1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @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} -combine -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}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -Wabi -Wctor-dtor-privacy @gol
186 -Wnon-virtual-dtor -Wreorder @gol
187 -Weffc++ -Wno-deprecated @gol
188 -Wno-non-template-friend -Wold-style-cast @gol
189 -Woverloaded-virtual -Wno-pmf-conversions @gol
190 -Wsign-promo -Wsynth}
192 @item Objective-C Language Options
193 @xref{Objective-C Dialect Options,,Options Controlling Objective-C Dialect}.
195 -fconstant-string-class=@var{class-name} @gol
196 -fgnu-runtime -fnext-runtime @gol
197 -fno-nil-receivers @gol
198 -fobjc-exceptions @gol
199 -freplace-objc-classes @gol
202 -Wno-protocol -Wselector -Wundeclared-selector}
204 @item Language Independent Options
205 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
206 @gccoptlist{-fmessage-length=@var{n} @gol
207 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
209 @item Warning Options
210 @xref{Warning Options,,Options to Request or Suppress Warnings}.
211 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
212 -w -Wextra -Wall -Waggregate-return @gol
213 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
214 -Wconversion -Wno-deprecated-declarations @gol
215 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
216 -Werror -Werror-implicit-function-declaration @gol
217 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
218 -Wno-format-extra-args -Wformat-nonliteral @gol
219 -Wformat-security -Wformat-y2k @gol
220 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
221 -Wimport -Wno-import -Winit-self -Winline @gol
222 -Wno-invalid-offsetof -Winvalid-pch @gol
223 -Wlarger-than-@var{len} -Wlong-long @gol
224 -Wmain -Wmissing-braces @gol
225 -Wmissing-format-attribute -Wmissing-include-dirs @gol
226 -Wmissing-noreturn @gol
227 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
228 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
229 -Wreturn-type -Wsequence-point -Wshadow @gol
230 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
231 -Wswitch -Wswitch-default -Wswitch-enum @gol
232 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
233 -Wunknown-pragmas -Wunreachable-code @gol
234 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
235 -Wunused-value -Wunused-variable -Wwrite-strings @gol
238 @item C-only Warning Options
239 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
240 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
241 -Wstrict-prototypes -Wtraditional @gol
242 -Wdeclaration-after-statement}
244 @item Debugging Options
245 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
246 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
247 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
248 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
250 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
251 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
252 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
253 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
255 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
256 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
257 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
258 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
259 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
266 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
267 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
268 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
269 -ftest-coverage -ftime-report -fvar-tracking @gol
270 -g -g@var{level} -gcoff -gdwarf-2 @gol
271 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
272 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
273 -print-multi-directory -print-multi-lib @gol
274 -print-prog-name=@var{program} -print-search-dirs -Q @gol
277 @item Optimization Options
278 @xref{Optimize Options,,Options that Control Optimization}.
279 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
280 -falign-labels=@var{n} -falign-loops=@var{n} @gol
281 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
282 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
283 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
284 -fcaller-saves -fcprop-registers @gol
285 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
286 -fdelayed-branch -fdelete-null-pointer-checks @gol
287 -fexpensive-optimizations -ffast-math -ffloat-store @gol
288 -fforce-addr -fforce-mem -ffunction-sections @gol
289 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
290 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
291 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
292 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
293 -fmodulo-sched -fmove-all-movables -fnew-ra -fno-branch-count-reg @gol
294 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
295 -fno-function-cse -fno-guess-branch-probability @gol
296 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
297 -funsafe-math-optimizations -ffinite-math-only @gol
298 -fno-trapping-math -fno-zero-initialized-in-bss @gol
299 -fomit-frame-pointer -foptimize-register-move @gol
300 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
301 -fprofile-generate -fprofile-use @gol
302 -freduce-all-givs -fregmove -frename-registers @gol
303 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
304 -frerun-cse-after-loop -frerun-loop-opt @gol
305 -frounding-math -fschedule-insns -fschedule-insns2 @gol
306 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
307 -fsched-spec-load-dangerous @gol
308 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
309 -fsched2-use-superblocks @gol
310 -fsched2-use-traces -fsignaling-nans @gol
311 -fsingle-precision-constant @gol
312 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
313 -funroll-all-loops -funroll-loops -fpeel-loops @gol
314 -funswitch-loops -fold-unroll-loops -fold-unroll-all-loops @gol
315 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
317 -ftree-dominator-opts -ftree-dse -ftree-copyrename @gol
318 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre @gol
319 --param @var{name}=@var{value}
320 -O -O0 -O1 -O2 -O3 -Os}
322 @item Preprocessor Options
323 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
324 @gccoptlist{-A@var{question}=@var{answer} @gol
325 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
326 -C -dD -dI -dM -dN @gol
327 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
328 -idirafter @var{dir} @gol
329 -include @var{file} -imacros @var{file} @gol
330 -iprefix @var{file} -iwithprefix @var{dir} @gol
331 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
332 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
333 -P -fworking-directory -remap @gol
334 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
335 -Xpreprocessor @var{option}}
337 @item Assembler Option
338 @xref{Assembler Options,,Passing Options to the Assembler}.
