Call
6300 Standard for all functions, even if this is not strictly necessary for
6301 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6302 with this option will cause the stack frames not to be generated for
6303 leaf functions. The default is @option{-mno-apcs-frame}.
6307 This is a synonym for @option{-mapcs-frame}.
6311 Generate code for a processor running with a 26-bit program counter,
6312 and conforming to the function calling standards for the APCS 26-bit
6313 option. This option replaces the @option{-m2} and @option{-m3} options
6314 of previous releases of the compiler.
6318 Generate code for a processor running with a 32-bit program counter,
6319 and conforming to the function calling standards for the APCS 32-bit
6320 option. This option replaces the @option{-m6} option of previous releases
6324 @c not currently implemented
6325 @item -mapcs-stack-check
6326 @opindex mapcs-stack-check
6327 Generate code to check the amount of stack space available upon entry to
6328 every function (that actually uses some stack space). If there is
6329 insufficient space available then either the function
6330 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6331 called, depending upon the amount of stack space required. The run time
6332 system is required to provide these functions. The default is
6333 @option{-mno-apcs-stack-check}, since this produces smaller code.
6335 @c not currently implemented
6337 @opindex mapcs-float
6338 Pass floating point arguments using the float point registers. This is
6339 one of the variants of the APCS@. This option is recommended if the
6340 target hardware has a floating point unit or if a lot of floating point
6341 arithmetic is going to be performed by the code. The default is
6342 @option{-mno-apcs-float}, since integer only code is slightly increased in
6343 size if @option{-mapcs-float} is used.
6345 @c not currently implemented
6346 @item -mapcs-reentrant
6347 @opindex mapcs-reentrant
6348 Generate reentrant, position independent code. The default is
6349 @option{-mno-apcs-reentrant}.
6352 @item -mthumb-interwork
6353 @opindex mthumb-interwork
6354 Generate code which supports calling between the ARM and Thumb
6355 instruction sets. Without this option the two instruction sets cannot
6356 be reliably used inside one program. The default is
6357 @option{-mno-thumb-interwork}, since slightly larger code is generated
6358 when @option{-mthumb-interwork} is specified.
6360 @item -mno-sched-prolog
6361 @opindex mno-sched-prolog
6362 Prevent the reordering of instructions in the function prolog, or the
6363 merging of those instruction with the instructions in the function's
6364 body. This means that all functions will start with a recognizable set
6365 of instructions (or in fact one of a choice from a small set of
6366 different function prologues), and this information can be used to
6367 locate the start if functions inside an executable piece of code. The
6368 default is @option{-msched-prolog}.
6371 @opindex mhard-float
6372 Generate output containing floating point instructions. This is the
6376 @opindex msoft-float
6377 Generate output containing library calls for floating point.
6378 @strong{Warning:} the requisite libraries are not available for all ARM
6379 targets. Normally the facilities of the machine's usual C compiler are
6380 used, but this cannot be done directly in cross-compilation. You must make
6381 your own arrangements to provide suitable library functions for
6384 @option{-msoft-float} changes the calling convention in the output file;
6385 therefore, it is only useful if you compile @emph{all} of a program with
6386 this option. In particular, you need to compile @file{libgcc.a}, the
6387 library that comes with GCC, with @option{-msoft-float} in order for
6390 @item -mlittle-endian
6391 @opindex mlittle-endian
6392 Generate code for a processor running in little-endian mode. This is
6393 the default for all standard configurations.
6396 @opindex mbig-endian
6397 Generate code for a processor running in big-endian mode; the default is
6398 to compile code for a little-endian processor.
6400 @item -mwords-little-endian
6401 @opindex mwords-little-endian
6402 This option only applies when generating code for big-endian processors.
6403 Generate code for a little-endian word order but a big-endian byte
6404 order. That is, a byte order of the form @samp{32107654}. Note: this
6405 option should only be used if you require compatibility with code for
6406 big-endian ARM processors generated by versions of the compiler prior to
6409 @item -malignment-traps
6410 @opindex malignment-traps
6411 Generate code that will not trap if the MMU has alignment traps enabled.
6412 On ARM architectures prior to ARMv4, there were no instructions to
6413 access half-word objects stored in memory. However, when reading from
6414 memory a feature of the ARM architecture allows a word load to be used,
6415 even if the address is unaligned, and the processor core will rotate the
6416 data as it is being loaded. This option tells the compiler that such
6417 misaligned accesses will cause a MMU trap and that it should instead
6418 synthesize the access as a series of byte accesses. The compiler can
6419 still use word accesses to load half-word data if it knows that the
6420 address is aligned to a word boundary.
6422 This option is ignored when compiling for ARM architecture 4 or later,
6423 since these processors have instructions to directly access half-word
6426 @item -mno-alignment-traps
6427 @opindex mno-alignment-traps
6428 Generate code that assumes that the MMU will not trap unaligned
6429 accesses. This produces better code when the target instruction set
6430 does not have half-word memory operations (i.e.@: implementations prior to
6433 Note that you cannot use this option to access unaligned word objects,
6434 since the processor will only fetch one 32-bit aligned object from
6437 The default setting for most targets is @option{-mno-alignment-traps}, since
6438 this produces better code when there are no half-word memory
6439 instructions available.
6441 @item -mshort-load-bytes
6442 @itemx -mno-short-load-words
6443 @opindex mshort-load-bytes
6444 @opindex mno-short-load-words
6445 These are deprecated aliases for @option{-malignment-traps}.
6447 @item -mno-short-load-bytes
6448 @itemx -mshort-load-words
6449 @opindex mno-short-load-bytes
6450 @opindex mshort-load-words
6451 This are deprecated aliases for @option{-mno-alignment-traps}.
6453 @item -mcpu=@var{name}
6455 This specifies the name of the target ARM processor. GCC uses this name
6456 to determine what kind of instructions it can emit when generating
6457 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6458 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6459 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6460 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6461 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6462 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6463 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6464 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6465 @samp{arm920t}, @samp{arm940t}, @samp{arm9tdmi}, @samp{arm10tdmi},
6466 @samp{arm1020t}, @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
6468 @itemx -mtune=@var{name}
6470 This option is very similar to the @option{-mcpu=} option, except that
6471 instead of specifying the actual target processor type, and hence
6472 restricting which instructions can be used, it specifies that GCC should
6473 tune the performance of the code as if the target were of the type
6474 specified in this option, but still choosing the instructions that it
6475 will generate based on the cpu specified by a @option{-mcpu=} option.
6476 For some ARM implementations better performance can be obtained by using
6479 @item -march=@var{name}
6481 This specifies the name of the target ARM architecture. GCC uses this
6482 name to determine what kind of instructions it can emit when generating
6483 assembly code. This option can be used in conjunction with or instead
6484 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6485 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6486 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{iwmmxt}, @samp{ep9312}.
6488 @item -mfpe=@var{number}
6489 @itemx -mfp=@var{number}
6492 This specifies the version of the floating point emulation available on
6493 the target. Permissible values are 2 and 3. @option{-mfp=} is a synonym
6494 for @option{-mfpe=}, for compatibility with older versions of GCC@.
6496 @item -mstructure-size-boundary=@var{n}
6497 @opindex mstructure-size-boundary
6498 The size of all structures and unions will be rounded up to a multiple
6499 of the number of bits set by this option. Permissible values are 8 and
6500 32. The default value varies for different toolchains. For the COFF
6501 targeted toolchain the default value is 8. Specifying the larger number
6502 can produce faster, more efficient code, but can also increase the size
6503 of the program. The two values are potentially incompatible. Code
6504 compiled with one value cannot necessarily expect to work with code or
6505 libraries compiled with the other value, if they exchange information
6506 using structures or unions.
6508 @item -mabort-on-noreturn
6509 @opindex mabort-on-noreturn
6510 Generate a call to the function @code{abort} at the end of a
6511 @code{noreturn} function. It will be executed if the function tries to
6515 @itemx -mno-long-calls
6516 @opindex mlong-calls
6517 @opindex mno-long-calls
6518 Tells the compiler to perform function calls by first loading the
6519 address of the function into a register and then performing a subroutine
6520 call on this register. This switch is needed if the target function
6521 will lie outside of the 64 megabyte addressing range of the offset based
6522 version of subroutine call instruction.
6524 Even if this switch is enabled, not all function calls will be turned
6525 into long calls. The heuristic is that static functions, functions
6526 which have the @samp{short-call} attribute, functions that are inside
6527 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6528 definitions have already been compiled within the current compilation
6529 unit, will not be turned into long calls. The exception to this rule is
6530 that weak function definitions, functions with the @samp{long-call}
6531 attribute or the @samp{section} attribute, and functions that are within
6532 the scope of a @samp{#pragma long_calls} directive, will always be
6533 turned into long calls.
6535 This feature is not enabled by default. Specifying
6536 @option{-mno-long-calls} will restore the default behavior, as will
6537 placing the function calls within the scope of a @samp{#pragma
6538 long_calls_off} directive. Note these switches have no effect on how
6539 the compiler generates code to handle function calls via function
6542 @item -mnop-fun-dllimport
6543 @opindex mnop-fun-dllimport
6544 Disable support for the @code{dllimport} attribute.
6546 @item -msingle-pic-base
6547 @opindex msingle-pic-base
6548 Treat the register used for PIC addressing as read-only, rather than
6549 loading it in the prologue for each function. The run-time system is
6550 responsible for initializing this register with an appropriate value
6551 before execution begins.
6553 @item -mpic-register=@var{reg}
6554 @opindex mpic-register
6555 Specify the register to be used for PIC addressing. The default is R10
6556 unless stack-checking is enabled, when R9 is used.
6558 @item -mcirrus-fix-invalid-insns
6559 @opindex mcirrus-fix-invalid-insns
6560 @opindex mno-cirrus-fix-invalid-insns
6561 Insert NOPs into the instruction stream to in order to work around
6562 problems with invalid Maverick instruction combinations. This option
6563 is only valid if the @option{-mcpu=ep9312} option has been used to
6564 enable generation of instructions for the Cirrus Maverick floating
6565 point co-processor. This option is not enabled by default, since the
6566 problem is only present in older Maverick implementations. The default
6567 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6570 @item -mpoke-function-name
6571 @opindex mpoke-function-name
6572 Write the name of each function into the text section, directly
6573 preceding the function prologue. The generated code is similar to this:
6577 .ascii "arm_poke_function_name", 0
6580 .word 0xff000000 + (t1 - t0)
6581 arm_poke_function_name
6583 stmfd sp!, @{fp, ip, lr, pc@}
6587 When performing a stack backtrace, code can inspect the value of
6588 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6589 location @code{pc - 12} and the top 8 bits are set, then we know that
6590 there is a function name embedded immediately preceding this location
6591 and has length @code{((pc[-3]) & 0xff000000)}.
6595 Generate code for the 16-bit Thumb instruction set. The default is to
6596 use the 32-bit ARM instruction set.
6599 @opindex mtpcs-frame
6600 Generate a stack frame that is compliant with the Thumb Procedure Call
6601 Standard for all non-leaf functions. (A leaf function is one that does
6602 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6604 @item -mtpcs-leaf-frame
6605 @opindex mtpcs-leaf-frame
6606 Generate a stack frame that is compliant with the Thumb Procedure Call
6607 Standard for all leaf functions. (A leaf function is one that does
6608 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6610 @item -mcallee-super-interworking
6611 @opindex mcallee-super-interworking
6612 Gives all externally visible functions in the file being compiled an ARM
6613 instruction set header which switches to Thumb mode before executing the
6614 rest of the function. This allows these functions to be called from
6615 non-interworking code.
6617 @item -mcaller-super-interworking
6618 @opindex mcaller-super-interworking
6619 Allows calls via function pointers (including virtual functions) to
6620 execute correctly regardless of whether the target code has been
6621 compiled for interworking or not. There is a small overhead in the cost
6622 of executing a function pointer if this option is enabled.
6626 @node MN10200 Options
6627 @subsection MN10200 Options
6628 @cindex MN10200 options
6630 These @option{-m} options are defined for Matsushita MN10200 architectures:
6635 Indicate to the linker that it should perform a relaxation optimization pass
6636 to shorten branches, calls and absolute memory addresses. This option only
6637 has an effect when used on the command line for the final link step.
6639 This option makes symbolic debugging impossible.
6642 @node MN10300 Options
6643 @subsection MN10300 Options
6644 @cindex MN10300 options
6646 These @option{-m} options are defined for Matsushita MN10300 architectures:
6651 Generate code to avoid bugs in the multiply instructions for the MN10300
6652 processors. This is the default.
6655 @opindex mno-mult-bug
6656 Do not generate code to avoid bugs in the multiply instructions for the
6661 Generate code which uses features specific to the AM33 processor.
6665 Do not generate code which uses features specific to the AM33 processor. This
6670 Do not link in the C run-time initialization object file.
6674 Indicate to the linker that it should perform a relaxation optimization pass
6675 to shorten branches, calls and absolute memory addresses. This option only
6676 has an effect when used on the command line for the final link step.
6678 This option makes symbolic debugging impossible.
6682 @node M32R/D Options
6683 @subsection M32R/D Options
6684 @cindex M32R/D options
6686 These @option{-m} options are defined for Mitsubishi M32R/D architectures:
6691 Generate code for the M32R/X@.
6695 Generate code for the M32R@. This is the default.
6697 @item -mcode-model=small
6698 @opindex mcode-model=small
6699 Assume all objects live in the lower 16MB of memory (so that their addresses
6700 can be loaded with the @code{ld24} instruction), and assume all subroutines
6701 are reachable with the @code{bl} instruction.
6702 This is the default.
6704 The addressability of a particular object can be set with the
6705 @code{model} attribute.
6707 @item -mcode-model=medium
6708 @opindex mcode-model=medium
6709 Assume objects may be anywhere in the 32-bit address space (the compiler
6710 will generate @code{seth/add3} instructions to load their addresses), and
6711 assume all subroutines are reachable with the @code{bl} instruction.
6713 @item -mcode-model=large
6714 @opindex mcode-model=large
6715 Assume objects may be anywhere in the 32-bit address space (the compiler
6716 will generate @code{seth/add3} instructions to load their addresses), and
6717 assume subroutines may not be reachable with the @code{bl} instruction
6718 (the compiler will generate the much slower @code{seth/add3/jl}
6719 instruction sequence).
6722 @opindex msdata=none
6723 Disable use of the small data area. Variables will be put into
6724 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
6725 @code{section} attribute has been specified).
6726 This is the default.
6728 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
6729 Objects may be explicitly put in the small data area with the
6730 @code{section} attribute using one of these sections.
6733 @opindex msdata=sdata
6734 Put small global and static data in the small data area, but do not
6735 generate special code to reference them.
6739 Put small global and static data in the small data area, and generate
6740 special instructions to reference them.
6744 @cindex smaller data references
6745 Put global and static objects less than or equal to @var{num} bytes
6746 into the small data or bss sections instead of the normal data or bss
6747 sections. The default value of @var{num} is 8.
6748 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
6749 for this option to have any effect.
6751 All modules should be compiled with the same @option{-G @var{num}} value.
6752 Compiling with different values of @var{num} may or may not work; if it
6753 doesn't the linker will give an error message---incorrect code will not be
6759 @subsection M88K Options
6760 @cindex M88k options
6762 These @samp{-m} options are defined for Motorola 88k architectures:
6767 Generate code that works well on both the m88100 and the
6772 Generate code that works best for the m88100, but that also
6777 Generate code that works best for the m88110, and may not run
6782 Obsolete option to be removed from the next revision.
6785 @item -midentify-revision
6786 @opindex midentify-revision
6787 @cindex identifying source, compiler (88k)
6788 Include an @code{ident} directive in the assembler output recording the
6789 source file name, compiler name and version, timestamp, and compilation
6792 @item -mno-underscores
6793 @opindex mno-underscores
6794 @cindex underscores, avoiding (88k)
6795 In assembler output, emit symbol names without adding an underscore
6796 character at the beginning of each name. The default is to use an
6797 underscore as prefix on each name.
6799 @item -mocs-debug-info
6800 @itemx -mno-ocs-debug-info
6801 @opindex mocs-debug-info
6802 @opindex mno-ocs-debug-info
6804 @cindex debugging, 88k OCS
6805 Include (or omit) additional debugging information (about registers used
6806 in each stack frame) as specified in the 88open Object Compatibility
6807 Standard, ``OCS''@. This extra information allows debugging of code that
6808 has had the frame pointer eliminated. The default for SVr4 and Delta 88
6809 SVr3.2 is to include this information; other 88k configurations omit this
6810 information by default.
6812 @item -mocs-frame-position
6813 @opindex mocs-frame-position
6814 @cindex register positions in frame (88k)
6815 When emitting COFF debugging information for automatic variables and
6816 parameters stored on the stack, use the offset from the canonical frame
6817 address, which is the stack pointer (register 31) on entry to the
6818 function. The SVr4 and Delta88 SVr3.2, and BCS configurations use
6819 @option{-mocs-frame-position}; other 88k configurations have the default
6820 @option{-mno-ocs-frame-position}.
6822 @item -mno-ocs-frame-position
6823 @opindex mno-ocs-frame-position
6824 @cindex register positions in frame (88k)
6825 When emitting COFF debugging information for automatic variables and
6826 parameters stored on the stack, use the offset from the frame pointer
6827 register (register 30). When this option is in effect, the frame
6828 pointer is not eliminated when debugging information is selected by the
6831 @item -moptimize-arg-area
6832 @opindex moptimize-arg-area
6833 @cindex arguments in frame (88k)
6834 Save space by reorganizing the stack frame. This option generates code
6835 that does not agree with the 88open specifications, but uses less
6838 @itemx -mno-optimize-arg-area
6839 @opindex mno-optimize-arg-area
6840 Do not reorganize the stack frame to save space. This is the default.
6841 The generated conforms to the specification, but uses more memory.
6843 @item -mshort-data-@var{num}
6844 @opindex mshort-data
6845 @cindex smaller data references (88k)
6846 @cindex r0-relative references (88k)
6847 Generate smaller data references by making them relative to @code{r0},
6848 which allows loading a value using a single instruction (rather than the
6849 usual two). You control which data references are affected by
6850 specifying @var{num} with this option. For example, if you specify
6851 @option{-mshort-data-512}, then the data references affected are those
6852 involving displacements of less than 512 bytes.
6853 @option{-mshort-data-@var{num}} is not effective for @var{num} greater
6856 @item -mserialize-volatile
6857 @opindex mserialize-volatile
6858 @itemx -mno-serialize-volatile
6859 @opindex mno-serialize-volatile
6860 @cindex sequential consistency on 88k
6861 Do, or don't, generate code to guarantee sequential consistency
6862 of volatile memory references. By default, consistency is
6865 The order of memory references made by the MC88110 processor does
6866 not always match the order of the instructions requesting those
6867 references. In particular, a load instruction may execute before
6868 a preceding store instruction. Such reordering violates
6869 sequential consistency of volatile memory references, when there
6870 are multiple processors. When consistency must be guaranteed,
6871 GCC generates special instructions, as needed, to force
6872 execution in the proper order.
6874 The MC88100 processor does not reorder memory references and so
6875 always provides sequential consistency. However, by default, GCC
6876 generates the special instructions to guarantee consistency
6877 even when you use @option{-m88100}, so that the code may be run on an
6878 MC88110 processor. If you intend to run your code only on the
6879 MC88100 processor, you may use @option{-mno-serialize-volatile}.
6881 The extra code generated to guarantee consistency may affect the
6882 performance of your application. If you know that you can safely
6883 forgo this guarantee, you may use @option{-mno-serialize-volatile}.
6889 @cindex assembler syntax, 88k
6891 Turn on (@option{-msvr4}) or off (@option{-msvr3}) compiler extensions
6892 related to System V release 4 (SVr4). This controls the following:
6896 Which variant of the assembler syntax to emit.
6898 @option{-msvr4} makes the C preprocessor recognize @samp{#pragma weak}
6899 that is used on System V release 4.
6901 @option{-msvr4} makes GCC issue additional declaration directives used in
6905 @option{-msvr4} is the default for the m88k-motorola-sysv4 configuration.
6906 @option{-msvr3} is the default for all other m88k configurations.
6908 @item -mversion-03.00
6909 @opindex mversion-03.00
6910 This option is obsolete, and is ignored.
6911 @c ??? which asm syntax better for GAS? option there too?