339 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
342 @xref{Link Options,,Options for Linking}.
343 @gccoptlist{@var{object-file-name} -l@var{library} @gol
344 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
345 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
346 -Wl,@var{option} -Xlinker @var{option} @gol
349 @item Directory Options
350 @xref{Directory Options,,Options for Directory Search}.
351 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
354 @c I wrote this xref this way to avoid overfull hbox. -- rms
355 @xref{Target Options}.
356 @gccoptlist{-V @var{version} -b @var{machine}}
358 @item Machine Dependent Options
359 @xref{Submodel Options,,Hardware Models and Configurations}.
360 @c This list is ordered alphanumerically by subsection name.
361 @c Try and put the significant identifier (CPU or system) first,
362 @c so users have a clue at guessing where the ones they want will be.
365 @gccoptlist{-EB -EL @gol
366 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
367 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
370 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
371 -mabi=@var{name} @gol
372 -mapcs-stack-check -mno-apcs-stack-check @gol
373 -mapcs-float -mno-apcs-float @gol
374 -mapcs-reentrant -mno-apcs-reentrant @gol
375 -msched-prolog -mno-sched-prolog @gol
376 -mlittle-endian -mbig-endian -mwords-little-endian @gol
377 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
378 -mthumb-interwork -mno-thumb-interwork @gol
379 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
380 -mstructure-size-boundary=@var{n} @gol
381 -mabort-on-noreturn @gol
382 -mlong-calls -mno-long-calls @gol
383 -msingle-pic-base -mno-single-pic-base @gol
384 -mpic-register=@var{reg} @gol
385 -mnop-fun-dllimport @gol
386 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
387 -mpoke-function-name @gol
389 -mtpcs-frame -mtpcs-leaf-frame @gol
390 -mcaller-super-interworking -mcallee-super-interworking}
393 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
394 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
397 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
398 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
399 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
400 -mstack-align -mdata-align -mconst-align @gol
401 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
402 -melf -maout -melinux -mlinux -sim -sim2 @gol
403 -mmul-bug-workaround -mno-mul-bug-workaround}
405 @emph{Darwin Options}
406 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
407 -arch_only -bind_at_load -bundle -bundle_loader @gol
408 -client_name -compatibility_version -current_version @gol
409 -dependency-file -dylib_file -dylinker_install_name @gol
410 -dynamic -dynamiclib -exported_symbols_list @gol
411 -filelist -flat_namespace -force_cpusubtype_ALL @gol
412 -force_flat_namespace -headerpad_max_install_names @gol
413 -image_base -init -install_name -keep_private_externs @gol
414 -multi_module -multiply_defined -multiply_defined_unused @gol
415 -noall_load -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
416 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
417 -private_bundle -read_only_relocs -sectalign @gol
418 -sectobjectsymbols -whyload -seg1addr @gol
419 -sectcreate -sectobjectsymbols -sectorder @gol
420 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
421 -segprot -segs_read_only_addr -segs_read_write_addr @gol
422 -single_module -static -sub_library -sub_umbrella @gol
423 -twolevel_namespace -umbrella -undefined @gol
424 -unexported_symbols_list -weak_reference_mismatches @gol
425 -whatsloaded -F -gused -gfull}
427 @emph{DEC Alpha Options}
428 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
429 -mieee -mieee-with-inexact -mieee-conformant @gol
430 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
431 -mtrap-precision=@var{mode} -mbuild-constants @gol
432 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
433 -mbwx -mmax -mfix -mcix @gol
434 -mfloat-vax -mfloat-ieee @gol
435 -mexplicit-relocs -msmall-data -mlarge-data @gol
436 -msmall-text -mlarge-text @gol
437 -mmemory-latency=@var{time}}
439 @emph{DEC Alpha/VMS Options}
440 @gccoptlist{-mvms-return-codes}
443 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
444 -mhard-float -msoft-float @gol
445 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
446 -mdouble -mno-double @gol
447 -mmedia -mno-media -mmuladd -mno-muladd @gol
448 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic -mlinked-fp @gol
449 -mlibrary-pic -macc-4 -macc-8 @gol
450 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
451 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
452 -mvliw-branch -mno-vliw-branch @gol
453 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
454 -mno-nested-cond-exec -mtomcat-stats @gol
457 @emph{H8/300 Options}
458 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
461 @gccoptlist{-march=@var{architecture-type} @gol
462 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
463 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
464 -mfixed-range=@var{register-range} @gol
465 -mjump-in-delay -mlinker-opt -mlong-calls @gol
466 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
467 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
468 -mno-jump-in-delay -mno-long-load-store @gol
469 -mno-portable-runtime -mno-soft-float @gol
470 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
471 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
472 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
473 -nolibdld -static -threads}
475 @emph{i386 and x86-64 Options}
476 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
477 -mfpmath=@var{unit} @gol
478 -masm=@var{dialect} -mno-fancy-math-387 @gol
479 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
480 -mno-wide-multiply -mrtd -malign-double @gol
481 -mpreferred-stack-boundary=@var{num} @gol
482 -mmmx -msse -msse2 -msse3 -m3dnow @gol
483 -mthreads -mno-align-stringops -minline-all-stringops @gol
484 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
485 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
486 -mno-red-zone -mno-tls-direct-seg-refs @gol
487 -mcmodel=@var{code-model} @gol
491 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
492 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
493 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
494 -minline-float-divide-max-throughput @gol
495 -minline-int-divide-min-latency @gol
496 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
497 -mfixed-range=@var{register-range}}
499 @emph{M32R/D Options}
500 @gccoptlist{-m32r2 -m32rx -m32r @gol
502 -malign-loops -mno-align-loops @gol
503 -missue-rate=@var{number} @gol
504 -mbranch-cost=@var{number} @gol
505 -mmodel=@var{code-size-model-type} @gol
506 -msdata=@var{sdata-type} @gol
507 -mno-flush-func -mflush-func=@var{name} @gol
508 -mno-flush-trap -mflush-trap=@var{number} @gol
511 @emph{M680x0 Options}
512 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
513 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
514 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
515 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
516 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
518 @emph{M68hc1x Options}
519 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
520 -mauto-incdec -minmax -mlong-calls -mshort @gol
521 -msoft-reg-count=@var{count}}
524 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
525 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
526 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
527 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
528 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
531 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
532 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
533 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
534 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
535 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
536 -mint64 -mlong64 -mlong32 @gol
537 -G@var{num} -membedded-data -mno-embedded-data @gol
538 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
539 -msplit-addresses -mno-split-addresses @gol
540 -mexplicit-relocs -mno-explicit-relocs @gol
541 -mcheck-zero-division -mno-check-zero-division @gol
542 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
543 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
544 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
545 -mfix-vr4120 -mno-fix-vr4120 -mfix-sb1 -mno-fix-sb1 @gol
546 -mflush-func=@var{func} -mno-flush-func @gol
547 -mbranch-likely -mno-branch-likely @gol
548 -mfp-exceptions -mno-fp-exceptions @gol
549 -mvr4130-align -mno-vr4130-align}
552 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
553 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
554 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
555 -mno-base-addresses -msingle-exit -mno-single-exit}
557 @emph{MN10300 Options}
558 @gccoptlist{-mmult-bug -mno-mult-bug @gol
559 -mam33 -mno-am33 @gol
560 -mam33-2 -mno-am33-2 @gol
564 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
565 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
566 -mregparam -mnoregparam -msb -mnosb @gol
567 -mbitfield -mnobitfield -mhimem -mnohimem}
569 @emph{PDP-11 Options}
570 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
571 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
572 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
573 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