6913 @item -mno-check-zero-division
6914 @itemx -mcheck-zero-division
6915 @opindex mno-check-zero-division
6916 @opindex mcheck-zero-division
6917 @cindex zero division on 88k
6918 Do, or don't, generate code to guarantee that integer division by
6919 zero will be detected. By default, detection is guaranteed.
6921 Some models of the MC88100 processor fail to trap upon integer
6922 division by zero under certain conditions. By default, when
6923 compiling code that might be run on such a processor, GCC
6924 generates code that explicitly checks for zero-valued divisors
6925 and traps with exception number 503 when one is detected. Use of
6926 @option{-mno-check-zero-division} suppresses such checking for code
6927 generated to run on an MC88100 processor.
6929 GCC assumes that the MC88110 processor correctly detects all instances
6930 of integer division by zero. When @option{-m88110} is specified, no
6931 explicit checks for zero-valued divisors are generated, and both
6932 @option{-mcheck-zero-division} and @option{-mno-check-zero-division} are
6935 @item -muse-div-instruction
6936 @opindex muse-div-instruction
6937 @cindex divide instruction, 88k
6938 Use the div instruction for signed integer division on the
6939 MC88100 processor. By default, the div instruction is not used.
6941 On the MC88100 processor the signed integer division instruction
6942 div) traps to the operating system on a negative operand. The
6943 operating system transparently completes the operation, but at a
6944 large cost in execution time. By default, when compiling code
6945 that might be run on an MC88100 processor, GCC emulates signed
6946 integer division using the unsigned integer division instruction
6947 divu), thereby avoiding the large penalty of a trap to the
6948 operating system. Such emulation has its own, smaller, execution
6949 cost in both time and space. To the extent that your code's
6950 important signed integer division operations are performed on two
6951 nonnegative operands, it may be desirable to use the div
6952 instruction directly.
6954 On the MC88110 processor the div instruction (also known as the
6955 divs instruction) processes negative operands without trapping to
6956 the operating system. When @option{-m88110} is specified,
6957 @option{-muse-div-instruction} is ignored, and the div instruction is used
6958 for signed integer division.
6960 Note that the result of dividing @code{INT_MIN} by @minus{}1 is undefined. In
6961 particular, the behavior of such a division with and without
6962 @option{-muse-div-instruction} may differ.
6964 @item -mtrap-large-shift
6965 @itemx -mhandle-large-shift
6966 @opindex mtrap-large-shift
6967 @opindex mhandle-large-shift
6968 @cindex bit shift overflow (88k)
6969 @cindex large bit shifts (88k)
6970 Include code to detect bit-shifts of more than 31 bits; respectively,
6971 trap such shifts or emit code to handle them properly. By default GCC
6972 makes no special provision for large bit shifts.
6974 @item -mwarn-passed-structs
6975 @opindex mwarn-passed-structs
6976 @cindex structure passing (88k)
6977 Warn when a function passes a struct as an argument or result.
6978 Structure-passing conventions have changed during the evolution of the C
6979 language, and are often the source of portability problems. By default,
6980 GCC issues no such warning.
6983 @c break page here to avoid unsightly interparagraph stretch.
6987 @node RS/6000 and PowerPC Options
6988 @subsection IBM RS/6000 and PowerPC Options
6989 @cindex RS/6000 and PowerPC Options
6990 @cindex IBM RS/6000 and PowerPC Options
6992 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
7000 @itemx -mpowerpc-gpopt
7001 @itemx -mno-powerpc-gpopt
7002 @itemx -mpowerpc-gfxopt
7003 @itemx -mno-powerpc-gfxopt
7005 @itemx -mno-powerpc64
7011 @opindex mno-powerpc
7012 @opindex mpowerpc-gpopt
7013 @opindex mno-powerpc-gpopt
7014 @opindex mpowerpc-gfxopt
7015 @opindex mno-powerpc-gfxopt
7017 @opindex mno-powerpc64
7018 GCC supports two related instruction set architectures for the
7019 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
7020 instructions supported by the @samp{rios} chip set used in the original
7021 RS/6000 systems and the @dfn{PowerPC} instruction set is the
7022 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
7023 the IBM 4xx microprocessors.
7025 Neither architecture is a subset of the other. However there is a
7026 large common subset of instructions supported by both. An MQ
7027 register is included in processors supporting the POWER architecture.
7029 You use these options to specify which instructions are available on the
7030 processor you are using. The default value of these options is
7031 determined when configuring GCC@. Specifying the
7032 @option{-mcpu=@var{cpu_type}} overrides the specification of these
7033 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
7034 rather than the options listed above.
7036 The @option{-mpower} option allows GCC to generate instructions that
7037 are found only in the POWER architecture and to use the MQ register.
7038 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
7039 to generate instructions that are present in the POWER2 architecture but
7040 not the original POWER architecture.
7042 The @option{-mpowerpc} option allows GCC to generate instructions that
7043 are found only in the 32-bit subset of the PowerPC architecture.
7044 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
7045 GCC to use the optional PowerPC architecture instructions in the
7046 General Purpose group, including floating-point square root. Specifying
7047 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
7048 use the optional PowerPC architecture instructions in the Graphics
7049 group, including floating-point select.
7051 The @option{-mpowerpc64} option allows GCC to generate the additional
7052 64-bit instructions that are found in the full PowerPC64 architecture
7053 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
7054 @option{-mno-powerpc64}.
7056 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
7057 will use only the instructions in the common subset of both
7058 architectures plus some special AIX common-mode calls, and will not use
7059 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
7060 permits GCC to use any instruction from either architecture and to
7061 allow use of the MQ register; specify this for the Motorola MPC601.
7063 @item -mnew-mnemonics
7064 @itemx -mold-mnemonics
7065 @opindex mnew-mnemonics
7066 @opindex mold-mnemonics
7067 Select which mnemonics to use in the generated assembler code. With
7068 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
7069 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
7070 assembler mnemonics defined for the POWER architecture. Instructions
7071 defined in only one architecture have only one mnemonic; GCC uses that
7072 mnemonic irrespective of which of these options is specified.
7074 GCC defaults to the mnemonics appropriate for the architecture in
7075 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
7076 value of these option. Unless you are building a cross-compiler, you
7077 should normally not specify either @option{-mnew-mnemonics} or
7078 @option{-mold-mnemonics}, but should instead accept the default.
7080 @item -mcpu=@var{cpu_type}
7082 Set architecture type, register usage, choice of mnemonics, and
7083 instruction scheduling parameters for machine type @var{cpu_type}.
7084 Supported values for @var{cpu_type} are @samp{rios}, @samp{rios1},
7085 @samp{rsc}, @samp{rios2}, @samp{rs64a}, @samp{601}, @samp{602},
7086 @samp{603}, @samp{603e}, @samp{604}, @samp{604e}, @samp{620},
7087 @samp{630}, @samp{740}, @samp{7400}, @samp{7450}, @samp{750},
7088 @samp{power}, @samp{power2}, @samp{powerpc}, @samp{403}, @samp{505},
7089 @samp{801}, @samp{821}, @samp{823}, and @samp{860} and @samp{common}.
7091 @option{-mcpu=common} selects a completely generic processor. Code
7092 generated under this option will run on any POWER or PowerPC processor.
7093 GCC will use only the instructions in the common subset of both
7094 architectures, and will not use the MQ register. GCC assumes a generic
7095 processor model for scheduling purposes.
7097 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
7098 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
7099 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
7100 types, with an appropriate, generic processor model assumed for
7101 scheduling purposes.
7103 The other options specify a specific processor. Code generated under
7104 those options will run best on that processor, and may not run at all on
7107 The @option{-mcpu} options automatically enable or disable other
7108 @option{-m} options as follows:
7112 @option{-mno-power}, @option{-mno-powerpc}
7119 @option{-mpower}, @option{-mno-powerpc}, @option{-mno-new-mnemonics}
7134 @option{-mno-power}, @option{-mpowerpc}, @option{-mnew-mnemonics}
7137 @option{-mpower}, @option{-mpowerpc}, @option{-mnew-mnemonics}
7142 @option{-mno-power}, @option{-mpowerpc}, @option{-mnew-mnemonics}, @option{-msoft-float}
7145 @item -mtune=@var{cpu_type}
7147 Set the instruction scheduling parameters for machine type
7148 @var{cpu_type}, but do not set the architecture type, register usage, or
7149 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
7150 values for @var{cpu_type} are used for @option{-mtune} as for
7151 @option{-mcpu}. If both are specified, the code generated will use the
7152 architecture, registers, and mnemonics set by @option{-mcpu}, but the
7153 scheduling parameters set by @option{-mtune}.
7158 @opindex mno-altivec
7159 These switches enable or disable the use of built-in functions that
7160 allow access to the AltiVec instruction set. You may also need to set
7161 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
7166 Extend the current ABI with SPE ABI extensions. This does not change
7167 the default ABI, instead it adds the SPE ABI extensions to the current
7171 @opindex mabi=no-spe
7172 Disable Booke SPE ABI extensions for the current ABI.
7174 @item -misel=@var{yes/no}
7177 This switch enables or disables the generation of ISEL instructions.
7179 @item -mspe=@var{yes/no}
7182 This switch enables or disables the generation of SPE simd
7185 @item -mfloat-gprs=@var{yes/no}
7187 @opindex mfloat-gprs
7188 This switch enables or disables the generation of floating point
7189 operations on the general purpose registers for architectures that
7190 support it. This option is currently only available on the MPC8540.
7193 @itemx -mno-fp-in-toc
7194 @itemx -mno-sum-in-toc
7195 @itemx -mminimal-toc
7197 @opindex mno-fp-in-toc
7198 @opindex mno-sum-in-toc
7199 @opindex mminimal-toc
7200 Modify generation of the TOC (Table Of Contents), which is created for
7201 every executable file. The @option{-mfull-toc} option is selected by
7202 default. In that case, GCC will allocate at least one TOC entry for
7203 each unique non-automatic variable reference in your program. GCC
7204 will also place floating-point constants in the TOC@. However, only
7205 16,384 entries are available in the TOC@.
7207 If you receive a linker error message that saying you have overflowed
7208 the available TOC space, you can reduce the amount of TOC space used
7209 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
7210 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
7211 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
7212 generate code to calculate the sum of an address and a constant at
7213 run-time instead of putting that sum into the TOC@. You may specify one
7214 or both of these options. Each causes GCC to produce very slightly
7215 slower and larger code at the expense of conserving TOC space.
7217 If you still run out of space in the TOC even when you specify both of
7218 these options, specify @option{-mminimal-toc} instead. This option causes
7219 GCC to make only one TOC entry for every file. When you specify this
7220 option, GCC will produce code that is slower and larger but which
7221 uses extremely little TOC space. You may wish to use this option
7222 only on files that contain less frequently executed code.
7228 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
7229 @code{long} type, and the infrastructure needed to support them.
7230 Specifying @option{-maix64} implies @option{-mpowerpc64} and
7231 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
7232 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
7237 @opindex mno-xl-call
7238 On AIX, pass floating-point arguments to prototyped functions beyond the
7239 register save area (RSA) on the stack in addition to argument FPRs. The
7240 AIX calling convention was extended but not initially documented to
7241 handle an obscure K&R C case of calling a function that takes the
7242 address of its arguments with fewer arguments than declared. AIX XL
7243 compilers access floating point arguments which do not fit in the
7244 RSA from the stack when a subroutine is compiled without
7245 optimization. Because always storing floating-point arguments on the
7246 stack is inefficient and rarely needed, this option is not enabled by
7247 default and only is necessary when calling subroutines compiled by AIX
7248 XL compilers without optimization.
7252 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
7253 application written to use message passing with special startup code to
7254 enable the application to run. The system must have PE installed in the
7255 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
7256 must be overridden with the @option{-specs=} option to specify the
7257 appropriate directory location. The Parallel Environment does not
7258 support threads, so the @option{-mpe} option and the @option{-pthread}
7259 option are incompatible.
7261 @item -malign-natural
7262 @itemx -malign-power
7263 @opindex malign-natural
7264 @opindex malign-power
7265 On AIX, Darwin, and 64-bit PowerPC Linux, the option
7266 @option{-malign-natural} overrides the ABI-defined alignment of larger
7267 types, such as floating-point doubles, on their natural size-based boundary.
7268 The option @option{-malign-power} instructs GCC to follow the ABI-specified
7269 alignment rules. GCC defaults to the standard alignment defined in the ABI.
7273 @opindex msoft-float
7274 @opindex mhard-float
7275 Generate code that does not use (uses) the floating-point register set.
7276 Software floating point emulation is provided if you use the
7277 @option{-msoft-float} option, and pass the option to GCC when linking.
7280 @itemx -mno-multiple
7282 @opindex mno-multiple
7283 Generate code that uses (does not use) the load multiple word
7284 instructions and the store multiple word instructions. These
7285 instructions are generated by default on POWER systems, and not
7286 generated on PowerPC systems. Do not use @option{-mmultiple} on little
7287 endian PowerPC systems, since those instructions do not work when the
7288 processor is in little endian mode. The exceptions are PPC740 and
7289 PPC750 which permit the instructions usage in little endian mode.
7295 Generate code that uses (does not use) the load string instructions
7296 and the store string word instructions to save multiple registers and
7297 do small block moves. These instructions are generated by default on
7298 POWER systems, and not generated on PowerPC systems. Do not use
7299 @option{-mstring} on little endian PowerPC systems, since those
7300 instructions do not work when the processor is in little endian mode.
7301 The exceptions are PPC740 and PPC750 which permit the instructions
7302 usage in little endian mode.
7308 Generate code that uses (does not use) the load or store instructions
7309 that update the base register to the address of the calculated memory
7310 location. These instructions are generated by default. If you use
7311 @option{-mno-update}, there is a small window between the time that the
7312 stack pointer is updated and the address of the previous frame is
7313 stored, which means code that walks the stack frame across interrupts or
7314 signals may get corrupted data.
7317 @itemx -mno-fused-madd
7318 @opindex mfused-madd
7319 @opindex mno-fused-madd
7320 Generate code that uses (does not use) the floating point multiply and
7321 accumulate instructions. These instructions are generated by default if
7322 hardware floating is used.
7324 @item -mno-bit-align
7326 @opindex mno-bit-align
7328 On System V.4 and embedded PowerPC systems do not (do) force structures
7329 and unions that contain bit-fields to be aligned to the base type of the
7332 For example, by default a structure containing nothing but 8
7333 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
7334 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
7335 the structure would be aligned to a 1 byte boundary and be one byte in
7338 @item -mno-strict-align
7339 @itemx -mstrict-align
7340 @opindex mno-strict-align
7341 @opindex mstrict-align
7342 On System V.4 and embedded PowerPC systems do not (do) assume that
7343 unaligned memory references will be handled by the system.
7346 @itemx -mno-relocatable
7347 @opindex mrelocatable
7348 @opindex mno-relocatable
7349 On embedded PowerPC systems generate code that allows (does not allow)
7350 the program to be relocated to a different address at runtime. If you
7351 use @option{-mrelocatable} on any module, all objects linked together must
7352 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
7354 @item -mrelocatable-lib
7355 @itemx -mno-relocatable-lib
7356 @opindex mrelocatable-lib
7357 @opindex mno-relocatable-lib
7358 On embedded PowerPC systems generate code that allows (does not allow)
7359 the program to be relocated to a different address at runtime. Modules
7360 compiled with @option{-mrelocatable-lib} can be linked with either modules
7361 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
7362 with modules compiled with the @option{-mrelocatable} options.
7368 On System V.4 and embedded PowerPC systems do not (do) assume that
7369 register 2 contains a pointer to a global area pointing to the addresses
7370 used in the program.
7373 @itemx -mlittle-endian
7375 @opindex mlittle-endian
7376 On System V.4 and embedded PowerPC systems compile code for the
7377 processor in little endian mode. The @option{-mlittle-endian} option is
7378 the same as @option{-mlittle}.
7383 @opindex mbig-endian
7384 On System V.4 and embedded PowerPC systems compile code for the
7385 processor in big endian mode. The @option{-mbig-endian} option is
7386 the same as @option{-mbig}.
7388 @item -mdynamic-no-pic
7389 @opindex mdynamic-no-pic
7390 On Darwin and Mac OS X systems, compile code so that it is not
7391 relocatable, but that its external references are relocatable. The
7392 resulting code is suitable for applications, but not shared
7397 On System V.4 and embedded PowerPC systems compile code using calling
7398 conventions that adheres to the March 1995 draft of the System V
7399 Application Binary Interface, PowerPC processor supplement. This is the
7400 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
7402 @item -mcall-sysv-eabi
7403 @opindex mcall-sysv-eabi
7404 Specify both @option{-mcall-sysv} and @option{-meabi} options.
7406 @item -mcall-sysv-noeabi
7407 @opindex mcall-sysv-noeabi
7408 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
7410 @item -mcall-solaris
7411 @opindex mcall-solaris
7412 On System V.4 and embedded PowerPC systems compile code for the Solaris
7416 @opindex mcall-linux
7417 On System V.4 and embedded PowerPC systems compile code for the
7418 Linux-based GNU system.
7422 On System V.4 and embedded PowerPC systems compile code for the
7423 Hurd-based GNU system.
7426 @opindex mcall-netbsd
7427 On System V.4 and embedded PowerPC systems compile code for the
7428 NetBSD operating system.
7430 @item -maix-struct-return
7431 @opindex maix-struct-return
7432 Return all structures in memory (as specified by the AIX ABI)@.
7434 @item -msvr4-struct-return
7435 @opindex msvr4-struct-return
7436 Return structures smaller than 8 bytes in registers (as specified by the
7440 @opindex mabi=altivec
7441 Extend the current ABI with AltiVec ABI extensions. This does not
7442 change the default ABI, instead it adds the AltiVec ABI extensions to
7445 @item -mabi=no-altivec
7446 @opindex mabi=no-altivec
7447 Disable AltiVec ABI extensions for the current ABI.
7450 @itemx -mno-prototype
7452 @opindex mno-prototype
7453 On System V.4 and embedded PowerPC systems assume that all calls to
7454 variable argument functions are properly prototyped. Otherwise, the
7455 compiler must insert an instruction before every non prototyped call to
7456 set or clear bit 6 of the condition code register (@var{CR}) to
7457 indicate whether floating point values were passed in the floating point
7458 registers in case the function takes a variable arguments. With
7459 @option{-mprototype}, only calls to prototyped variable argument functions
7460 will set or clear the bit.
7464 On embedded PowerPC systems, assume that the startup module is called
7465 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
7466 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
7471 On embedded PowerPC systems, assume that the startup module is called
7472 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
7477 On embedded PowerPC systems, assume that the startup module is called
7478 @file{crt0.o} and the standard C libraries are @file{libads.a} and
7482 @opindex myellowknife
7483 On embedded PowerPC systems, assume that the startup module is called
7484 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
7489 On System V.4 and embedded PowerPC systems, specify that you are
7490 compiling for a VxWorks system.
7494 Specify that you are compiling for the WindISS simulation environment.
7498 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
7499 header to indicate that @samp{eabi} extended relocations are used.
7505 On System V.4 and embedded PowerPC systems do (do not) adhere to the
7506 Embedded Applications Binary Interface (eabi) which is a set of
7507 modifications to the System V.4 specifications. Selecting @option{-meabi}
7508 means that the stack is aligned to an 8 byte boundary, a function
7509 @code{__eabi} is called to from @code{main} to set up the eabi
7510 environment, and the @option{-msdata} option can use both @code{r2} and
7511 @code{r13} to point to two separate small data areas. Selecting
7512 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
7513 do not call an initialization function from @code{main}, and the
7514 @option{-msdata} option will only use @code{r13} to point to a single
7515 small data area. The @option{-meabi} option is on by default if you
7516 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
7519 @opindex msdata=eabi
7520 On System V.4 and embedded PowerPC systems, put small initialized
7521 @code{const} global and static data in the @samp{.sdata2} section, which
7522 is pointed to by register @code{r2}. Put small initialized
7523 non-@code{const} global and static data in the @samp{.sdata} section,
7524 which is pointed to by register @code{r13}. Put small uninitialized
7525 global and static data in the @samp{.sbss} section, which is adjacent to
7526 the @samp{.sdata} section. The @option{-msdata=eabi} option is
7527 incompatible with the @option{-mrelocatable} option. The
7528 @option{-msdata=eabi} option also sets the @option{-memb} option.
7531 @opindex msdata=sysv
7532 On System V.4 and embedded PowerPC systems, put small global and static
7533 data in the @samp{.sdata} section, which is pointed to by register
7534 @code{r13}. Put small uninitialized global and static data in the
7535 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
7536 The @option{-msdata=sysv} option is incompatible with the
7537 @option{-mrelocatable} option.
7539 @item -msdata=default
7541 @opindex msdata=default
7543 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
7544 compile code the same as @option{-msdata=eabi}, otherwise compile code the
7545 same as @option{-msdata=sysv}.
7548 @opindex msdata-data
7549 On System V.4 and embedded PowerPC systems, put small global and static
7550 data in the @samp{.sdata} section. Put small uninitialized global and
7551 static data in the @samp{.sbss} section. Do not use register @code{r13}
7552 to address small data however. This is the default behavior unless
7553 other @option{-msdata} options are used.
7557 @opindex msdata=none
7559 On embedded PowerPC systems, put all initialized global and static data
7560 in the @samp{.data} section, and all uninitialized data in the
7561 @samp{.bss} section.
7565 @cindex smaller data references (PowerPC)
7566 @cindex .sdata/.sdata2 references (PowerPC)
7567 On embedded PowerPC systems, put global and static items less than or
7568 equal to @var{num} bytes into the small data or bss sections instead of
7569 the normal data or bss section. By default, @var{num} is 8. The
7570 @option{-G @var{num}} switch is also passed to the linker.
7571 All modules should be compiled with the same @option{-G @var{num}} value.
7574 @itemx -mno-regnames
7576 @opindex mno-regnames
7577 On System V.4 and embedded PowerPC systems do (do not) emit register
7578 names in the assembly language output using symbolic forms.
7581 @itemx -mno-longcall
7583 @opindex mno-longcall
7584 Default to making all function calls via pointers, so that functions
7585 which reside further than 64 megabytes (67,108,864 bytes) from the
7586 current location can be called. This setting can be overridden by the
7587 @code{shortcall} function attribute, or by @code{#pragma longcall(0)}.
7589 Some linkers are capable of detecting out-of-range calls and generating
7590 glue code on the fly. On these systems, long calls are unnecessary and
7591 generate slower code. As of this writing, the AIX linker can do this,
7592 as can the GNU linker for PowerPC/64. It is planned to add this feature
7593 to the GNU linker for 32-bit PowerPC systems as well.
7595 In the future, we may cause GCC to ignore all longcall specifications
7596 when the linker is known to generate glue.
7600 Adds support for multithreading with the @dfn{pthreads} library.
7601 This option sets flags for both the preprocessor and linker.
7605 @node Darwin Options
7606 @subsection Darwin Options
7607 @cindex Darwin options
7609 These options are defined for all architectures running the Darwin operating
7610 system. They are useful for compatibility with other Mac OS compilers.
7615 Loads all members of static archive libraries.
7616 See man ld(1) for more information.
7618 @item -arch_errors_fatal
7619 @opindex arch_errors_fatal
7620 Cause the errors having to do with files that have the wrong architecture
7624 @opindex bind_at_load
7625 Causes the output file to be marked such that the dynamic linker will
7626 bind all undefined references when the file is loaded or launched.
7630 Produce a Mach-o bundle format file.
7631 See man ld(1) for more information.
7633 @item -bundle_loader @var{executable}
7634 @opindex bundle_loader
7635 This specifies the @var{executable} that will be loading the build
7636 output file being linked. See man ld(1) for more information.
7638 @item -allowable_client @var{client_name}
7642 @item -compatibility_version
7643 @item -current_version
7644 @item -dependency-file
7646 @item -dylinker_install_name
7649 @item -exported_symbols_list
7651 @item -flat_namespace
7652 @item -force_cpusubtype_ALL
7653 @item -force_flat_namespace
7654 @item -headerpad_max_install_names
7658 @item -keep_private_externs
7660 @item -multiply_defined
7661 @item -multiply_defined_unused
7663 @item -nofixprebinding
7666 @item -noseglinkedit
7667 @item -pagezero_size
7669 @item -prebind_all_twolevel_modules
7670 @item -private_bundle
7671 @item -read_only_relocs
7673 @item -sectobjectsymbols
7677 @item -sectobjectsymbols
7679 @item -seg_addr_table
7680 @item -seg_addr_table_filename
7683 @item -segs_read_only_addr
7684 @item -segs_read_write_addr
7685 @item -single_module
7689 @item -twolevel_namespace
7692 @item -unexported_symbols_list
7693 @item -weak_reference_mismatches
7696 @opindex allowable_client
7698 @opindex client_name
7699 @opindex compatibility_version
7700 @opindex current_version
7701 @opindex dependency-file
7703 @opindex dylinker_install_name
7706 @opindex exported_symbols_list
7708 @opindex flat_namespace
7709 @opindex force_cpusubtype_ALL
7710 @opindex force_flat_namespace
7711 @opindex headerpad_max_install_names
7714 @opindex install_name
7715 @opindex keep_private_externs
7716 @opindex multi_module
7717 @opindex multiply_defined
7718 @opindex multiply_defined_unused
7720 @opindex nofixprebinding
7721 @opindex nomultidefs
7723 @opindex noseglinkedit
7724 @opindex pagezero_size
7726 @opindex prebind_all_twolevel_modules
7727 @opindex private_bundle
7728 @opindex read_only_relocs
7730 @opindex sectobjectsymbols
7734 @opindex sectobjectsymbols
7736 @opindex seg_addr_table
7737 @opindex seg_addr_table_filename
7738 @opindex seglinkedit
7740 @opindex segs_read_only_addr
7741 @opindex segs_read_write_addr
7742 @opindex single_module
7744 @opindex sub_library
7745 @opindex sub_umbrella
7746 @opindex twolevel_namespace
7749 @opindex unexported_symbols_list
7750 @opindex weak_reference_mismatches
7751 @opindex whatsloaded
7753 This options are available for Darwin linker. Darwin linker man page
7754 describes them in detail.
7759 @subsection IBM RT Options
7761 @cindex IBM RT options
7763 These @samp{-m} options are defined for the IBM RT PC:
7767 @opindex min-line-mul
7768 Use an in-line code sequence for integer multiplies. This is the
7771 @item -mcall-lib-mul
7772 @opindex mcall-lib-mul
7773 Call @code{lmul$$} for integer multiples.
7775 @item -mfull-fp-blocks
7776 @opindex mfull-fp-blocks
7777 Generate full-size floating point data blocks, including the minimum
7778 amount of scratch space recommended by IBM@. This is the default.
7780 @item -mminimum-fp-blocks
7781 @opindex mminimum-fp-blocks
7782 Do not include extra scratch space in floating point data blocks. This
7783 results in smaller code, but slower execution, since scratch space must
7784 be allocated dynamically.
7786 @cindex @file{stdarg.h} and RT PC
7787 @item -mfp-arg-in-fpregs
7788 @opindex mfp-arg-in-fpregs
7789 Use a calling sequence incompatible with the IBM calling convention in
7790 which floating point arguments are passed in floating point registers.
7791 Note that @code{stdarg.h} will not work with floating point operands
7792 if this option is specified.
7794 @item -mfp-arg-in-gregs
7795 @opindex mfp-arg-in-gregs
7796 Use the normal calling convention for floating point arguments. This is
7799 @item -mhc-struct-return
7800 @opindex mhc-struct-return
7801 Return structures of more than one word in memory, rather than in a
7802 register. This provides compatibility with the MetaWare HighC (hc)
7803 compiler. Use the option @option{-fpcc-struct-return} for compatibility
7804 with the Portable C Compiler (pcc).
7806 @item -mnohc-struct-return
7807 @opindex mnohc-struct-return
7808 Return some structures of more than one word in registers, when
7809 convenient. This is the default. For compatibility with the
7810 IBM-supplied compilers, use the option @option{-fpcc-struct-return} or the
7811 option @option{-mhc-struct-return}.
7815 @subsection MIPS Options
7816 @cindex MIPS options
7818 These @samp{-m} options are defined for the MIPS family of computers:
7822 @item -march=@var{arch}
7824 Generate code that will run on @var{arch}, which can be the name of a
7825 generic MIPS ISA, or the name of a particular processor.
7827 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
7828 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
7829 The processor names are:
7830 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
7832 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
7833 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
7837 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4300},
7838 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
7839 The special value @samp{from-abi} selects the
7840 most compatible architecture for the selected ABI (that is,
7841 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
7843 In processor names, a final @samp{000} can be abbreviated as @samp{k}
7844 (for example, @samp{-march=r2k}). Prefixes are optional, and
7845 @samp{vr} may be written @samp{r}.
7847 GCC defines two macros based on the value of this option. The first
7848 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
7849 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
7850 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
7851 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
7852 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
7854 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
7855 above. In other words, it will have the full prefix and will not
7856 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
7857 the macro names the resolved architecture (either @samp{"mips1"} or
7858 @samp{"mips3"}). It names the default architecture when no
7859 @option{-march} option is given.
7861 @item -mtune=@var{arch}
7863 Optimize for @var{arch}. Among other things, this option controls
7864 the way instructions are scheduled, and the perceived cost of arithmetic
7865 operations. The list of @var{arch} values is the same as for
7868 When this option is not used, GCC will optimize for the processor
7869 specified by @option{-march}. By using @option{-march} and
7870 @option{-mtune} together, it is possible to generate code that will
7871 run on a family of processors, but optimize the code for one
7872 particular member of that family.
7874 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
7875 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
7876 @samp{-march} ones described above.
7880 Equivalent to @samp{-march=mips1}.
7884 Equivalent to @samp{-march=mips2}.
7888 Equivalent to @samp{-march=mips3}.
7892 Equivalent to @samp{-march=mips4}.
7896 Equivalent to @samp{-march=mips32}.
7900 Equivalent to @samp{-march=mips32r2}.
7904 Equivalent to @samp{-march=mips64}.
7907 @itemx -mno-fused-madd
7908 @opindex mfused-madd
7909 @opindex mno-fused-madd
7910 Generate code that uses (does not use) the floating point multiply and
7911 accumulate instructions, when they are available. These instructions
7912 are generated by default if they are available, but this may be
7913 undesirable if the extra precision causes problems or on certain chips
7914 in the mode where denormals are rounded to zero where denormals
7915 generated by multiply and accumulate instructions cause exceptions
7920 Assume that floating point registers are 32 bits wide.
7924 Assume that floating point registers are 64 bits wide.
7928 Assume that general purpose registers are 32 bits wide.
7932 Assume that general purpose registers are 64 bits wide.
7936 Force int and long types to be 64 bits wide. See @option{-mlong32} for an
7937 explanation of the default, and the width of pointers.
7941 Force long types to be 64 bits wide. See @option{-mlong32} for an
7942 explanation of the default, and the width of pointers.
7946 Force long, int, and pointer types to be 32 bits wide.
7948 The default size of ints, longs and pointers depends on the ABI@. All
7949 the supported ABIs use 32-bit ints. The n64 ABI uses 64-bit longs, as
7950 does the 64-bit Cygnus EABI; the others use 32-bit longs. Pointers
7951 are the same size as longs, or the same size as integer registers,
7952 whichever is smaller.
7966 Generate code for the given ABI@.
7968 Note that there are two embedded ABIs: @option{-mabi=eabi}
7969 selects the one defined by Cygnus while @option{-meabi=meabi}
7970 selects the one defined by MIPS@. Both these ABIs have
7971 32-bit and 64-bit variants. Normally, GCC will generate
7972 64-bit code when you select a 64-bit architecture, but you
7973 can use @option{-mgp32} to get 32-bit code instead.
7975 @item -mabi-fake-default
7976 @opindex mabi-fake-default
7977 You don't want to know what this option does. No, really. I mean
7978 it. Move on to the next option.
7980 What? You're still here? Oh, well@enddots{} Ok, here's the deal. GCC
7981 wants the default set of options to get the root of the multilib tree,
7982 and the shared library SONAMEs without any multilib-indicating
7983 suffixes. This is not convenience for @samp{mips64-linux-gnu}, since
7984 we want to default to the N32 ABI, while still being binary-compatible
7985 with @samp{mips-linux-gnu} if you stick to the O32 ABI@. Being
7986 binary-compatible means shared libraries should have the same SONAMEs,
7987 and libraries should live in the same location. Having O32 libraries
7988 in a sub-directory named say @file{o32} is not acceptable.
7990 So we trick GCC into believing that O32 is the default ABI, except
7991 that we override the default with some internal command-line
7992 processing magic. Problem is, if we stopped at that, and you then
7993 created a multilib-aware package that used the output of @command{gcc
7994 -print-multi-lib} to decide which multilibs to build, and how, and
7995 you'd find yourself in an awkward situation when you found out that
7996 some of the options listed ended up mapping to the same multilib, and
7997 none of your libraries was actually built for the multilib that
7998 @option{-print-multi-lib} claims to be the default. So we added this
7999 option that disables the default switcher, falling back to GCC's
8000 original notion of the default library. Confused yet?
8002 For short: don't ever use this option, unless you find it in the list
8003 of additional options to be used when building for multilibs, in the
8004 output of @option{gcc -print-multi-lib}.
8008 Generate code for the MIPS assembler, and invoke @file{mips-tfile} to
8009 add normal debug information. This is the default for all
8010 platforms except for the OSF/1 reference platform, using the OSF/rose
8011 object format. If the either of the @option{-gstabs} or @option{-gstabs+}
8012 switches are used, the @file{mips-tfile} program will encapsulate the
8013 stabs within MIPS ECOFF@.
8017 Generate code for the GNU assembler. This is the default on the OSF/1
8018 reference platform, using the OSF/rose object format. Also, this is
8019 the default if the configure option @option{--with-gnu-as} is used.
8021 @item -msplit-addresses
8022 @itemx -mno-split-addresses
8023 @opindex msplit-addresses
8024 @opindex mno-split-addresses
8025 Generate code to load the high and low parts of address constants separately.
8026 This allows GCC to optimize away redundant loads of the high order
8027 bits of addresses. This optimization requires GNU as and GNU ld.
8028 This optimization is enabled by default for some embedded targets where
8029 GNU as and GNU ld are standard.
8035 The @option{-mrnames} switch says to output code using the MIPS software
8036 names for the registers, instead of the hardware names (ie, @var{a0}
8037 instead of @var{$4}). The only known assembler that supports this option
8038 is the Algorithmics assembler.
8044 The @option{-mmemcpy} switch makes all block moves call the appropriate
8045 string function (@samp{memcpy} or @samp{bcopy}) instead of possibly
8046 generating inline code.
8049 @itemx -mno-mips-tfile
8050 @opindex mmips-tfile
8051 @opindex mno-mips-tfile
8052 The @option{-mno-mips-tfile} switch causes the compiler not
8053 postprocess the object file with the @file{mips-tfile} program,
8054 after the MIPS assembler has generated it to add debug support. If
8055 @file{mips-tfile} is not run, then no local variables will be
8056 available to the debugger. In addition, @file{stage2} and
8057 @file{stage3} objects will have the temporary file names passed to the
8058 assembler embedded in the object file, which means the objects will
8059 not compare the same. The @option{-mno-mips-tfile} switch should only
8060 be used when there are bugs in the @file{mips-tfile} program that
8061 prevents compilation.
8064 @opindex msoft-float
8065 Generate output containing library calls for floating point.
8066 @strong{Warning:} the requisite libraries are not part of GCC@.
8067 Normally the facilities of the machine's usual C compiler are used, but
8068 this can't be done directly in cross-compilation. You must make your
8069 own arrangements to provide suitable library functions for
8073 @opindex mhard-float
8074 Generate output containing floating point instructions. This is the
8075 default if you use the unmodified sources.
8078 @itemx -mno-abicalls
8080 @opindex mno-abicalls
8081 Emit (or do not emit) the pseudo operations @samp{.abicalls},
8082 @samp{.cpload}, and @samp{.cprestore} that some System V.4 ports use for
8083 position independent code.
8089 Lift (or do not lift) the usual restrictions on the size of the global
8092 GCC normally uses a single instruction to load values from the GOT.
8093 While this is relatively efficient, it will only work if the GOT
8094 is smaller than about 64k. Anything larger will cause the linker
8095 to report an error such as:
8097 @cindex relocation truncated to fit (MIPS)
8099 relocation truncated to fit: R_MIPS_GOT16 foobar
8102 If this happens, you should recompile your code with @option{-mxgot}.
8103 It should then work with very large GOTs, although it will also be
8104 less efficient, since it will take three instructions to fetch the
8105 value of a global symbol.
8107 Note that some linkers can create multiple GOTs. If you have such a
8108 linker, you should only need to use @option{-mxgot} when a single object
8109 file accesses more than 64k's worth of GOT entries. Very few do.
8111 These options have no effect unless GCC is generating position
8115 @itemx -mno-long-calls
8116 @opindex mlong-calls
8117 @opindex mno-long-calls
8118 Do all calls with the @samp{JALR} instruction, which requires
8119 loading up a function's address into a register before the call.
8120 You need to use this switch, if you call outside of the current
8121 512 megabyte segment to functions that are not through pointers.
8123 @item -membedded-pic
8124 @itemx -mno-embedded-pic
8125 @opindex membedded-pic
8126 @opindex mno-embedded-pic
8127 Generate PIC code suitable for some embedded systems. All calls are
8128 made using PC relative address, and all data is addressed using the $gp
8129 register. No more than 65536 bytes of global data may be used. This
8130 requires GNU as and GNU ld which do most of the work. This currently
8131 only works on targets which use ECOFF; it does not work with ELF@.
8133 @item -membedded-data
8134 @itemx -mno-embedded-data
8135 @opindex membedded-data
8136 @opindex mno-embedded-data
8137 Allocate variables to the read-only data section first if possible, then
8138 next in the small data section if possible, otherwise in data. This gives
8139 slightly slower code than the default, but reduces the amount of RAM required
8140 when executing, and thus may be preferred for some embedded systems.
8142 @item -muninit-const-in-rodata
8143 @itemx -mno-uninit-const-in-rodata
8144 @opindex muninit-const-in-rodata
8145 @opindex mno-uninit-const-in-rodata
8146 When used together with @option{-membedded-data}, it will always store uninitialized
8147 const variables in the read-only data section.
8149 @item -msingle-float
8150 @itemx -mdouble-float
8151 @opindex msingle-float
8152 @opindex mdouble-float
8153 The @option{-msingle-float} switch tells gcc to assume that the floating
8154 point coprocessor only supports single precision operations, as on the
8155 @samp{r4650} chip. The @option{-mdouble-float} switch permits gcc to use
8156 double precision operations. This is the default.
8162 Permit use of the @samp{mad}, @samp{madu} and @samp{mul} instructions,
8163 as on the @samp{r4650} chip.
8167 Turns on @option{-msingle-float}, @option{-mmad}, and, at least for now,
8168 @option{-mcpu=r4650}.
8174 Enable 16-bit instructions.
8178 Use the entry and exit pseudo ops. This option can only be used with
8183 Compile code for the processor in little endian mode.
8184 The requisite libraries are assumed to exist.
8188 Compile code for the processor in big endian mode.
8189 The requisite libraries are assumed to exist.
8193 @cindex smaller data references (MIPS)
8194 @cindex gp-relative references (MIPS)
8195 Put global and static items less than or equal to @var{num} bytes into
8196 the small data or bss sections instead of the normal data or bss
8197 section. This allows the assembler to emit one word memory reference
8198 instructions based on the global pointer (@var{gp} or @var{$28}),
8199 instead of the normal two words used. By default, @var{num} is 8 when
8200 the MIPS assembler is used, and 0 when the GNU assembler is used. The
8201 @option{-G @var{num}} switch is also passed to the assembler and linker.
8202 All modules should be compiled with the same @option{-G @var{num}}
8207 Tell the MIPS assembler to not run its preprocessor over user
8208 assembler files (with a @samp{.s} suffix) when assembling them.
8212 Pass an option to gas which will cause nops to be inserted if
8213 the read of the destination register of an mfhi or mflo instruction
8214 occurs in the following two instructions.
8219 Work around certain SB-1 CPU core errata.
8220 (This flag currently works around the SB-1 revision 2
8221 ``F1'' and ``F2'' floating point errata.)
8225 Do not include the default crt0.
8227 @item -mflush-func=@var{func}
8228 @itemx -mno-flush-func
8229 @opindex mflush-func
8230 Specifies the function to call to flush the I and D caches, or to not
8231 call any such function. If called, the function must take the same
8232 arguments as the common @code{_flush_func()}, that is, the address of the
8233 memory range for which the cache is being flushed, the size of the
8234 memory range, and the number 3 (to flush both caches). The default
8235 depends on the target gcc was configured for, but commonly is either
8236 @samp{_flush_func} or @samp{__cpu_flush}.