574 -mbranch-expensive -mbranch-cheap @gol
575 -msplit -mno-split -munix-asm -mdec-asm}
577 @emph{PowerPC Options}
578 See RS/6000 and PowerPC Options.
580 @emph{RS/6000 and PowerPC Options}
581 @gccoptlist{-mcpu=@var{cpu-type} @gol
582 -mtune=@var{cpu-type} @gol
583 -mpower -mno-power -mpower2 -mno-power2 @gol
584 -mpowerpc -mpowerpc64 -mno-powerpc @gol
585 -maltivec -mno-altivec @gol
586 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
587 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
588 -mnew-mnemonics -mold-mnemonics @gol
589 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
590 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
591 -malign-power -malign-natural @gol
592 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
593 -mstring -mno-string -mupdate -mno-update @gol
594 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
595 -mstrict-align -mno-strict-align -mrelocatable @gol
596 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
597 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
598 -mdynamic-no-pic @gol
599 -mprioritize-restricted-insns=@var{priority} @gol
600 -msched-costly-dep=@var{dependence_type} @gol
601 -minsert-sched-nops=@var{scheme} @gol
602 -mcall-sysv -mcall-netbsd @gol
603 -maix-struct-return -msvr4-struct-return @gol
604 -mabi=altivec -mabi=no-altivec @gol
605 -mabi=spe -mabi=no-spe @gol
606 -misel=yes -misel=no @gol
607 -mspe=yes -mspe=no @gol
608 -mfloat-gprs=yes -mfloat-gprs=no @gol
609 -mprototype -mno-prototype @gol
610 -msim -mmvme -mads -myellowknife -memb -msdata @gol
611 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
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 @gol
618 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd}
621 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
622 -m4-nofpu -m4-single-only -m4-single -m4 @gol
623 -m5-64media -m5-64media-nofpu @gol
624 -m5-32media -m5-32media-nofpu @gol
625 -m5-compact -m5-compact-nofpu @gol
626 -mb -ml -mdalign -mrelax @gol
627 -mbigtable -mfmovd -mhitachi -mnomacsave @gol
628 -mieee -misize -mpadstruct -mspace @gol
629 -mprefergot -musermode}
632 @gccoptlist{-mcpu=@var{cpu-type} @gol
633 -mtune=@var{cpu-type} @gol
634 -mcmodel=@var{code-model} @gol
635 -m32 -m64 -mapp-regs -mno-app-regs @gol
636 -mfaster-structs -mno-faster-structs @gol
637 -mfpu -mno-fpu -mhard-float -msoft-float @gol
638 -mhard-quad-float -msoft-quad-float @gol
639 -mimpure-text -mno-impure-text -mlittle-endian @gol
640 -mstack-bias -mno-stack-bias @gol
641 -munaligned-doubles -mno-unaligned-doubles @gol
642 -mv8plus -mno-v8plus -mvis -mno-vis
645 @emph{System V Options}
646 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
648 @emph{TMS320C3x/C4x Options}
649 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
650 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
651 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
652 -mparallel-insns -mparallel-mpy -mpreserve-float}
655 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
656 -mprolog-function -mno-prolog-function -mspace @gol
657 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
658 -mapp-regs -mno-app-regs @gol
659 -mdisable-callt -mno-disable-callt @gol
665 @gccoptlist{-mg -mgnu -munix}
667 @emph{x86-64 Options}
668 See i386 and x86-64 Options.
670 @emph{Xstormy16 Options}
673 @emph{Xtensa Options}
674 @gccoptlist{-mconst16 -mno-const16 @gol
675 -mfused-madd -mno-fused-madd @gol
676 -mtext-section-literals -mno-text-section-literals @gol
677 -mtarget-align -mno-target-align @gol
678 -mlongcalls -mno-longcalls}
680 @emph{zSeries Options}
681 See S/390 and zSeries Options.
683 @item Code Generation Options
684 @xref{Code Gen Options,,Options for Code Generation Conventions}.
685 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
686 -ffixed-@var{reg} -fexceptions @gol
687 -fnon-call-exceptions -funwind-tables @gol
688 -fasynchronous-unwind-tables @gol
689 -finhibit-size-directive -finstrument-functions @gol
690 -fno-common -fno-ident @gol
691 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
692 -freg-struct-return -fshared-data -fshort-enums @gol
693 -fshort-double -fshort-wchar @gol
694 -fverbose-asm -fpack-struct -fstack-check @gol
695 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
696 -fargument-alias -fargument-noalias @gol
697 -fargument-noalias-global -fleading-underscore @gol
698 -ftls-model=@var{model} @gol
699 -ftrapv -fwrapv -fbounds-check}
703 * Overall Options:: Controlling the kind of output:
704 an executable, object files, assembler files,
705 or preprocessed source.
706 * C Dialect Options:: Controlling the variant of C language compiled.
707 * C++ Dialect Options:: Variations on C++.
708 * Objective-C Dialect Options:: Variations on Objective-C.