8238 @item -mbranch-likely
8239 @itemx -mno-branch-likely
8240 @opindex mbranch-likely
8241 @opindex mno-branch-likely
8242 Enable or disable use of Branch Likely instructions, regardless of the
8243 default for the selected architecture. By default, Branch Likely
8244 instructions may be generated if they are supported by the selected
8245 architecture. An exception is for the MIPS32 and MIPS64 architectures
8246 and processors which implement those architectures; for those, Branch
8247 Likely instructions will not be generated by default because the MIPS32
8248 and MIPS64 architectures specifically deprecate their use.
8251 @node i386 and x86-64 Options
8252 @subsection Intel 386 and AMD x86-64 Options
8253 @cindex i386 Options
8254 @cindex x86-64 Options
8255 @cindex Intel 386 Options
8256 @cindex AMD x86-64 Options
8258 These @samp{-m} options are defined for the i386 and x86-64 family of
8262 @item -mtune=@var{cpu-type}
8264 Tune to @var{cpu-type} everything applicable about the generated code, except
8265 for the ABI and the set of available instructions. The choices for
8266 @var{cpu-type} are @samp{i386}, @samp{i486}, @samp{i586}, @samp{i686},
8267 @samp{pentium}, @samp{pentium-mmx}, @samp{pentiumpro}, @samp{pentium2},
8268 @samp{pentium3}, @samp{pentium4}, @samp{k6}, @samp{k6-2}, @samp{k6-3},
8269 @samp{athlon}, @samp{athlon-tbird}, @samp{athlon-4}, @samp{athlon-xp},
8270 @samp{athlon-mp}, @samp{winchip-c6}, @samp{winchip2}, @samp{k8}, @samp{c3}
8273 While picking a specific @var{cpu-type} will schedule things appropriately
8274 for that particular chip, the compiler will not generate any code that
8275 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8276 being used. @samp{i586} is equivalent to @samp{pentium} and @samp{i686}
8277 is equivalent to @samp{pentiumpro}. @samp{k6} and @samp{athlon} are the
8278 AMD chips as opposed to the Intel ones.
8280 @item -march=@var{cpu-type}
8282 Generate instructions for the machine type @var{cpu-type}. The choices
8283 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8284 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8286 @item -mcpu=@var{cpu-type}
8288 A deprecated synonym for @option{-mtune}.
8297 @opindex mpentiumpro
8298 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8299 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8300 These synonyms are deprecated.
8302 @item -mfpmath=@var{unit}
8304 generate floating point arithmetics for selected unit @var{unit}. the choices
8309 Use the standard 387 floating point coprocessor present majority of chips and
8310 emulated otherwise. Code compiled with this option will run almost everywhere.
8311 The temporary results are computed in 80bit precision instead of precision
8312 specified by the type resulting in slightly different results compared to most
8313 of other chips. See @option{-ffloat-store} for more detailed description.
8315 This is the default choice for i386 compiler.
8318 Use scalar floating point instructions present in the SSE instruction set.
8319 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8320 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8321 instruction set supports only single precision arithmetics, thus the double and
8322 extended precision arithmetics is still done using 387. Later version, present
8323 only in Pentium4 and the future AMD x86-64 chips supports double precision
8326 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8327 @option{-msse2} switches to enable SSE extensions and make this option
8328 effective. For x86-64 compiler, these extensions are enabled by default.
8330 The resulting code should be considerably faster in majority of cases and avoid
8331 the numerical instability problems of 387 code, but may break some existing
8332 code that expects temporaries to be 80bit.
8334 This is the default choice for x86-64 compiler.
8337 Use all SSE extensions enabled by @option{-msse2} as well as the new
8338 SSE extensions in Prescott New Instructions. @option{-mpni} also
8339 enables 2 builtin functions, @code{__builtin_ia32_monitor} and
8340 @code{__builtin_ia32_mwait}, for new instructions @code{monitor} and
8344 Attempt to utilize both instruction sets at once. This effectively double the
8345 amount of available registers and on chips with separate execution units for
8346 387 and SSE the execution resources too. Use this option with care, as it is
8347 still experimental, because gcc register allocator does not model separate
8348 functional units well resulting in instable performance.
8351 @item -masm=@var{dialect}
8352 @opindex masm=@var{dialect}
8353 Output asm instructions using selected @var{dialect}. Supported choices are
8354 @samp{intel} or @samp{att} (the default one).
8359 @opindex mno-ieee-fp
8360 Control whether or not the compiler uses IEEE floating point
8361 comparisons. These handle correctly the case where the result of a
8362 comparison is unordered.
8365 @opindex msoft-float
8366 Generate output containing library calls for floating point.
8367 @strong{Warning:} the requisite libraries are not part of GCC@.
8368 Normally the facilities of the machine's usual C compiler are used, but
8369 this can't be done directly in cross-compilation. You must make your
8370 own arrangements to provide suitable library functions for
8373 On machines where a function returns floating point results in the 80387
8374 register stack, some floating point opcodes may be emitted even if
8375 @option{-msoft-float} is used.
8377 @item -mno-fp-ret-in-387
8378 @opindex mno-fp-ret-in-387
8379 Do not use the FPU registers for return values of functions.
8381 The usual calling convention has functions return values of types
8382 @code{float} and @code{double} in an FPU register, even if there
8383 is no FPU@. The idea is that the operating system should emulate
8386 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8387 in ordinary CPU registers instead.
8389 @item -mno-fancy-math-387
8390 @opindex mno-fancy-math-387
8391 Some 387 emulators do not support the @code{sin}, @code{cos} and
8392 @code{sqrt} instructions for the 387. Specify this option to avoid
8393 generating those instructions. This option is the default on FreeBSD,
8394 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8395 indicates that the target cpu will always have an FPU and so the
8396 instruction will not need emulation. As of revision 2.6.1, these
8397 instructions are not generated unless you also use the
8398 @option{-funsafe-math-optimizations} switch.
8400 @item -malign-double
8401 @itemx -mno-align-double
8402 @opindex malign-double
8403 @opindex mno-align-double
8404 Control whether GCC aligns @code{double}, @code{long double}, and
8405 @code{long long} variables on a two word boundary or a one word
8406 boundary. Aligning @code{double} variables on a two word boundary will
8407 produce code that runs somewhat faster on a @samp{Pentium} at the
8408 expense of more memory.
8410 @strong{Warning:} if you use the @option{-malign-double} switch,
8411 structures containing the above types will be aligned differently than
8412 the published application binary interface specifications for the 386
8413 and will not be binary compatible with structures in code compiled
8414 without that switch.
8416 @item -m96bit-long-double
8417 @item -m128bit-long-double
8418 @opindex m96bit-long-double
8419 @opindex m128bit-long-double
8420 These switches control the size of @code{long double} type. The i386
8421 application binary interface specifies the size to be 96 bits,
8422 so @option{-m96bit-long-double} is the default in 32 bit mode.
8424 Modern architectures (Pentium and newer) would prefer @code{long double}
8425 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8426 conforming to the ABI, this would not be possible. So specifying a
8427 @option{-m128bit-long-double} will align @code{long double}
8428 to a 16 byte boundary by padding the @code{long double} with an additional
8431 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8432 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8434 Notice that neither of these options enable any extra precision over the x87
8435 standard of 80 bits for a @code{long double}.
8437 @strong{Warning:} if you override the default value for your target ABI, the
8438 structures and arrays containing @code{long double} will change their size as
8439 well as function calling convention for function taking @code{long double}
8440 will be modified. Hence they will not be binary compatible with arrays or
8441 structures in code compiled without that switch.
8445 @itemx -mno-svr3-shlib
8446 @opindex msvr3-shlib
8447 @opindex mno-svr3-shlib
8448 Control whether GCC places uninitialized local variables into the
8449 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8450 into @code{bss}. These options are meaningful only on System V Release 3.
8454 Use a different function-calling convention, in which functions that
8455 take a fixed number of arguments return with the @code{ret} @var{num}
8456 instruction, which pops their arguments while returning. This saves one
8457 instruction in the caller since there is no need to pop the arguments
8460 You can specify that an individual function is called with this calling
8461 sequence with the function attribute @samp{stdcall}. You can also
8462 override the @option{-mrtd} option by using the function attribute
8463 @samp{cdecl}. @xref{Function Attributes}.
8465 @strong{Warning:} this calling convention is incompatible with the one
8466 normally used on Unix, so you cannot use it if you need to call
8467 libraries compiled with the Unix compiler.
8469 Also, you must provide function prototypes for all functions that
8470 take variable numbers of arguments (including @code{printf});
8471 otherwise incorrect code will be generated for calls to those
8474 In addition, seriously incorrect code will result if you call a
8475 function with too many arguments. (Normally, extra arguments are
8476 harmlessly ignored.)
8478 @item -mregparm=@var{num}
8480 Control how many registers are used to pass integer arguments. By
8481 default, no registers are used to pass arguments, and at most 3
8482 registers can be used. You can control this behavior for a specific
8483 function by using the function attribute @samp{regparm}.
8484 @xref{Function Attributes}.
8486 @strong{Warning:} if you use this switch, and
8487 @var{num} is nonzero, then you must build all modules with the same
8488 value, including any libraries. This includes the system libraries and
8491 @item -mpreferred-stack-boundary=@var{num}
8492 @opindex mpreferred-stack-boundary
8493 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8494 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8495 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8496 size (@option{-Os}), in which case the default is the minimum correct
8497 alignment (4 bytes for x86, and 8 bytes for x86-64).
8499 On Pentium and PentiumPro, @code{double} and @code{long double} values
8500 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8501 suffer significant run time performance penalties. On Pentium III, the
8502 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8503 penalties if it is not 16 byte aligned.
8505 To ensure proper alignment of this values on the stack, the stack boundary
8506 must be as aligned as that required by any value stored on the stack.
8507 Further, every function must be generated such that it keeps the stack
8508 aligned. Thus calling a function compiled with a higher preferred
8509 stack boundary from a function compiled with a lower preferred stack
8510 boundary will most likely misalign the stack. It is recommended that
8511 libraries that use callbacks always use the default setting.
8513 This extra alignment does consume extra stack space, and generally
8514 increases code size. Code that is sensitive to stack space usage, such
8515 as embedded systems and operating system kernels, may want to reduce the
8516 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8534 These switches enable or disable the use of built-in functions that allow
8535 direct access to the MMX, SSE and 3Dnow extensions of the instruction set.
8537 @xref{X86 Built-in Functions}, for details of the functions enabled
8538 and disabled by these switches.
8540 To have SSE/SSE2 instructions generated automatically from floating-point
8541 code, see @option{-mfpmath=sse}.
8544 @itemx -mno-push-args
8546 @opindex mno-push-args
8547 Use PUSH operations to store outgoing parameters. This method is shorter
8548 and usually equally fast as method using SUB/MOV operations and is enabled
8549 by default. In some cases disabling it may improve performance because of
8550 improved scheduling and reduced dependencies.
8552 @item -maccumulate-outgoing-args
8553 @opindex maccumulate-outgoing-args
8554 If enabled, the maximum amount of space required for outgoing arguments will be
8555 computed in the function prologue. This is faster on most modern CPUs
8556 because of reduced dependencies, improved scheduling and reduced stack usage
8557 when preferred stack boundary is not equal to 2. The drawback is a notable
8558 increase in code size. This switch implies @option{-mno-push-args}.
8562 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8563 on thread-safe exception handling must compile and link all code with the
8564 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8565 @option{-D_MT}; when linking, it links in a special thread helper library
8566 @option{-lmingwthrd} which cleans up per thread exception handling data.
8568 @item -mno-align-stringops
8569 @opindex mno-align-stringops
8570 Do not align destination of inlined string operations. This switch reduces
8571 code size and improves performance in case the destination is already aligned,
8572 but gcc don't know about it.
8574 @item -minline-all-stringops
8575 @opindex minline-all-stringops
8576 By default GCC inlines string operations only when destination is known to be
8577 aligned at least to 4 byte boundary. This enables more inlining, increase code
8578 size, but may improve performance of code that depends on fast memcpy, strlen
8579 and memset for short lengths.
8581 @item -momit-leaf-frame-pointer
8582 @opindex momit-leaf-frame-pointer
8583 Don't keep the frame pointer in a register for leaf functions. This
8584 avoids the instructions to save, set up and restore frame pointers and
8585 makes an extra register available in leaf functions. The option
8586 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8587 which might make debugging harder.
8589 @item -mtls-direct-seg-refs
8590 @itemx -mno-tls-direct-seg-refs
8591 @opindex mtls-direct-seg-refs
8592 Controls whether TLS variables may be accessed with offsets from the
8593 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8594 or whether the thread base pointer must be added. Whether or not this
8595 is legal depends on the operating system, and whether it maps the
8596 segment to cover the entire TLS area.
8598 For systems that use GNU libc, the default is on.
8601 These @samp{-m} switches are supported in addition to the above
8602 on AMD x86-64 processors in 64-bit environments.
8609 Generate code for a 32-bit or 64-bit environment.
8610 The 32-bit environment sets int, long and pointer to 32 bits and
8611 generates code that runs on any i386 system.
8612 The 64-bit environment sets int to 32 bits and long and pointer
8613 to 64 bits and generates code for AMD's x86-64 architecture.
8616 @opindex no-red-zone
8617 Do not use a so called red zone for x86-64 code. The red zone is mandated
8618 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8619 stack pointer that will not be modified by signal or interrupt handlers
8620 and therefore can be used for temporary data without adjusting the stack
8621 pointer. The flag @option{-mno-red-zone} disables this red zone.
8623 @item -mcmodel=small
8624 @opindex mcmodel=small
8625 Generate code for the small code model: the program and its symbols must
8626 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8627 Programs can be statically or dynamically linked. This is the default
8630 @item -mcmodel=kernel
8631 @opindex mcmodel=kernel
8632 Generate code for the kernel code model. The kernel runs in the
8633 negative 2 GB of the address space.
8634 This model has to be used for Linux kernel code.
8636 @item -mcmodel=medium
8637 @opindex mcmodel=medium
8638 Generate code for the medium model: The program is linked in the lower 2
8639 GB of the address space but symbols can be located anywhere in the
8640 address space. Programs can be statically or dynamically linked, but
8641 building of shared libraries are not supported with the medium model.
8643 @item -mcmodel=large
8644 @opindex mcmodel=large
8645 Generate code for the large model: This model makes no assumptions
8646 about addresses and sizes of sections. Currently GCC does not implement
8651 @subsection HPPA Options
8652 @cindex HPPA Options
8654 These @samp{-m} options are defined for the HPPA family of computers:
8657 @item -march=@var{architecture-type}
8659 Generate code for the specified architecture. The choices for
8660 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8661 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8662 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8663 architecture option for your machine. Code compiled for lower numbered
8664 architectures will run on higher numbered architectures, but not the
8667 PA 2.0 support currently requires gas snapshot 19990413 or later. The
8668 next release of binutils (current is 2.9.1) will probably contain PA 2.0
8672 @itemx -mpa-risc-1-1
8673 @itemx -mpa-risc-2-0
8674 @opindex mpa-risc-1-0
8675 @opindex mpa-risc-1-1
8676 @opindex mpa-risc-2-0
8677 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8680 @opindex mbig-switch
8681 Generate code suitable for big switch tables. Use this option only if
8682 the assembler/linker complain about out of range branches within a switch
8685 @item -mjump-in-delay
8686 @opindex mjump-in-delay
8687 Fill delay slots of function calls with unconditional jump instructions
8688 by modifying the return pointer for the function call to be the target
8689 of the conditional jump.
8691 @item -mdisable-fpregs
8692 @opindex mdisable-fpregs
8693 Prevent floating point registers from being used in any manner. This is
8694 necessary for compiling kernels which perform lazy context switching of
8695 floating point registers. If you use this option and attempt to perform
8696 floating point operations, the compiler will abort.
8698 @item -mdisable-indexing
8699 @opindex mdisable-indexing
8700 Prevent the compiler from using indexing address modes. This avoids some
8701 rather obscure problems when compiling MIG generated code under MACH@.
8703 @item -mno-space-regs
8704 @opindex mno-space-regs
8705 Generate code that assumes the target has no space registers. This allows
8706 GCC to generate faster indirect calls and use unscaled index address modes.
8708 Such code is suitable for level 0 PA systems and kernels.
8710 @item -mfast-indirect-calls
8711 @opindex mfast-indirect-calls
8712 Generate code that assumes calls never cross space boundaries. This
8713 allows GCC to emit code which performs faster indirect calls.
8715 This option will not work in the presence of shared libraries or nested
8718 @item -mlong-load-store
8719 @opindex mlong-load-store
8720 Generate 3-instruction load and store sequences as sometimes required by
8721 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8724 @item -mportable-runtime
8725 @opindex mportable-runtime
8726 Use the portable calling conventions proposed by HP for ELF systems.
8730 Enable the use of assembler directives only GAS understands.
8732 @item -mschedule=@var{cpu-type}
8734 Schedule code according to the constraints for the machine type
8735 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8736 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8737 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8738 proper scheduling option for your machine. The default scheduling is
8742 @opindex mlinker-opt
8743 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8744 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8745 linkers in which they give bogus error messages when linking some programs.
8748 @opindex msoft-float
8749 Generate output containing library calls for floating point.
8750 @strong{Warning:} the requisite libraries are not available for all HPPA
8751 targets. Normally the facilities of the machine's usual C compiler are
8752 used, but this cannot be done directly in cross-compilation. You must make
8753 your own arrangements to provide suitable library functions for
8754 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8755 does provide software floating point support.
8757 @option{-msoft-float} changes the calling convention in the output file;
8758 therefore, it is only useful if you compile @emph{all} of a program with
8759 this option. In particular, you need to compile @file{libgcc.a}, the
8760 library that comes with GCC, with @option{-msoft-float} in order for
8765 Generate the predefine, @code{_SIO}, for server IO. The default is
8766 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8767 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
8768 options are available under HP-UX and HI-UX.
8772 Use GNU ld specific options. This passes @option{-shared} to ld when
8773 building a shared library. It is the default when GCC is configured,
8774 explicitly or implicitly, with the GNU linker. This option does not
8775 have any affect on which ld is called, it only changes what parameters
8776 are passed to that ld. The ld that is called is determined by the
8777 @option{--with-ld} configure option, gcc's program search path, and
8778 finally by the user's @env{PATH}. The linker used by GCC can be printed
8779 using @samp{which `gcc -print-prog-name=ld`}.
8783 Use HP ld specific options. This passes @option{-b} to ld when building
8784 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8785 links. It is the default when GCC is configured, explicitly or
8786 implicitly, with the HP linker. This option does not have any affect on
8787 which ld is called, it only changes what parameters are passed to that
8788 ld. The ld that is called is determined by the @option{--with-ld}
8789 configure option, gcc's program search path, and finally by the user's
8790 @env{PATH}. The linker used by GCC can be printed using @samp{which
8791 `gcc -print-prog-name=ld`}.
8794 @opindex mno-long-calls
8795 Generate code that uses long call sequences. This ensures that a call
8796 is always able to reach linker generated stubs. The default is to generate
8797 long calls only when the distance from the call site to the beginning
8798 of the function or translation unit, as the case may be, exceeds a
8799 predefined limit set by the branch type being used. The limits for
8800 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8801 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8804 Distances are measured from the beginning of functions when using the
8805 @option{-ffunction-sections} option, or when using the @option{-mgas}
8806 and @option{-mno-portable-runtime} options together under HP-UX with
8809 It is normally not desirable to use this option as it will degrade
8810 performance. However, it may be useful in large applications,
8811 particularly when partial linking is used to build the application.
8813 The types of long calls used depends on the capabilities of the
8814 assembler and linker, and the type of code being generated. The
8815 impact on systems that support long absolute calls, and long pic
8816 symbol-difference or pc-relative calls should be relatively small.
8817 However, an indirect call is used on 32-bit ELF systems in pic code
8818 and it is quite long.
8822 Suppress the generation of link options to search libdld.sl when the
8823 @option{-static} option is specified on HP-UX 10 and later.
8827 The HP-UX implementation of setlocale in libc has a dependency on
8828 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8829 when the @option{-static} option is specified, special link options
8830 are needed to resolve this dependency.
8832 On HP-UX 10 and later, the GCC driver adds the necessary options to
8833 link with libdld.sl when the @option{-static} option is specified.
8834 This causes the resulting binary to be dynamic. On the 64-bit port,
8835 the linkers generate dynamic binaries by default in any case. The
8836 @option{-nolibdld} option can be used to prevent the GCC driver from
8837 adding these link options.
8841 Add support for multithreading with the @dfn{dce thread} library
8842 under HP-UX. This option sets flags for both the preprocessor and
8846 @node Intel 960 Options
8847 @subsection Intel 960 Options
8849 These @samp{-m} options are defined for the Intel 960 implementations:
8852 @item -m@var{cpu-type}
8860 Assume the defaults for the machine type @var{cpu-type} for some of
8861 the other options, including instruction scheduling, floating point
8862 support, and addressing modes. The choices for @var{cpu-type} are
8863 @samp{ka}, @samp{kb}, @samp{mc}, @samp{ca}, @samp{cf},
8864 @samp{sa}, and @samp{sb}.
8871 @opindex msoft-float
8872 The @option{-mnumerics} option indicates that the processor does support
8873 floating-point instructions. The @option{-msoft-float} option indicates
8874 that floating-point support should not be assumed.
8876 @item -mleaf-procedures
8877 @itemx -mno-leaf-procedures
8878 @opindex mleaf-procedures
8879 @opindex mno-leaf-procedures
8880 Do (or do not) attempt to alter leaf procedures to be callable with the
8881 @code{bal} instruction as well as @code{call}. This will result in more
8882 efficient code for explicit calls when the @code{bal} instruction can be
8883 substituted by the assembler or linker, but less efficient code in other
8884 cases, such as calls via function pointers, or using a linker that doesn't
8885 support this optimization.
8888 @itemx -mno-tail-call
8890 @opindex mno-tail-call
8891 Do (or do not) make additional attempts (beyond those of the
8892 machine-independent portions of the compiler) to optimize tail-recursive
8893 calls into branches. You may not want to do this because the detection of
8894 cases where this is not valid is not totally complete. The default is
8895 @option{-mno-tail-call}.
8897 @item -mcomplex-addr
8898 @itemx -mno-complex-addr
8899 @opindex mcomplex-addr
8900 @opindex mno-complex-addr
8901 Assume (or do not assume) that the use of a complex addressing mode is a
8902 win on this implementation of the i960. Complex addressing modes may not
8903 be worthwhile on the K-series, but they definitely are on the C-series.
8904 The default is currently @option{-mcomplex-addr} for all processors except
8908 @itemx -mno-code-align
8909 @opindex mcode-align
8910 @opindex mno-code-align
8911 Align code to 8-byte boundaries for faster fetching (or don't bother).
8912 Currently turned on by default for C-series implementations only.
8915 @item -mclean-linkage
8916 @itemx -mno-clean-linkage
8917 @opindex mclean-linkage
8918 @opindex mno-clean-linkage
8919 These options are not fully implemented.
8923 @itemx -mic2.0-compat
8924 @itemx -mic3.0-compat
8926 @opindex mic2.0-compat
8927 @opindex mic3.0-compat
8928 Enable compatibility with iC960 v2.0 or v3.0.
8932 @opindex masm-compat
8934 Enable compatibility with the iC960 assembler.
8936 @item -mstrict-align
8937 @itemx -mno-strict-align
8938 @opindex mstrict-align
8939 @opindex mno-strict-align
8940 Do not permit (do permit) unaligned accesses.
8944 Enable structure-alignment compatibility with Intel's gcc release version
8945 1.3 (based on gcc 1.37). This option implies @option{-mstrict-align}.
8947 @item -mlong-double-64
8948 @opindex mlong-double-64
8949 Implement type @samp{long double} as 64-bit floating point numbers.
8950 Without the option @samp{long double} is implemented by 80-bit
8951 floating point numbers. The only reason we have it because there is
8952 no 128-bit @samp{long double} support in @samp{fp-bit.c} yet. So it
8953 is only useful for people using soft-float targets. Otherwise, we
8954 should recommend against use of it.
8958 @node DEC Alpha Options
8959 @subsection DEC Alpha Options
8961 These @samp{-m} options are defined for the DEC Alpha implementations:
8964 @item -mno-soft-float
8966 @opindex mno-soft-float
8967 @opindex msoft-float
8968 Use (do not use) the hardware floating-point instructions for
8969 floating-point operations. When @option{-msoft-float} is specified,
8970 functions in @file{libgcc.a} will be used to perform floating-point
8971 operations. Unless they are replaced by routines that emulate the
8972 floating-point operations, or compiled in such a way as to call such
8973 emulations routines, these routines will issue floating-point
8974 operations. If you are compiling for an Alpha without floating-point
8975 operations, you must ensure that the library is built so as not to call
8978 Note that Alpha implementations without floating-point operations are
8979 required to have floating-point registers.
8984 @opindex mno-fp-regs
8985 Generate code that uses (does not use) the floating-point register set.
8986 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8987 register set is not used, floating point operands are passed in integer
8988 registers as if they were integers and floating-point results are passed
8989 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8990 so any function with a floating-point argument or return value called by code
8991 compiled with @option{-mno-fp-regs} must also be compiled with that
8994 A typical use of this option is building a kernel that does not use,
8995 and hence need not save and restore, any floating-point registers.
8999 The Alpha architecture implements floating-point hardware optimized for
9000 maximum performance. It is mostly compliant with the IEEE floating
9001 point standard. However, for full compliance, software assistance is
9002 required. This option generates code fully IEEE compliant code
9003 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9004 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9005 defined during compilation. The resulting code is less efficient but is
9006 able to correctly support denormalized numbers and exceptional IEEE
9007 values such as not-a-number and plus/minus infinity. Other Alpha
9008 compilers call this option @option{-ieee_with_no_inexact}.
9010 @item -mieee-with-inexact
9011 @opindex mieee-with-inexact
9012 This is like @option{-mieee} except the generated code also maintains
9013 the IEEE @var{inexact-flag}. Turning on this option causes the
9014 generated code to implement fully-compliant IEEE math. In addition to
9015 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9016 macro. On some Alpha implementations the resulting code may execute
9017 significantly slower than the code generated by default. Since there is
9018 very little code that depends on the @var{inexact-flag}, you should
9019 normally not specify this option. Other Alpha compilers call this
9020 option @option{-ieee_with_inexact}.
9022 @item -mfp-trap-mode=@var{trap-mode}
9023 @opindex mfp-trap-mode
9024 This option controls what floating-point related traps are enabled.
9025 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9026 The trap mode can be set to one of four values:
9030 This is the default (normal) setting. The only traps that are enabled
9031 are the ones that cannot be disabled in software (e.g., division by zero
9035 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9039 Like @samp{su}, but the instructions are marked to be safe for software
9040 completion (see Alpha architecture manual for details).
9043 Like @samp{su}, but inexact traps are enabled as well.
9046 @item -mfp-rounding-mode=@var{rounding-mode}
9047 @opindex mfp-rounding-mode
9048 Selects the IEEE rounding mode. Other Alpha compilers call this option
9049 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9054 Normal IEEE rounding mode. Floating point numbers are rounded towards
9055 the nearest machine number or towards the even machine number in case
9059 Round towards minus infinity.
9062 Chopped rounding mode. Floating point numbers are rounded towards zero.
9065 Dynamic rounding mode. A field in the floating point control register
9066 (@var{fpcr}, see Alpha architecture reference manual) controls the
9067 rounding mode in effect. The C library initializes this register for
9068 rounding towards plus infinity. Thus, unless your program modifies the
9069 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9072 @item -mtrap-precision=@var{trap-precision}
9073 @opindex mtrap-precision
9074 In the Alpha architecture, floating point traps are imprecise. This
9075 means without software assistance it is impossible to recover from a
9076 floating trap and program execution normally needs to be terminated.
9077 GCC can generate code that can assist operating system trap handlers
9078 in determining the exact location that caused a floating point trap.
9079 Depending on the requirements of an application, different levels of
9080 precisions can be selected:
9084 Program precision. This option is the default and means a trap handler
9085 can only identify which program caused a floating point exception.
9088 Function precision. The trap handler can determine the function that
9089 caused a floating point exception.
9092 Instruction precision. The trap handler can determine the exact
9093 instruction that caused a floating point exception.
9096 Other Alpha compilers provide the equivalent options called
9097 @option{-scope_safe} and @option{-resumption_safe}.
9099 @item -mieee-conformant
9100 @opindex mieee-conformant
9101 This option marks the generated code as IEEE conformant. You must not
9102 use this option unless you also specify @option{-mtrap-precision=i} and either
9103 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9104 is to emit the line @samp{.eflag 48} in the function prologue of the
9105 generated assembly file. Under DEC Unix, this has the effect that
9106 IEEE-conformant math library routines will be linked in.
9108 @item -mbuild-constants
9109 @opindex mbuild-constants
9110 Normally GCC examines a 32- or 64-bit integer constant to
9111 see if it can construct it from smaller constants in two or three
9112 instructions. If it cannot, it will output the constant as a literal and
9113 generate code to load it from the data segment at runtime.
9115 Use this option to require GCC to construct @emph{all} integer constants
9116 using code, even if it takes more instructions (the maximum is six).
9118 You would typically use this option to build a shared library dynamic
9119 loader. Itself a shared library, it must relocate itself in memory
9120 before it can find the variables and constants in its own data segment.
9126 Select whether to generate code to be assembled by the vendor-supplied
9127 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9145 Indicate whether GCC should generate code to use the optional BWX,
9146 CIX, FIX and MAX instruction sets. The default is to use the instruction
9147 sets supported by the CPU type specified via @option{-mcpu=} option or that
9148 of the CPU on which GCC was built if none was specified.
9153 @opindex mfloat-ieee
9154 Generate code that uses (does not use) VAX F and G floating point
9155 arithmetic instead of IEEE single and double precision.
9157 @item -mexplicit-relocs
9158 @itemx -mno-explicit-relocs
9159 @opindex mexplicit-relocs
9160 @opindex mno-explicit-relocs
9161 Older Alpha assemblers provided no way to generate symbol relocations
9162 except via assembler macros. Use of these macros does not allow
9163 optimal instruction scheduling. GNU binutils as of version 2.12
9164 supports a new syntax that allows the compiler to explicitly mark
9165 which relocations should apply to which instructions. This option
9166 is mostly useful for debugging, as GCC detects the capabilities of
9167 the assembler when it is built and sets the default accordingly.
9171 @opindex msmall-data
9172 @opindex mlarge-data
9173 When @option{-mexplicit-relocs} is in effect, static data is
9174 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9175 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9176 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9177 16-bit relocations off of the @code{$gp} register. This limits the
9178 size of the small data area to 64KB, but allows the variables to be
9179 directly accessed via a single instruction.
9181 The default is @option{-mlarge-data}. With this option the data area
9182 is limited to just below 2GB. Programs that require more than 2GB of
9183 data must use @code{malloc} or @code{mmap} to allocate the data in the
9184 heap instead of in the program's data segment.
9186 When generating code for shared libraries, @option{-fpic} implies
9187 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9191 @opindex msmall-text
9192 @opindex mlarge-text
9193 When @option{-msmall-text} is used, the compiler assumes that the
9194 code of the entire program (or shared library) fits in 4MB, and is
9195 thus reachable with a branch instruction. When @option{-msmall-data}
9196 is used, the compiler can assume that all local symbols share the
9197 same @code{$gp} value, and thus reduce the number of instructions
9198 required for a function call from 4 to 1.
9200 The default is @option{-mlarge-text}.
9202 @item -mcpu=@var{cpu_type}
9204 Set the instruction set and instruction scheduling parameters for
9205 machine type @var{cpu_type}. You can specify either the @samp{EV}
9206 style name or the corresponding chip number. GCC supports scheduling
9207 parameters for the EV4, EV5 and EV6 family of processors and will
9208 choose the default values for the instruction set from the processor
9209 you specify. If you do not specify a processor type, GCC will default
9210 to the processor on which the compiler was built.
9212 Supported values for @var{cpu_type} are
9218 Schedules as an EV4 and has no instruction set extensions.
9222 Schedules as an EV5 and has no instruction set extensions.
9226 Schedules as an EV5 and supports the BWX extension.
9231 Schedules as an EV5 and supports the BWX and MAX extensions.
9235 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9239 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9242 @item -mtune=@var{cpu_type}
9244 Set only the instruction scheduling parameters for machine type
9245 @var{cpu_type}. The instruction set is not changed.
9247 @item -mmemory-latency=@var{time}
9248 @opindex mmemory-latency
9249 Sets the latency the scheduler should assume for typical memory
9250 references as seen by the application. This number is highly
9251 dependent on the memory access patterns used by the application
9252 and the size of the external cache on the machine.
9254 Valid options for @var{time} are
9258 A decimal number representing clock cycles.
9264 The compiler contains estimates of the number of clock cycles for
9265 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9266 (also called Dcache, Scache, and Bcache), as well as to main memory.
9267 Note that L3 is only valid for EV5.
9272 @node DEC Alpha/VMS Options
9273 @subsection DEC Alpha/VMS Options
9275 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9278 @item -mvms-return-codes
9279 @opindex mvms-return-codes
9280 Return VMS condition codes from main. The default is to return POSIX
9281 style condition (e.g.@ error) codes.
9284 @node H8/300 Options
9285 @subsection H8/300 Options
9287 These @samp{-m} options are defined for the H8/300 implementations:
9292 Shorten some address references at link time, when possible; uses the
9293 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9294 ld.info, Using ld}, for a fuller description.
9298 Generate code for the H8/300H@.
9302 Generate code for the H8S@.
9306 Generate code for the H8S and H8/300H in the normal mode. This switch
9307 must be used either with -mh or -ms.
9311 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9315 Make @code{int} data 32 bits by default.
9319 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9320 The default for the H8/300H and H8S is to align longs and floats on 4
9322 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9323 This option has no effect on the H8/300.
9327 @subsection SH Options
9329 These @samp{-m} options are defined for the SH implementations:
9334 Generate code for the SH1.
9338 Generate code for the SH2.
9341 Generate code for the SH2e.
9345 Generate code for the SH3.
9349 Generate code for the SH3e.
9353 Generate code for the SH4 without a floating-point unit.
9355 @item -m4-single-only
9356 @opindex m4-single-only
9357 Generate code for the SH4 with a floating-point unit that only
9358 supports single-precision arithmetic.
9362 Generate code for the SH4 assuming the floating-point unit is in
9363 single-precision mode by default.
9367 Generate code for the SH4.
9371 Compile code for the processor in big endian mode.
9375 Compile code for the processor in little endian mode.
9379 Align doubles at 64-bit boundaries. Note that this changes the calling
9380 conventions, and thus some functions from the standard C library will
9381 not work unless you recompile it first with @option{-mdalign}.
9385 Shorten some address references at link time, when possible; uses the
9386 linker option @option{-relax}.
9390 Use 32-bit offsets in @code{switch} tables. The default is to use
9395 Enable the use of the instruction @code{fmovd}.
9399 Comply with the calling conventions defined by Renesas.
9403 Mark the @code{MAC} register as call-clobbered, even if
9404 @option{-mhitachi} is given.
9408 Increase IEEE-compliance of floating-point code.
9412 Dump instruction size and location in the assembly code.
9416 This option is deprecated. It pads structures to multiple of 4 bytes,
9417 which is incompatible with the SH ABI@.
9421 Optimize for space instead of speed. Implied by @option{-Os}.
9425 When generating position-independent code, emit function calls using
9426 the Global Offset Table instead of the Procedure Linkage Table.
9430 Generate a library function call to invalidate instruction cache
9431 entries, after fixing up a trampoline. This library function call
9432 doesn't assume it can write to the whole memory address space. This
9433 is the default when the target is @code{sh-*-linux*}.
9436 @node System V Options
9437 @subsection Options for System V
9439 These additional options are available on System V Release 4 for
9440 compatibility with other compilers on those systems:
9445 Create a shared object.
9446 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
9450 Identify the versions of each tool used by the compiler, in a
9451 @code{.ident} assembler directive in the output.
9455 Refrain from adding @code{.ident} directives to the output file (this is
9458 @item -YP,@var{dirs}
9460 Search the directories @var{dirs}, and no others, for libraries
9461 specified with @option{-l}.
9465 Look in the directory @var{dir} to find the M4 preprocessor.
9466 The assembler uses this option.
9467 @c This is supposed to go with a -Yd for predefined M4 macro files, but
9468 @c the generic assembler that comes with Solaris takes just -Ym.
9471 @node TMS320C3x/C4x Options
9472 @subsection TMS320C3x/C4x Options
9473 @cindex TMS320C3x/C4x Options
9475 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
9479 @item -mcpu=@var{cpu_type}
9481 Set the instruction set, register set, and instruction scheduling
9482 parameters for machine type @var{cpu_type}. Supported values for
9483 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
9484 @samp{c44}. The default is @samp{c40} to generate code for the
9489 @itemx -msmall-memory
9491 @opindex mbig-memory
9493 @opindex msmall-memory
9495 Generates code for the big or small memory model. The small memory
9496 model assumed that all data fits into one 64K word page. At run-time
9497 the data page (DP) register must be set to point to the 64K page
9498 containing the .bss and .data program sections. The big memory model is
9499 the default and requires reloading of the DP register for every direct
9506 Allow (disallow) allocation of general integer operands into the block
9513 Enable (disable) generation of code using decrement and branch,
9514 DBcond(D), instructions. This is enabled by default for the C4x. To be
9515 on the safe side, this is disabled for the C3x, since the maximum
9516 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
9517 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
9518 that it can utilize the decrement and branch instruction, but will give
9519 up if there is more than one memory reference in the loop. Thus a loop
9520 where the loop counter is decremented can generate slightly more
9521 efficient code, in cases where the RPTB instruction cannot be utilized.
9523 @item -mdp-isr-reload
9525 @opindex mdp-isr-reload
9527 Force the DP register to be saved on entry to an interrupt service
9528 routine (ISR), reloaded to point to the data section, and restored on
9529 exit from the ISR@. This should not be required unless someone has
9530 violated the small memory model by modifying the DP register, say within
9537 For the C3x use the 24-bit MPYI instruction for integer multiplies
9538 instead of a library call to guarantee 32-bit results. Note that if one
9539 of the operands is a constant, then the multiplication will be performed
9540 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
9541 then squaring operations are performed inline instead of a library call.
9544 @itemx -mno-fast-fix
9546 @opindex mno-fast-fix
9547 The C3x/C4x FIX instruction to convert a floating point value to an
9548 integer value chooses the nearest integer less than or equal to the
9549 floating point value rather than to the nearest integer. Thus if the
9550 floating point number is negative, the result will be incorrectly
9551 truncated an additional code is necessary to detect and correct this
9552 case. This option can be used to disable generation of the additional
9553 code required to correct the result.
9559 Enable (disable) generation of repeat block sequences using the RPTB
9560 instruction for zero overhead looping. The RPTB construct is only used
9561 for innermost loops that do not call functions or jump across the loop
9562 boundaries. There is no advantage having nested RPTB loops due to the
9563 overhead required to save and restore the RC, RS, and RE registers.
9564 This is enabled by default with @option{-O2}.
9566 @item -mrpts=@var{count}
9570 Enable (disable) the use of the single instruction repeat instruction
9571 RPTS@. If a repeat block contains a single instruction, and the loop
9572 count can be guaranteed to be less than the value @var{count}, GCC will
9573 emit a RPTS instruction instead of a RPTB@. If no value is specified,
9574 then a RPTS will be emitted even if the loop count cannot be determined
9575 at compile time. Note that the repeated instruction following RPTS does
9576 not have to be reloaded from memory each iteration, thus freeing up the
9577 CPU buses for operands. However, since interrupts are blocked by this
9578 instruction, it is disabled by default.
9580 @item -mloop-unsigned
9581 @itemx -mno-loop-unsigned
9582 @opindex mloop-unsigned
9583 @opindex mno-loop-unsigned
9584 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
9585 is @math{2^{31} + 1} since these instructions test if the iteration count is
9586 negative to terminate the loop. If the iteration count is unsigned
9587 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
9588 exceeded. This switch allows an unsigned iteration count.
9592 Try to emit an assembler syntax that the TI assembler (asm30) is happy
9593 with. This also enforces compatibility with the API employed by the TI
9594 C3x C compiler. For example, long doubles are passed as structures
9595 rather than in floating point registers.
9601 Generate code that uses registers (stack) for passing arguments to functions.
9602 By default, arguments are passed in registers where possible rather
9603 than by pushing arguments on to the stack.
9605 @item -mparallel-insns
9606 @itemx -mno-parallel-insns
9607 @opindex mparallel-insns
9608 @opindex mno-parallel-insns
9609 Allow the generation of parallel instructions. This is enabled by
9610 default with @option{-O2}.
9612 @item -mparallel-mpy
9613 @itemx -mno-parallel-mpy
9614 @opindex mparallel-mpy
9615 @opindex mno-parallel-mpy
9616 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
9617 provided @option{-mparallel-insns} is also specified. These instructions have
9618 tight register constraints which can pessimize the code generation
9624 @subsection V850 Options
9625 @cindex V850 Options
9627 These @samp{-m} options are defined for V850 implementations:
9631 @itemx -mno-long-calls
9632 @opindex mlong-calls
9633 @opindex mno-long-calls
9634 Treat all calls as being far away (near). If calls are assumed to be
9635 far away, the compiler will always load the functions address up into a
9636 register, and call indirect through the pointer.
9642 Do not optimize (do optimize) basic blocks that use the same index
9643 pointer 4 or more times to copy pointer into the @code{ep} register, and
9644 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
9645 option is on by default if you optimize.
9647 @item -mno-prolog-function
9648 @itemx -mprolog-function
9649 @opindex mno-prolog-function
9650 @opindex mprolog-function
9651 Do not use (do use) external functions to save and restore registers
9652 at the prologue and epilogue of a function. The external functions
9653 are slower, but use less code space if more than one function saves
9654 the same number of registers. The @option{-mprolog-function} option
9655 is on by default if you optimize.
9659 Try to make the code as small as possible. At present, this just turns
9660 on the @option{-mep} and @option{-mprolog-function} options.
9664 Put static or global variables whose size is @var{n} bytes or less into
9665 the tiny data area that register @code{ep} points to. The tiny data
9666 area can hold up to 256 bytes in total (128 bytes for byte references).
9670 Put static or global variables whose size is @var{n} bytes or less into
9671 the small data area that register @code{gp} points to. The small data
9672 area can hold up to 64 kilobytes.
9676 Put static or global variables whose size is @var{n} bytes or less into
9677 the first 32 kilobytes of memory.
9681 Specify that the target processor is the V850.
9684 @opindex mbig-switch
9685 Generate code suitable for big switch tables. Use this option only if
9686 the assembler/linker complain about out of range branches within a switch
9691 This option will cause r2 and r5 to be used in the code generated by
9692 the compiler. This setting is the default.
9695 @opindex mno-app-regs
9696 This option will cause r2 and r5 to be treated as fixed registers.
9700 Specify that the target processor is the V850E1. The preprocessor
9701 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
9702 this option is used.
9706 Specify that the target processor is the V850E. The preprocessor
9707 constant @samp{__v850e__} will be defined if this option is used.
9709 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
9710 are defined then a default target processor will be chosen and the
9711 relevant @samp{__v850*__} preprocessor constant will be defined.
9713 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
9714 defined, regardless of which processor variant is the target.
9716 @item -mdisable-callt
9717 @opindex mdisable-callt
9718 This option will suppress generation of the CALLT instruction for the
9719 v850e and v850e1 flavors of the v850 architecture. The default is
9720 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
9725 @subsection ARC Options
9728 These options are defined for ARC implementations:
9733 Compile code for little endian mode. This is the default.
9737 Compile code for big endian mode.
9740 @opindex mmangle-cpu
9741 Prepend the name of the cpu to all public symbol names.
9742 In multiple-processor systems, there are many ARC variants with different
9743 instruction and register set characteristics. This flag prevents code
9744 compiled for one cpu to be linked with code compiled for another.
9745 No facility exists for handling variants that are ``almost identical''.
9746 This is an all or nothing option.
9748 @item -mcpu=@var{cpu}
9750 Compile code for ARC variant @var{cpu}.
9751 Which variants are supported depend on the configuration.
9752 All variants support @option{-mcpu=base}, this is the default.
9754 @item -mtext=@var{text-section}
9755 @itemx -mdata=@var{data-section}
9756 @itemx -mrodata=@var{readonly-data-section}
9760 Put functions, data, and readonly data in @var{text-section},
9761 @var{data-section}, and @var{readonly-data-section} respectively
9762 by default. This can be overridden with the @code{section} attribute.
9763 @xref{Variable Attributes}.
9768 @subsection NS32K Options
9769 @cindex NS32K options
9771 These are the @samp{-m} options defined for the 32000 series. The default
9772 values for these options depends on which style of 32000 was selected when
9773 the compiler was configured; the defaults for the most common choices are
9781 Generate output for a 32032. This is the default
9782 when the compiler is configured for 32032 and 32016 based systems.
9788 Generate output for a 32332. This is the default
9789 when the compiler is configured for 32332-based systems.
9795 Generate output for a 32532. This is the default
9796 when the compiler is configured for 32532-based systems.
9800 Generate output containing 32081 instructions for floating point.
9801 This is the default for all systems.
9805 Generate output containing 32381 instructions for floating point. This
9806 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9807 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9811 Try and generate multiply-add floating point instructions @code{polyF}
9812 and @code{dotF}. This option is only available if the @option{-m32381}
9813 option is in effect. Using these instructions requires changes to
9814 register allocation which generally has a negative impact on
9815 performance. This option should only be enabled when compiling code
9816 particularly likely to make heavy use of multiply-add instructions.
9819 @opindex mnomulti-add
9820 Do not try and generate multiply-add floating point instructions
9821 @code{polyF} and @code{dotF}. This is the default on all platforms.
9824 @opindex msoft-float
9825 Generate output containing library calls for floating point.
9826 @strong{Warning:} the requisite libraries may not be available.
9828 @item -mieee-compare
9829 @itemx -mno-ieee-compare
9830 @opindex mieee-compare
9831 @opindex mno-ieee-compare
9832 Control whether or not the compiler uses IEEE floating point
9833 comparisons. These handle correctly the case where the result of a
9834 comparison is unordered.
9835 @strong{Warning:} the requisite kernel support may not be available.
9838 @opindex mnobitfield
9839 Do not use the bit-field instructions. On some machines it is faster to
9840 use shifting and masking operations. This is the default for the pc532.
9844 Do use the bit-field instructions. This is the default for all platforms
9849 Use a different function-calling convention, in which functions
9850 that take a fixed number of arguments return pop their
9851 arguments on return with the @code{ret} instruction.
9853 This calling convention is incompatible with the one normally
9854 used on Unix, so you cannot use it if you need to call libraries
9855 compiled with the Unix compiler.
9857 Also, you must provide function prototypes for all functions that
9858 take variable numbers of arguments (including @code{printf});
9859 otherwise incorrect code will be generated for calls to those
9862 In addition, seriously incorrect code will result if you call a
9863 function with too many arguments. (Normally, extra arguments are
9864 harmlessly ignored.)
9866 This option takes its name from the 680x0 @code{rtd} instruction.
9871 Use a different function-calling convention where the first two arguments
9872 are passed in registers.
9874 This calling convention is incompatible with the one normally
9875 used on Unix, so you cannot use it if you need to call libraries
9876 compiled with the Unix compiler.
9879 @opindex mnoregparam
9880 Do not pass any arguments in registers. This is the default for all
9885 It is OK to use the sb as an index register which is always loaded with
9886 zero. This is the default for the pc532-netbsd target.
9890 The sb register is not available for use or has not been initialized to
9891 zero by the run time system. This is the default for all targets except
9892 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9893 @option{-fpic} is set.
9897 Many ns32000 series addressing modes use displacements of up to 512MB@.
9898 If an address is above 512MB then displacements from zero can not be used.
9899 This option causes code to be generated which can be loaded above 512MB@.
9900 This may be useful for operating systems or ROM code.
9904 Assume code will be loaded in the first 512MB of virtual address space.
9905 This is the default for all platforms.
9911 @subsection AVR Options
9914 These options are defined for AVR implementations:
9917 @item -mmcu=@var{mcu}
9919 Specify ATMEL AVR instruction set or MCU type.
9921 Instruction set avr1 is for the minimal AVR core, not supported by the C
9922 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
9923 attiny11, attiny12, attiny15, attiny28).
9925 Instruction set avr2 (default) is for the classic AVR core with up to
9926 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
9927 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
9928 at90c8534, at90s8535).
9930 Instruction set avr3 is for the classic AVR core with up to 128K program
9931 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
9933 Instruction set avr4 is for the enhanced AVR core with up to 8K program
9934 memory space (MCU types: atmega8, atmega83, atmega85).
9936 Instruction set avr5 is for the enhanced AVR core with up to 128K program
9937 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
9938 atmega64, atmega128, at43usb355, at94k).
9942 Output instruction sizes to the asm file.
9944 @item -minit-stack=@var{N}
9945 @opindex minit-stack
9946 Specify the initial stack address, which may be a symbol or numeric value,
9947 @samp{__stack} is the default.
9949 @item -mno-interrupts
9950 @opindex mno-interrupts
9951 Generated code is not compatible with hardware interrupts.
9952 Code size will be smaller.
9954 @item -mcall-prologues
9955 @opindex mcall-prologues
9956 Functions prologues/epilogues expanded as call to appropriate
9957 subroutines. Code size will be smaller.
9959 @item -mno-tablejump
9960 @opindex mno-tablejump
9961 Do not generate tablejump insns which sometimes increase code size.
9964 @opindex mtiny-stack
9965 Change only the low 8 bits of the stack pointer.
9969 @subsection MCore Options
9970 @cindex MCore options
9972 These are the @samp{-m} options defined for the Motorola M*Core
9980 @opindex mno-hardlit
9981 Inline constants into the code stream if it can be done in two
9982 instructions or less.
9988 Use the divide instruction. (Enabled by default).
9990 @item -mrelax-immediate
9991 @itemx -mno-relax-immediate
9992 @opindex mrelax-immediate
9993 @opindex mno-relax-immediate
9994 Allow arbitrary sized immediates in bit operations.
9996 @item -mwide-bitfields
9997 @itemx -mno-wide-bitfields
9998 @opindex mwide-bitfields
9999 @opindex mno-wide-bitfields
10000 Always treat bit-fields as int-sized.
10002 @item -m4byte-functions
10003 @itemx -mno-4byte-functions
10004 @opindex m4byte-functions
10005 @opindex mno-4byte-functions
10006 Force all functions to be aligned to a four byte boundary.
10008 @item -mcallgraph-data
10009 @itemx -mno-callgraph-data
10010 @opindex mcallgraph-data
10011 @opindex mno-callgraph-data
10012 Emit callgraph information.
10015 @itemx -mno-slow-bytes
10016 @opindex mslow-bytes
10017 @opindex mno-slow-bytes
10018 Prefer word access when reading byte quantities.
10020 @item -mlittle-endian
10021 @itemx -mbig-endian
10022 @opindex mlittle-endian
10023 @opindex mbig-endian
10024 Generate code for a little endian target.
10030 Generate code for the 210 processor.
10033 @node IA-64 Options
10034 @subsection IA-64 Options
10035 @cindex IA-64 Options
10037 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10041 @opindex mbig-endian
10042 Generate code for a big endian target. This is the default for HP-UX@.
10044 @item -mlittle-endian
10045 @opindex mlittle-endian
10046 Generate code for a little endian target. This is the default for AIX5
10052 @opindex mno-gnu-as
10053 Generate (or don't) code for the GNU assembler. This is the default.
10054 @c Also, this is the default if the configure option @option{--with-gnu-as}
10060 @opindex mno-gnu-ld
10061 Generate (or don't) code for the GNU linker. This is the default.
10062 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10067 Generate code that does not use a global pointer register. The result
10068 is not position independent code, and violates the IA-64 ABI@.
10070 @item -mvolatile-asm-stop
10071 @itemx -mno-volatile-asm-stop
10072 @opindex mvolatile-asm-stop
10073 @opindex mno-volatile-asm-stop
10074 Generate (or don't) a stop bit immediately before and after volatile asm
10079 Generate code that works around Itanium B step errata.
10081 @item -mregister-names
10082 @itemx -mno-register-names
10083 @opindex mregister-names
10084 @opindex mno-register-names
10085 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10086 the stacked registers. This may make assembler output more readable.
10092 Disable (or enable) optimizations that use the small data section. This may
10093 be useful for working around optimizer bugs.
10095 @item -mconstant-gp
10096 @opindex mconstant-gp
10097 Generate code that uses a single constant global pointer value. This is
10098 useful when compiling kernel code.
10102 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10103 This is useful when compiling firmware code.
10105 @item -minline-float-divide-min-latency
10106 @opindex minline-float-divide-min-latency
10107 Generate code for inline divides of floating point values
10108 using the minimum latency algorithm.
10110 @item -minline-float-divide-max-throughput
10111 @opindex minline-float-divide-max-throughput
10112 Generate code for inline divides of floating point values
10113 using the maximum throughput algorithm.
10115 @item -minline-int-divide-min-latency
10116 @opindex minline-int-divide-min-latency
10117 Generate code for inline divides of integer values
10118 using the minimum latency algorithm.
10120 @item -minline-int-divide-max-throughput
10121 @opindex minline-int-divide-max-throughput
10122 Generate code for inline divides of integer values
10123 using the maximum throughput algorithm.
10125 @item -mno-dwarf2-asm
10126 @itemx -mdwarf2-asm
10127 @opindex mno-dwarf2-asm
10128 @opindex mdwarf2-asm
10129 Don't (or do) generate assembler code for the DWARF2 line number debugging
10130 info. This may be useful when not using the GNU assembler.
10132 @item -mfixed-range=@var{register-range}
10133 @opindex mfixed-range
10134 Generate code treating the given register range as fixed registers.
10135 A fixed register is one that the register allocator can not use. This is
10136 useful when compiling kernel code. A register range is specified as
10137 two registers separated by a dash. Multiple register ranges can be
10138 specified separated by a comma.
10140 @item -mearly-stop-bits
10141 @itemx -mno-early-stop-bits
10142 @opindex mearly-stop-bits
10143 @opindex mno-early-stop-bits
10144 Allow stop bits to be placed earlier than immediately preceding the
10145 instruction that triggered the stop bit. This can improve instruction
10146 scheduling, but does not always do so.
10150 @subsection D30V Options
10151 @cindex D30V Options
10153 These @samp{-m} options are defined for D30V implementations:
10158 Link the @samp{.text}, @samp{.data}, @samp{.bss}, @samp{.strings},
10159 @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections into external
10160 memory, which starts at location @code{0x80000000}.
10163 @opindex mextmemory
10164 Same as the @option{-mextmem} switch.
10168 Link the @samp{.text} section into onchip text memory, which starts at
10169 location @code{0x0}. Also link @samp{.data}, @samp{.bss},
10170 @samp{.strings}, @samp{.rodata}, @samp{.rodata1}, @samp{.data1} sections
10171 into onchip data memory, which starts at location @code{0x20000000}.
10173 @item -mno-asm-optimize
10174 @itemx -masm-optimize
10175 @opindex mno-asm-optimize
10176 @opindex masm-optimize
10177 Disable (enable) passing @option{-O} to the assembler when optimizing.
10178 The assembler uses the @option{-O} option to automatically parallelize
10179 adjacent short instructions where possible.
10181 @item -mbranch-cost=@var{n}
10182 @opindex mbranch-cost
10183 Increase the internal costs of branches to @var{n}. Higher costs means
10184 that the compiler will issue more instructions to avoid doing a branch.
10187 @item -mcond-exec=@var{n}
10188 @opindex mcond-exec
10189 Specify the maximum number of conditionally executed instructions that
10190 replace a branch. The default is 4.
10193 @node S/390 and zSeries Options
10194 @subsection S/390 and zSeries Options
10195 @cindex S/390 and zSeries Options
10197 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10201 @itemx -msoft-float
10202 @opindex mhard-float
10203 @opindex msoft-float
10204 Use (do not use) the hardware floating-point instructions and registers
10205 for floating-point operations. When @option{-msoft-float} is specified,
10206 functions in @file{libgcc.a} will be used to perform floating-point
10207 operations. When @option{-mhard-float} is specified, the compiler
10208 generates IEEE floating-point instructions. This is the default.
10211 @itemx -mno-backchain
10212 @opindex mbackchain
10213 @opindex mno-backchain
10214 Generate (or do not generate) code which maintains an explicit
10215 backchain within the stack frame that points to the caller's frame.
10216 This is currently needed to allow debugging. The default is to
10217 generate the backchain.
10220 @itemx -mno-small-exec
10221 @opindex msmall-exec
10222 @opindex mno-small-exec
10223 Generate (or do not generate) code using the @code{bras} instruction
10224 to do subroutine calls.
10225 This only works reliably if the total executable size does not
10226 exceed 64k. The default is to use the @code{basr} instruction instead,
10227 which does not have this limitation.
10233 When @option{-m31} is specified, generate code compliant to the
10234 Linux for S/390 ABI@. When @option{-m64} is specified, generate
10235 code compliant to the Linux for zSeries ABI@. This allows GCC in
10236 particular to generate 64-bit instructions. For the @samp{s390}
10237 targets, the default is @option{-m31}, while the @samp{s390x}
10238 targets default to @option{-m64}.
10244 When @option{-mzarch} is specified, generate code using the
10245 instructions available on z/Architecture.
10246 When @option{-mesa} is specified, generate code using the
10247 instructions available on ESA/390. Note that @option{-mesa} is
10248 not possible with @option{-m64}.
10249 When generating code compliant to the Linux for S/390 ABI,
10250 the default is @option{-mesa}. When generating code compliant
10251 to the Linux for zSeries ABI, the default is @option{-mzarch}.
10257 Generate (or do not generate) code using the @code{mvcle} instruction
10258 to perform block moves. When @option{-mno-mvcle} is specified,
10259 use a @code{mvc} loop instead. This is the default.
10265 Print (or do not print) additional debug information when compiling.
10266 The default is to not print debug information.
10268 @item -march=@var{cpu-type}
10270 Generate code that will run on @var{cpu-type}, which is the name of a system
10271 representing a certain processor type. Possible values for
10272 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10273 When generating code using the instructions available on z/Architecture,
10274 the default is @option{-march=z900}. Otherwise, the default is
10275 @option{-march=g5}.
10277 @item -mtune=@var{cpu-type}
10279 Tune to @var{cpu-type} everything applicable about the generated code,
10280 except for the ABI and the set of available instructions.
10281 The list of @var{cpu-type} values is the same as for @option{-march}.
10282 The default is the value used for @option{-march}.
10287 @subsection CRIS Options
10288 @cindex CRIS Options
10290 These options are defined specifically for the CRIS ports.
10293 @item -march=@var{architecture-type}
10294 @itemx -mcpu=@var{architecture-type}
10297 Generate code for the specified architecture. The choices for
10298 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10299 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
10300 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10303 @item -mtune=@var{architecture-type}
10305 Tune to @var{architecture-type} everything applicable about the generated
10306 code, except for the ABI and the set of available instructions. The
10307 choices for @var{architecture-type} are the same as for
10308 @option{-march=@var{architecture-type}}.
10310 @item -mmax-stack-frame=@var{n}
10311 @opindex mmax-stack-frame
10312 Warn when the stack frame of a function exceeds @var{n} bytes.
10314 @item -melinux-stacksize=@var{n}
10315 @opindex melinux-stacksize
10316 Only available with the @samp{cris-axis-aout} target. Arranges for
10317 indications in the program to the kernel loader that the stack of the
10318 program should be set to @var{n} bytes.
10324 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10325 @option{-march=v3} and @option{-march=v8} respectively.
10329 Enable CRIS-specific verbose debug-related information in the assembly
10330 code. This option also has the effect to turn off the @samp{#NO_APP}
10331 formatted-code indicator to the assembler at the beginning of the
10336 Do not use condition-code results from previous instruction; always emit
10337 compare and test instructions before use of condition codes.
10339 @item -mno-side-effects
10340 @opindex mno-side-effects
10341 Do not emit instructions with side-effects in addressing modes other than
10344 @item -mstack-align
10345 @itemx -mno-stack-align
10346 @itemx -mdata-align
10347 @itemx -mno-data-align
10348 @itemx -mconst-align
10349 @itemx -mno-const-align
10350 @opindex mstack-align
10351 @opindex mno-stack-align
10352 @opindex mdata-align
10353 @opindex mno-data-align
10354 @opindex mconst-align
10355 @opindex mno-const-align
10356 These options (no-options) arranges (eliminate arrangements) for the
10357 stack-frame, individual data and constants to be aligned for the maximum
10358 single data access size for the chosen CPU model. The default is to
10359 arrange for 32-bit alignment. ABI details such as structure layout are
10360 not affected by these options.
10368 Similar to the stack- data- and const-align options above, these options
10369 arrange for stack-frame, writable data and constants to all be 32-bit,
10370 16-bit or 8-bit aligned. The default is 32-bit alignment.
10372 @item -mno-prologue-epilogue
10373 @itemx -mprologue-epilogue
10374 @opindex mno-prologue-epilogue
10375 @opindex mprologue-epilogue
10376 With @option{-mno-prologue-epilogue}, the normal function prologue and
10377 epilogue that sets up the stack-frame are omitted and no return
10378 instructions or return sequences are generated in the code. Use this
10379 option only together with visual inspection of the compiled code: no
10380 warnings or errors are generated when call-saved registers must be saved,
10381 or storage for local variable needs to be allocated.
10385 @opindex mno-gotplt
10387 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10388 instruction sequences that load addresses for functions from the PLT part
10389 of the GOT rather than (traditional on other architectures) calls to the
10390 PLT. The default is @option{-mgotplt}.
10394 Legacy no-op option only recognized with the cris-axis-aout target.
10398 Legacy no-op option only recognized with the cris-axis-elf and
10399 cris-axis-linux-gnu targets.
10403 Only recognized with the cris-axis-aout target, where it selects a
10404 GNU/linux-like multilib, include files and instruction set for
10405 @option{-march=v8}.
10409 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10413 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
10414 to link with input-output functions from a simulator library. Code,
10415 initialized data and zero-initialized data are allocated consecutively.
10419 Like @option{-sim}, but pass linker options to locate initialized data at
10420 0x40000000 and zero-initialized data at 0x80000000.
10424 @subsection MMIX Options
10425 @cindex MMIX Options
10427 These options are defined for the MMIX:
10431 @itemx -mno-libfuncs
10433 @opindex mno-libfuncs
10434 Specify that intrinsic library functions are being compiled, passing all
10435 values in registers, no matter the size.
10438 @itemx -mno-epsilon
10440 @opindex mno-epsilon
10441 Generate floating-point comparison instructions that compare with respect
10442 to the @code{rE} epsilon register.
10444 @item -mabi=mmixware
10446 @opindex mabi-mmixware
10448 Generate code that passes function parameters and return values that (in
10449 the called function) are seen as registers @code{$0} and up, as opposed to
10450 the GNU ABI which uses global registers @code{$231} and up.
10452 @item -mzero-extend
10453 @itemx -mno-zero-extend
10454 @opindex mzero-extend
10455 @opindex mno-zero-extend
10456 When reading data from memory in sizes shorter than 64 bits, use (do not
10457 use) zero-extending load instructions by default, rather than
10458 sign-extending ones.
10461 @itemx -mno-knuthdiv
10463 @opindex mno-knuthdiv
10464 Make the result of a division yielding a remainder have the same sign as
10465 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10466 remainder follows the sign of the dividend. Both methods are
10467 arithmetically valid, the latter being almost exclusively used.
10469 @item -mtoplevel-symbols
10470 @itemx -mno-toplevel-symbols
10471 @opindex mtoplevel-symbols
10472 @opindex mno-toplevel-symbols
10473 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10474 code can be used with the @code{PREFIX} assembly directive.
10478 Generate an executable in the ELF format, rather than the default
10479 @samp{mmo} format used by the @command{mmix} simulator.
10481 @item -mbranch-predict
10482 @itemx -mno-branch-predict
10483 @opindex mbranch-predict
10484 @opindex mno-branch-predict
10485 Use (do not use) the probable-branch instructions, when static branch
10486 prediction indicates a probable branch.
10488 @item -mbase-addresses
10489 @itemx -mno-base-addresses
10490 @opindex mbase-addresses
10491 @opindex mno-base-addresses
10492 Generate (do not generate) code that uses @emph{base addresses}. Using a
10493 base address automatically generates a request (handled by the assembler
10494 and the linker) for a constant to be set up in a global register. The
10495 register is used for one or more base address requests within the range 0
10496 to 255 from the value held in the register. The generally leads to short
10497 and fast code, but the number of different data items that can be
10498 addressed is limited. This means that a program that uses lots of static
10499 data may require @option{-mno-base-addresses}.
10501 @item -msingle-exit
10502 @itemx -mno-single-exit
10503 @opindex msingle-exit
10504 @opindex mno-single-exit
10505 Force (do not force) generated code to have a single exit point in each
10509 @node PDP-11 Options
10510 @subsection PDP-11 Options
10511 @cindex PDP-11 Options
10513 These options are defined for the PDP-11:
10518 Use hardware FPP floating point. This is the default. (FIS floating
10519 point on the PDP-11/40 is not supported.)
10522 @opindex msoft-float
10523 Do not use hardware floating point.
10527 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10531 Return floating-point results in memory. This is the default.
10535 Generate code for a PDP-11/40.
10539 Generate code for a PDP-11/45. This is the default.
10543 Generate code for a PDP-11/10.
10545 @item -mbcopy-builtin
10546 @opindex bcopy-builtin
10547 Use inline @code{movstrhi} patterns for copying memory. This is the
10552 Do not use inline @code{movstrhi} patterns for copying memory.
10558 Use 16-bit @code{int}. This is the default.
10564 Use 32-bit @code{int}.
10567 @itemx -mno-float32
10569 @opindex mno-float32
10570 Use 64-bit @code{float}. This is the default.
10575 @opindex mno-float64
10576 Use 32-bit @code{float}.
10580 Use @code{abshi2} pattern. This is the default.
10584 Do not use @code{abshi2} pattern.
10586 @item -mbranch-expensive
10587 @opindex mbranch-expensive
10588 Pretend that branches are expensive. This is for experimenting with
10589 code generation only.
10591 @item -mbranch-cheap
10592 @opindex mbranch-cheap
10593 Do not pretend that branches are expensive. This is the default.
10597 Generate code for a system with split I&D.
10601 Generate code for a system without split I&D. This is the default.
10605 Use Unix assembler syntax. This is the default when configured for
10606 @samp{pdp11-*-bsd}.
10610 Use DEC assembler syntax. This is the default when configured for any
10611 PDP-11 target other than @samp{pdp11-*-bsd}.
10614 @node Xstormy16 Options
10615 @subsection Xstormy16 Options
10616 @cindex Xstormy16 Options
10618 These options are defined for Xstormy16:
10623 Choose startup files and linker script suitable for the simulator.
10627 @subsection FRV Options
10628 @cindex FRV Options
10634 Only use the first 32 general purpose registers.
10639 Use all 64 general purpose registers.
10644 Use only the first 32 floating point registers.
10649 Use all 64 floating point registers
10652 @opindex mhard-float
10654 Use hardware instructions for floating point operations.
10657 @opindex msoft-float
10659 Use library routines for floating point operations.
10664 Dynamically allocate condition code registers.
10669 Do not try to dynamically allocate condition code registers, only
10670 use @code{icc0} and @code{fcc0}.
10675 Change ABI to use double word insns.
10680 Do not use double word instructions.
10685 Use floating point double instructions.
10688 @opindex mno-double
10690 Do not use floating point double instructions.
10695 Use media instructions.
10700 Do not use media instructions.
10705 Use multiply and add/subtract instructions.
10708 @opindex mno-muladd
10710 Do not use multiply and add/subtract instructions.
10712 @item -mlibrary-pic
10713 @opindex mlibrary-pic
10715 Enable PIC support for building libraries
10720 Use only the first four media accumulator registers.
10725 Use all eight media accumulator registers.
10730 Pack VLIW instructions.
10735 Do not pack VLIW instructions.
10738 @opindex mno-eflags
10740 Do not mark ABI switches in e_flags.
10743 @opindex mcond-move
10745 Enable the use of conditional-move instructions (default).
10747 This switch is mainly for debugging the compiler and will likely be removed
10748 in a future version.
10750 @item -mno-cond-move
10751 @opindex mno-cond-move
10753 Disable the use of conditional-move instructions.
10755 This switch is mainly for debugging the compiler and will likely be removed
10756 in a future version.
10761 Enable the use of conditional set instructions (default).
10763 This switch is mainly for debugging the compiler and will likely be removed
10764 in a future version.
10769 Disable the use of conditional set instructions.
10771 This switch is mainly for debugging the compiler and will likely be removed
10772 in a future version.
10775 @opindex mcond-exec
10777 Enable the use of conditional execution (default).
10779 This switch is mainly for debugging the compiler and will likely be removed
10780 in a future version.
10782 @item -mno-cond-exec
10783 @opindex mno-cond-exec
10785 Disable the use of conditional execution.
10787 This switch is mainly for debugging the compiler and will likely be removed
10788 in a future version.
10790 @item -mvliw-branch
10791 @opindex mvliw-branch
10793 Run a pass to pack branches into VLIW instructions (default).
10795 This switch is mainly for debugging the compiler and will likely be removed
10796 in a future version.
10798 @item -mno-vliw-branch
10799 @opindex mno-vliw-branch
10801 Do not run a pass to pack branches into VLIW instructions.
10803 This switch is mainly for debugging the compiler and will likely be removed
10804 in a future version.
10806 @item -mmulti-cond-exec
10807 @opindex mmulti-cond-exec
10809 Enable optimization of @code{&&} and @code{||} in conditional execution
10812 This switch is mainly for debugging the compiler and will likely be removed
10813 in a future version.
10815 @item -mno-multi-cond-exec
10816 @opindex mno-multi-cond-exec
10818 Disable optimization of @code{&&} and @code{||} in conditional execution.
10820 This switch is mainly for debugging the compiler and will likely be removed
10821 in a future version.
10823 @item -mnested-cond-exec
10824 @opindex mnested-cond-exec
10826 Enable nested conditional execution optimizations (default).
10828 This switch is mainly for debugging the compiler and will likely be removed
10829 in a future version.
10831 @item -mno-nested-cond-exec
10832 @opindex mno-nested-cond-exec
10834 Disable nested conditional execution optimizations.
10836 This switch is mainly for debugging the compiler and will likely be removed
10837 in a future version.
10839 @item -mtomcat-stats
10840 @opindex mtomcat-stats
10842 Cause gas to print out tomcat statistics.
10844 @item -mcpu=@var{cpu}
10847 Select the processor type for which to generate code. Possible values are
10848 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
10853 @node Xtensa Options
10854 @subsection Xtensa Options
10855 @cindex Xtensa Options
10857 These options are supported for Xtensa targets:
10861 @itemx -mno-const16
10863 @opindex mno-const16
10864 Enable or disable use of @code{CONST16} instructions for loading
10865 constant values. The @code{CONST16} instruction is currently not a
10866 standard option from Tensilica. When enabled, @code{CONST16}
10867 instructions are always used in place of the standard @code{L32R}
10868 instructions. The use of @code{CONST16} is enabled by default only if
10869 the @code{L32R} instruction is not available.
10872 @itemx -mno-fused-madd
10873 @opindex mfused-madd
10874 @opindex mno-fused-madd
10875 Enable or disable use of fused multiply/add and multiply/subtract
10876 instructions in the floating-point option. This has no effect if the
10877 floating-point option is not also enabled. Disabling fused multiply/add
10878 and multiply/subtract instructions forces the compiler to use separate
10879 instructions for the multiply and add/subtract operations. This may be
10880 desirable in some cases where strict IEEE 754-compliant results are
10881 required: the fused multiply add/subtract instructions do not round the
10882 intermediate result, thereby producing results with @emph{more} bits of
10883 precision than specified by the IEEE standard. Disabling fused multiply
10884 add/subtract instructions also ensures that the program output is not
10885 sensitive to the compiler's ability to combine multiply and add/subtract
10888 @item -mtext-section-literals
10889 @itemx -mno-text-section-literals
10890 @opindex mtext-section-literals
10891 @opindex mno-text-section-literals
10892 Control the treatment of literal pools. The default is
10893 @option{-mno-text-section-literals}, which places literals in a separate
10894 section in the output file. This allows the literal pool to be placed
10895 in a data RAM/ROM, and it also allows the linker to combine literal
10896 pools from separate object files to remove redundant literals and
10897 improve code size. With @option{-mtext-section-literals}, the literals
10898 are interspersed in the text section in order to keep them as close as
10899 possible to their references. This may be necessary for large assembly
10902 @item -mtarget-align
10903 @itemx -mno-target-align
10904 @opindex mtarget-align
10905 @opindex mno-target-align
10906 When this option is enabled, GCC instructs the assembler to
10907 automatically align instructions to reduce branch penalties at the
10908 expense of some code density. The assembler attempts to widen density
10909 instructions to align branch targets and the instructions following call
10910 instructions. If there are not enough preceding safe density
10911 instructions to align a target, no widening will be performed. The
10912 default is @option{-mtarget-align}. These options do not affect the
10913 treatment of auto-aligned instructions like @code{LOOP}, which the
10914 assembler will always align, either by widening density instructions or
10915 by inserting no-op instructions.
10918 @itemx -mno-longcalls
10919 @opindex mlongcalls
10920 @opindex mno-longcalls
10921 When this option is enabled, GCC instructs the assembler to translate
10922 direct calls to indirect calls unless it can determine that the target
10923 of a direct call is in the range allowed by the call instruction. This
10924 translation typically occurs for calls to functions in other source
10925 files. Specifically, the assembler translates a direct @code{CALL}
10926 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
10927 The default is @option{-mno-longcalls}. This option should be used in
10928 programs where the call target can potentially be out of range. This
10929 option is implemented in the assembler, not the compiler, so the
10930 assembly code generated by GCC will still show direct call
10931 instructions---look at the disassembled object code to see the actual
10932 instructions. Note that the assembler will use an indirect call for
10933 every cross-file call, not just those that really will be out of range.
10936 @node Code Gen Options
10937 @section Options for Code Generation Conventions
10938 @cindex code generation conventions
10939 @cindex options, code generation
10940 @cindex run-time options
10942 These machine-independent options control the interface conventions
10943 used in code generation.
10945 Most of them have both positive and negative forms; the negative form
10946 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
10947 one of the forms is listed---the one which is not the default. You
10948 can figure out the other form by either removing @samp{no-} or adding
10952 @item -fbounds-check
10953 @opindex fbounds-check
10954 For front-ends that support it, generate additional code to check that
10955 indices used to access arrays are within the declared range. This is
10956 currently only supported by the Java and Fortran 77 front-ends, where
10957 this option defaults to true and false respectively.
10961 This option generates traps for signed overflow on addition, subtraction,
10962 multiplication operations.
10966 This option instructs the compiler to assume that signed arithmetic
10967 overflow of addition, subtraction and multiplication wraps around
10968 using twos-complement representation. This flag enables some optimizations
10969 and disables other. This option is enabled by default for the Java
10970 front-end, as required by the Java language specification.
10973 @opindex fexceptions
10974 Enable exception handling. Generates extra code needed to propagate
10975 exceptions. For some targets, this implies GCC will generate frame
10976 unwind information for all functions, which can produce significant data
10977 size overhead, although it does not affect execution. If you do not
10978 specify this option, GCC will enable it by default for languages like
10979 C++ which normally require exception handling, and disable it for
10980 languages like C that do not normally require it. However, you may need
10981 to enable this option when compiling C code that needs to interoperate
10982 properly with exception handlers written in C++. You may also wish to
10983 disable this option if you are compiling older C++ programs that don't
10984 use exception handling.
10986 @item -fnon-call-exceptions
10987 @opindex fnon-call-exceptions
10988 Generate code that allows trapping instructions to throw exceptions.
10989 Note that this requires platform-specific runtime support that does
10990 not exist everywhere. Moreover, it only allows @emph{trapping}
10991 instructions to throw exceptions, i.e.@: memory references or floating
10992 point instructions. It does not allow exceptions to be thrown from
10993 arbitrary signal handlers such as @code{SIGALRM}.
10995 @item -funwind-tables
10996 @opindex funwind-tables
10997 Similar to @option{-fexceptions}, except that it will just generate any needed
10998 static data, but will not affect the generated code in any other way.
10999 You will normally not enable this option; instead, a language processor
11000 that needs this handling would enable it on your behalf.
11002 @item -fasynchronous-unwind-tables
11003 @opindex funwind-tables
11004 Generate unwind table in dwarf2 format, if supported by target machine. The
11005 table is exact at each instruction boundary, so it can be used for stack
11006 unwinding from asynchronous events (such as debugger or garbage collector).
11008 @item -fpcc-struct-return
11009 @opindex fpcc-struct-return
11010 Return ``short'' @code{struct} and @code{union} values in memory like
11011 longer ones, rather than in registers. This convention is less
11012 efficient, but it has the advantage of allowing intercallability between
11013 GCC-compiled files and files compiled with other compilers, particularly
11014 the Portable C Compiler (pcc).
11016 The precise convention for returning structures in memory depends
11017 on the target configuration macros.
11019 Short structures and unions are those whose size and alignment match
11020 that of some integer type.
11022 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11023 switch is not binary compatible with code compiled with the
11024 @option{-freg-struct-return} switch.
11025 Use it to conform to a non-default application binary interface.
11027 @item -freg-struct-return
11028 @opindex freg-struct-return
11029 Return @code{struct} and @code{union} values in registers when possible.
11030 This is more efficient for small structures than
11031 @option{-fpcc-struct-return}.
11033 If you specify neither @option{-fpcc-struct-return} nor
11034 @option{-freg-struct-return}, GCC defaults to whichever convention is
11035 standard for the target. If there is no standard convention, GCC
11036 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11037 the principal compiler. In those cases, we can choose the standard, and
11038 we chose the more efficient register return alternative.
11040 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11041 switch is not binary compatible with code compiled with the
11042 @option{-fpcc-struct-return} switch.
11043 Use it to conform to a non-default application binary interface.
11045 @item -fshort-enums
11046 @opindex fshort-enums
11047 Allocate to an @code{enum} type only as many bytes as it needs for the
11048 declared range of possible values. Specifically, the @code{enum} type
11049 will be equivalent to the smallest integer type which has enough room.
11051 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11052 code that is not binary compatible with code generated without that switch.
11053 Use it to conform to a non-default application binary interface.
11055 @item -fshort-double
11056 @opindex fshort-double
11057 Use the same size for @code{double} as for @code{float}.
11059 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11060 code that is not binary compatible with code generated without that switch.
11061 Use it to conform to a non-default application binary interface.
11063 @item -fshort-wchar
11064 @opindex fshort-wchar
11065 Override the underlying type for @samp{wchar_t} to be @samp{short
11066 unsigned int} instead of the default for the target. This option is
11067 useful for building programs to run under WINE@.
11069 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11070 code that is not binary compatible with code generated without that switch.
11071 Use it to conform to a non-default application binary interface.
11073 @item -fshared-data
11074 @opindex fshared-data
11075 Requests that the data and non-@code{const} variables of this
11076 compilation be shared data rather than private data. The distinction
11077 makes sense only on certain operating systems, where shared data is
11078 shared between processes running the same program, while private data
11079 exists in one copy per process.
11082 @opindex fno-common
11083 In C, allocate even uninitialized global variables in the data section of the
11084 object file, rather than generating them as common blocks. This has the
11085 effect that if the same variable is declared (without @code{extern}) in
11086 two different compilations, you will get an error when you link them.
11087 The only reason this might be useful is if you wish to verify that the
11088 program will work on other systems which always work this way.
11092 Ignore the @samp{#ident} directive.
11094 @item -fno-gnu-linker
11095 @opindex fno-gnu-linker
11096 Do not output global initializations (such as C++ constructors and
11097 destructors) in the form used by the GNU linker (on systems where the GNU
11098 linker is the standard method of handling them). Use this option when
11099 you want to use a non-GNU linker, which also requires using the
11100 @command{collect2} program to make sure the system linker includes
11101 constructors and destructors. (@command{collect2} is included in the GCC
11102 distribution.) For systems which @emph{must} use @command{collect2}, the
11103 compiler driver @command{gcc} is configured to do this automatically.
11105 @item -finhibit-size-directive
11106 @opindex finhibit-size-directive
11107 Don't output a @code{.size} assembler directive, or anything else that
11108 would cause trouble if the function is split in the middle, and the
11109 two halves are placed at locations far apart in memory. This option is
11110 used when compiling @file{crtstuff.c}; you should not need to use it
11113 @item -fverbose-asm
11114 @opindex fverbose-asm
11115 Put extra commentary information in the generated assembly code to
11116 make it more readable. This option is generally only of use to those
11117 who actually need to read the generated assembly code (perhaps while
11118 debugging the compiler itself).
11120 @option{-fno-verbose-asm}, the default, causes the
11121 extra information to be omitted and is useful when comparing two assembler
11126 @cindex global offset table
11128 Generate position-independent code (PIC) suitable for use in a shared
11129 library, if supported for the target machine. Such code accesses all
11130 constant addresses through a global offset table (GOT)@. The dynamic
11131 loader resolves the GOT entries when the program starts (the dynamic
11132 loader is not part of GCC; it is part of the operating system). If
11133 the GOT size for the linked executable exceeds a machine-specific
11134 maximum size, you get an error message from the linker indicating that
11135 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11136 instead. (These maximums are 16k on the m88k, 8k on the SPARC, and 32k
11137 on the m68k and RS/6000. The 386 has no such limit.)
11139 Position-independent code requires special support, and therefore works
11140 only on certain machines. For the 386, GCC supports PIC for System V
11141 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11142 position-independent.
11146 If supported for the target machine, emit position-independent code,
11147 suitable for dynamic linking and avoiding any limit on the size of the
11148 global offset table. This option makes a difference on the m68k, m88k,
11151 Position-independent code requires special support, and therefore works
11152 only on certain machines.
11158 These options are similar to @option{-fpic} and @option{-fPIC}, but
11159 generated position independent code can be only linked into executables.
11160 Usually these options are used when @option{-pie} GCC option will be
11161 used during linking.
11163 @item -ffixed-@var{reg}
11165 Treat the register named @var{reg} as a fixed register; generated code
11166 should never refer to it (except perhaps as a stack pointer, frame
11167 pointer or in some other fixed role).
11169 @var{reg} must be the name of a register. The register names accepted
11170 are machine-specific and are defined in the @code{REGISTER_NAMES}
11171 macro in the machine description macro file.
11173 This flag does not have a negative form, because it specifies a
11176 @item -fcall-used-@var{reg}
11177 @opindex fcall-used
11178 Treat the register named @var{reg} as an allocable register that is
11179 clobbered by function calls. It may be allocated for temporaries or
11180 variables that do not live across a call. Functions compiled this way
11181 will not save and restore the register @var{reg}.
11183 It is an error to used this flag with the frame pointer or stack pointer.
11184 Use of this flag for other registers that have fixed pervasive roles in
11185 the machine's execution model will produce disastrous results.
11187 This flag does not have a negative form, because it specifies a
11190 @item -fcall-saved-@var{reg}
11191 @opindex fcall-saved
11192 Treat the register named @var{reg} as an allocable register saved by
11193 functions. It may be allocated even for temporaries or variables that
11194 live across a call. Functions compiled this way will save and restore
11195 the register @var{reg} if they use it.
11197 It is an error to used this flag with the frame pointer or stack pointer.
11198 Use of this flag for other registers that have fixed pervasive roles in
11199 the machine's execution model will produce disastrous results.
11201 A different sort of disaster will result from the use of this flag for
11202 a register in which function values may be returned.
11204 This flag does not have a negative form, because it specifies a
11207 @item -fpack-struct
11208 @opindex fpack-struct
11209 Pack all structure members together without holes.
11211 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11212 code that is not binary compatible with code generated without that switch.
11213 Additionally, it makes the code suboptimal.
11214 Use it to conform to a non-default application binary interface.
11216 @item -finstrument-functions
11217 @opindex finstrument-functions
11218 Generate instrumentation calls for entry and exit to functions. Just
11219 after function entry and just before function exit, the following
11220 profiling functions will be called with the address of the current
11221 function and its call site. (On some platforms,
11222 @code{__builtin_return_address} does not work beyond the current
11223 function, so the call site information may not be available to the
11224 profiling functions otherwise.)
11227 void __cyg_profile_func_enter (void *this_fn,
11229 void __cyg_profile_func_exit (void *this_fn,
11233 The first argument is the address of the start of the current function,
11234 which may be looked up exactly in the symbol table.
11236 This instrumentation is also done for functions expanded inline in other
11237 functions. The profiling calls will indicate where, conceptually, the
11238 inline function is entered and exited. This means that addressable
11239 versions of such functions must be available. If all your uses of a
11240 function are expanded inline, this may mean an additional expansion of
11241 code size. If you use @samp{extern inline} in your C code, an
11242 addressable version of such functions must be provided. (This is
11243 normally the case anyways, but if you get lucky and the optimizer always
11244 expands the functions inline, you might have gotten away without
11245 providing static copies.)
11247 A function may be given the attribute @code{no_instrument_function}, in
11248 which case this instrumentation will not be done. This can be used, for
11249 example, for the profiling functions listed above, high-priority
11250 interrupt routines, and any functions from which the profiling functions
11251 cannot safely be called (perhaps signal handlers, if the profiling
11252 routines generate output or allocate memory).
11254 @item -fstack-check
11255 @opindex fstack-check
11256 Generate code to verify that you do not go beyond the boundary of the
11257 stack. You should specify this flag if you are running in an
11258 environment with multiple threads, but only rarely need to specify it in
11259 a single-threaded environment since stack overflow is automatically
11260 detected on nearly all systems if there is only one stack.
11262 Note that this switch does not actually cause checking to be done; the
11263 operating system must do that. The switch causes generation of code
11264 to ensure that the operating system sees the stack being extended.
11266 @item -fstack-limit-register=@var{reg}
11267 @itemx -fstack-limit-symbol=@var{sym}
11268 @itemx -fno-stack-limit
11269 @opindex fstack-limit-register
11270 @opindex fstack-limit-symbol
11271 @opindex fno-stack-limit
11272 Generate code to ensure that the stack does not grow beyond a certain value,
11273 either the value of a register or the address of a symbol. If the stack
11274 would grow beyond the value, a signal is raised. For most targets,
11275 the signal is raised before the stack overruns the boundary, so
11276 it is possible to catch the signal without taking special precautions.
11278 For instance, if the stack starts at absolute address @samp{0x80000000}
11279 and grows downwards, you can use the flags
11280 @option{-fstack-limit-symbol=__stack_limit} and
11281 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11282 of 128KB@. Note that this may only work with the GNU linker.
11284 @cindex aliasing of parameters
11285 @cindex parameters, aliased
11286 @item -fargument-alias
11287 @itemx -fargument-noalias
11288 @itemx -fargument-noalias-global
11289 @opindex fargument-alias
11290 @opindex fargument-noalias
11291 @opindex fargument-noalias-global
11292 Specify the possible relationships among parameters and between
11293 parameters and global data.
11295 @option{-fargument-alias} specifies that arguments (parameters) may
11296 alias each other and may alias global storage.@*
11297 @option{-fargument-noalias} specifies that arguments do not alias
11298 each other, but may alias global storage.@*
11299 @option{-fargument-noalias-global} specifies that arguments do not
11300 alias each other and do not alias global storage.
11302 Each language will automatically use whatever option is required by
11303 the language standard. You should not need to use these options yourself.
11305 @item -fleading-underscore
11306 @opindex fleading-underscore
11307 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11308 change the way C symbols are represented in the object file. One use
11309 is to help link with legacy assembly code.
11311 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11312 generate code that is not binary compatible with code generated without that
11313 switch. Use it to conform to a non-default application binary interface.
11314 Not all targets provide complete support for this switch.
11316 @item -ftls-model=@var{model}
11317 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11318 The @var{model} argument should be one of @code{global-dynamic},
11319 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11321 The default without @option{-fpic} is @code{initial-exec}; with
11322 @option{-fpic} the default is @code{global-dynamic}.
11327 @node Environment Variables
11328 @section Environment Variables Affecting GCC
11329 @cindex environment variables
11331 @c man begin ENVIRONMENT
11332 This section describes several environment variables that affect how GCC
11333 operates. Some of them work by specifying directories or prefixes to use
11334 when searching for various kinds of files. Some are used to specify other
11335 aspects of the compilation environment.
11337 Note that you can also specify places to search using options such as
11338 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11339 take precedence over places specified using environment variables, which
11340 in turn take precedence over those specified by the configuration of GCC@.
11341 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11342 GNU Compiler Collection (GCC) Internals}.
11347 @c @itemx LC_COLLATE
11349 @c @itemx LC_MONETARY
11350 @c @itemx LC_NUMERIC
11355 @c @findex LC_COLLATE
11356 @findex LC_MESSAGES
11357 @c @findex LC_MONETARY
11358 @c @findex LC_NUMERIC
11362 These environment variables control the way that GCC uses
11363 localization information that allow GCC to work with different
11364 national conventions. GCC inspects the locale categories
11365 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11366 so. These locale categories can be set to any value supported by your
11367 installation. A typical value is @samp{en_UK} for English in the United
11370 The @env{LC_CTYPE} environment variable specifies character
11371 classification. GCC uses it to determine the character boundaries in
11372 a string; this is needed for some multibyte encodings that contain quote
11373 and escape characters that would otherwise be interpreted as a string
11376 The @env{LC_MESSAGES} environment variable specifies the language to
11377 use in diagnostic messages.
11379 If the @env{LC_ALL} environment variable is set, it overrides the value
11380 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11381 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11382 environment variable. If none of these variables are set, GCC
11383 defaults to traditional C English behavior.
11387 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11388 files. GCC uses temporary files to hold the output of one stage of
11389 compilation which is to be used as input to the next stage: for example,
11390 the output of the preprocessor, which is the input to the compiler
11393 @item GCC_EXEC_PREFIX
11394 @findex GCC_EXEC_PREFIX
11395 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11396 names of the subprograms executed by the compiler. No slash is added
11397 when this prefix is combined with the name of a subprogram, but you can
11398 specify a prefix that ends with a slash if you wish.
11400 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11401 an appropriate prefix to use based on the pathname it was invoked with.
11403 If GCC cannot find the subprogram using the specified prefix, it
11404 tries looking in the usual places for the subprogram.
11406 The default value of @env{GCC_EXEC_PREFIX} is
11407 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11408 of @code{prefix} when you ran the @file{configure} script.
11410 Other prefixes specified with @option{-B} take precedence over this prefix.
11412 This prefix is also used for finding files such as @file{crt0.o} that are
11415 In addition, the prefix is used in an unusual way in finding the
11416 directories to search for header files. For each of the standard
11417 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11418 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11419 replacing that beginning with the specified prefix to produce an
11420 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11421 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11422 These alternate directories are searched first; the standard directories
11425 @item COMPILER_PATH
11426 @findex COMPILER_PATH
11427 The value of @env{COMPILER_PATH} is a colon-separated list of
11428 directories, much like @env{PATH}. GCC tries the directories thus
11429 specified when searching for subprograms, if it can't find the
11430 subprograms using @env{GCC_EXEC_PREFIX}.
11433 @findex LIBRARY_PATH
11434 The value of @env{LIBRARY_PATH} is a colon-separated list of
11435 directories, much like @env{PATH}. When configured as a native compiler,
11436 GCC tries the directories thus specified when searching for special
11437 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11438 using GCC also uses these directories when searching for ordinary
11439 libraries for the @option{-l} option (but directories specified with
11440 @option{-L} come first).
11444 @cindex locale definition
11445 This variable is used to pass locale information to the compiler. One way in
11446 which this information is used is to determine the character set to be used
11447 when character literals, string literals and comments are parsed in C and C++.
11448 When the compiler is configured to allow multibyte characters,
11449 the following values for @env{LANG} are recognized:
11453 Recognize JIS characters.
11455 Recognize SJIS characters.
11457 Recognize EUCJP characters.
11460 If @env{LANG} is not defined, or if it has some other value, then the
11461 compiler will use mblen and mbtowc as defined by the default locale to
11462 recognize and translate multibyte characters.
11466 Some additional environments variables affect the behavior of the
11469 @include cppenv.texi
11473 @node Precompiled Headers
11474 @section Using Precompiled Headers
11475 @cindex precompiled headers
11476 @cindex speed of compilation
11478 Often large projects have many header files that are included in every
11479 source file. The time the compiler takes to process these header files
11480 over and over again can account for nearly all of the time required to
11481 build the project. To make builds faster, GCC allows users to
11482 `precompile' a header file; then, if builds can use the precompiled
11483 header file they will be much faster.
11485 To create a precompiled header file, simply compile it as you would any
11486 other file, if necessary using the @option{-x} option to make the driver
11487 treat it as a C or C++ header file. You will probably want to use a
11488 tool like @command{make} to keep the precompiled header up-to-date when
11489 the headers it contains change.
11491 A precompiled header file will be searched for when @code{#include} is
11492 seen in the compilation. As it searches for the included file
11493 (@pxref{Search Path,,Search Path,cpp.info,The C Preprocessor}) the
11494 compiler looks for a precompiled header in each directory just before it
11495 looks for the include file in that directory. The name searched for is
11496 the name specified in the @code{#include} with @samp{.gch} appended. If
11497 the precompiled header file can't be used, it is ignored.
11499 For instance, if you have @code{#include "all.h"}, and you have
11500 @file{all.h.gch} in the same directory as @file{all.h}, then the
11501 precompiled header file will be used if possible, and the original
11502 header will be used otherwise.
11504 Alternatively, you might decide to put the precompiled header file in a
11505 directory and use @option{-I} to ensure that directory is searched
11506 before (or instead of) the directory containing the original header.
11507 Then, if you want to check that the precompiled header file is always
11508 used, you can put a file of the same name as the original header in this
11509 directory containing an @code{#error} command.
11511 This also works with @option{-include}. So yet another way to use
11512 precompiled headers, good for projects not designed with precompiled
11513 header files in mind, is to simply take most of the header files used by
11514 a project, include them from another header file, precompile that header
11515 file, and @option{-include} the precompiled header. If the header files
11516 have guards against multiple inclusion, they will be skipped because
11517 they've already been included (in the precompiled header).
11519 If you need to precompile the same header file for different
11520 languages, targets, or compiler options, you can instead make a
11521 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11522 header in the directory. (It doesn't matter what you call the files
11523 in the directory, every precompiled header in the directory will be
11524 considered.) The first precompiled header encountered in the
11525 directory that is valid for this compilation will be used; they're
11526 searched in no particular order.
11528 There are many other possibilities, limited only by your imagination,
11529 good sense, and the constraints of your build system.
11531 A precompiled header file can be used only when these conditions apply:
11535 Only one precompiled header can be used in a particular compilation.
11537 A precompiled header can't be used once the first C token is seen. You
11538 can have preprocessor directives before a precompiled header; you can
11539 even include a precompiled header from inside another header, so long as
11540 there are no C tokens before the @code{#include}.
11542 The precompiled header file must be produced for the same language as
11543 the current compilation. You can't use a C precompiled header for a C++
11546 The precompiled header file must be produced by the same compiler
11547 version and configuration as the current compilation is using.
11548 The easiest way to guarantee this is to use the same compiler binary
11549 for creating and using precompiled headers.
11551 Any macros defined before the precompiled header (including with
11552 @option{-D}) must either be defined in the same way as when the
11553 precompiled header was generated, or must not affect the precompiled
11554 header, which usually means that the they don't appear in the
11555 precompiled header at all.
11557 Certain command-line options must be defined in the same way as when the
11558 precompiled header was generated. At present, it's not clear which
11559 options are safe to change and which are not; the safest choice is to
11560 use exactly the same options when generating and using the precompiled
11564 For all of these but the last, the compiler will automatically ignore
11565 the precompiled header if the conditions aren't met. For the last item,
11566 some option changes will cause the precompiled header to be rejected,
11567 but not all incompatible option combinations have yet been found. If
11568 you find a new incompatible combination, please consider filing a bug
11569 report, see @ref{Bugs}.
11571 @node Running Protoize
11572 @section Running Protoize
11574 The program @code{protoize} is an optional part of GCC@. You can use
11575 it to add prototypes to a program, thus converting the program to ISO
11576 C in one respect. The companion program @code{unprotoize} does the
11577 reverse: it removes argument types from any prototypes that are found.
11579 When you run these programs, you must specify a set of source files as
11580 command line arguments. The conversion programs start out by compiling
11581 these files to see what functions they define. The information gathered
11582 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
11584 After scanning comes actual conversion. The specified files are all
11585 eligible to be converted; any files they include (whether sources or
11586 just headers) are eligible as well.
11588 But not all the eligible files are converted. By default,
11589 @code{protoize} and @code{unprotoize} convert only source and header
11590 files in the current directory. You can specify additional directories
11591 whose files should be converted with the @option{-d @var{directory}}
11592 option. You can also specify particular files to exclude with the
11593 @option{-x @var{file}} option. A file is converted if it is eligible, its
11594 directory name matches one of the specified directory names, and its
11595 name within the directory has not been excluded.
11597 Basic conversion with @code{protoize} consists of rewriting most
11598 function definitions and function declarations to specify the types of
11599 the arguments. The only ones not rewritten are those for varargs
11602 @code{protoize} optionally inserts prototype declarations at the
11603 beginning of the source file, to make them available for any calls that
11604 precede the function's definition. Or it can insert prototype
11605 declarations with block scope in the blocks where undeclared functions
11608 Basic conversion with @code{unprotoize} consists of rewriting most
11609 function declarations to remove any argument types, and rewriting
11610 function definitions to the old-style pre-ISO form.
11612 Both conversion programs print a warning for any function declaration or
11613 definition that they can't convert. You can suppress these warnings
11616 The output from @code{protoize} or @code{unprotoize} replaces the
11617 original source file. The original file is renamed to a name ending
11618 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
11619 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
11620 for DOS) file already exists, then the source file is simply discarded.
11622 @code{protoize} and @code{unprotoize} both depend on GCC itself to
11623 scan the program and collect information about the functions it uses.
11624 So neither of these programs will work until GCC is installed.
11626 Here is a table of the options you can use with @code{protoize} and
11627 @code{unprotoize}. Each option works with both programs unless
11631 @item -B @var{directory}
11632 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
11633 usual directory (normally @file{/usr/local/lib}). This file contains
11634 prototype information about standard system functions. This option
11635 applies only to @code{protoize}.
11637 @item -c @var{compilation-options}
11638 Use @var{compilation-options} as the options when running @command{gcc} to
11639 produce the @samp{.X} files. The special option @option{-aux-info} is
11640 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
11642 Note that the compilation options must be given as a single argument to
11643 @code{protoize} or @code{unprotoize}. If you want to specify several
11644 @command{gcc} options, you must quote the entire set of compilation options
11645 to make them a single word in the shell.
11647 There are certain @command{gcc} arguments that you cannot use, because they
11648 would produce the wrong kind of output. These include @option{-g},
11649 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
11650 the @var{compilation-options}, they are ignored.
11653 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
11654 systems) instead of @samp{.c}. This is convenient if you are converting
11655 a C program to C++. This option applies only to @code{protoize}.
11658 Add explicit global declarations. This means inserting explicit
11659 declarations at the beginning of each source file for each function
11660 that is called in the file and was not declared. These declarations
11661 precede the first function definition that contains a call to an
11662 undeclared function. This option applies only to @code{protoize}.
11664 @item -i @var{string}
11665 Indent old-style parameter declarations with the string @var{string}.
11666 This option applies only to @code{protoize}.
11668 @code{unprotoize} converts prototyped function definitions to old-style
11669 function definitions, where the arguments are declared between the
11670 argument list and the initial @samp{@{}. By default, @code{unprotoize}
11671 uses five spaces as the indentation. If you want to indent with just
11672 one space instead, use @option{-i " "}.
11675 Keep the @samp{.X} files. Normally, they are deleted after conversion
11679 Add explicit local declarations. @code{protoize} with @option{-l} inserts
11680 a prototype declaration for each function in each block which calls the
11681 function without any declaration. This option applies only to
11685 Make no real changes. This mode just prints information about the conversions
11686 that would have been done without @option{-n}.
11689 Make no @samp{.save} files. The original files are simply deleted.
11690 Use this option with caution.
11692 @item -p @var{program}
11693 Use the program @var{program} as the compiler. Normally, the name
11694 @file{gcc} is used.
11697 Work quietly. Most warnings are suppressed.
11700 Print the version number, just like @option{-v} for @command{gcc}.
11703 If you need special compiler options to compile one of your program's
11704 source files, then you should generate that file's @samp{.X} file
11705 specially, by running @command{gcc} on that source file with the
11706 appropriate options and the option @option{-aux-info}. Then run
11707 @code{protoize} on the entire set of files. @code{protoize} will use
11708 the existing @samp{.X} file because it is newer than the source file.
11712 gcc -Dfoo=bar file1.c -aux-info file1.X
11717 You need to include the special files along with the rest in the
11718 @code{protoize} command, even though their @samp{.X} files already
11719 exist, because otherwise they won't get converted.
11721 @xref{Protoize Caveats}, for more information on how to use
11722 @code{protoize} successfully.