709 * Language Independent Options:: Controlling how diagnostics should be
711 * Warning Options:: How picky should the compiler be?
712 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
713 * Optimize Options:: How much optimization?
714 * Preprocessor Options:: Controlling header files and macro definitions.
715 Also, getting dependency information for Make.
716 * Assembler Options:: Passing options to the assembler.
717 * Link Options:: Specifying libraries and so on.
718 * Directory Options:: Where to find header files and libraries.
719 Where to find the compiler executable files.
720 * Spec Files:: How to pass switches to sub-processes.
721 * Target Options:: Running a cross-compiler, or an old version of GCC.
724 @node Overall Options
725 @section Options Controlling the Kind of Output
727 Compilation can involve up to four stages: preprocessing, compilation
728 proper, assembly and linking, always in that order. GCC is capable of
729 preprocessing and compiling several files either into several
730 assembler input files, or into one assembler input file; then each
731 assembler input file produces an object file, and linking combines all
732 the object files (those newly compiled, and those specified as input)
733 into an executable file.
735 @cindex file name suffix
736 For any given input file, the file name suffix determines what kind of
741 C source code which must be preprocessed.
744 C source code which should not be preprocessed.
747 C++ source code which should not be preprocessed.
750 Objective-C source code. Note that you must link with the library
751 @file{libobjc.a} to make an Objective-C program work.
754 Objective-C source code which should not be preprocessed.
757 C or C++ header file to be turned into a precompiled header.
761 @itemx @var{file}.cxx
762 @itemx @var{file}.cpp
763 @itemx @var{file}.CPP
764 @itemx @var{file}.c++
766 C++ source code which must be preprocessed. Note that in @samp{.cxx},
767 the last two letters must both be literally @samp{x}. Likewise,
768 @samp{.C} refers to a literal capital C@.
772 C++ header file to be turned into a precompiled header.
775 @itemx @var{file}.for
776 @itemx @var{file}.FOR
777 Fortran source code which should not be preprocessed.
780 @itemx @var{file}.fpp
781 @itemx @var{file}.FPP
782 Fortran source code which must be preprocessed (with the traditional
786 Fortran source code which must be preprocessed with a RATFOR
787 preprocessor (not included with GCC)@.
790 @itemx @var{file}.f95
791 Fortran 90/95 source code which should not be preprocessed.
793 @c FIXME: Descriptions of Java file types.
800 Ada source code file which contains a library unit declaration (a
801 declaration of a package, subprogram, or generic, or a generic
802 instantiation), or a library unit renaming declaration (a package,
803 generic, or subprogram renaming declaration). Such files are also
806 @itemx @var{file}.adb
807 Ada source code file containing a library unit body (a subprogram or
808 package body). Such files are also called @dfn{bodies}.
810 @c GCC also knows about some suffixes for languages not yet included:
819 Assembler code which must be preprocessed.
822 An object file to be fed straight into linking.
823 Any file name with no recognized suffix is treated this way.
827 You can specify the input language explicitly with the @option{-x} option:
830 @item -x @var{language}
831 Specify explicitly the @var{language} for the following input files
832 (rather than letting the compiler choose a default based on the file
833 name suffix). This option applies to all following input files until
834 the next @option{-x} option. Possible values for @var{language} are:
836 c c-header cpp-output
837 c++ c++-header c++-cpp-output
838 objective-c objective-c-header objc-cpp-output
839 assembler assembler-with-cpp
841 f77 f77-cpp-input ratfor
848 Turn off any specification of a language, so that subsequent files are
849 handled according to their file name suffixes (as they are if @option{-x}
850 has not been used at all).
852 @item -pass-exit-codes
853 @opindex pass-exit-codes
854 Normally the @command{gcc} program will exit with the code of 1 if any
855 phase of the compiler returns a non-success return code. If you specify
856 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
857 numerically highest error produced by any phase that returned an error
861 If you only want some of the stages of compilation, you can use
862 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
863 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
864 @command{gcc} is to stop. Note that some combinations (for example,
865 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
870 Compile or assemble the source files, but do not link. The linking
871 stage simply is not done. The ultimate output is in the form of an
872 object file for each source file.
874 By default, the object file name for a source file is made by replacing
875 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
877 Unrecognized input files, not requiring compilation or assembly, are
882 Stop after the stage of compilation proper; do not assemble. The output
883 is in the form of an assembler code file for each non-assembler input
886 By default, the assembler file name for a source file is made by
887 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
889 Input files that don't require compilation are ignored.
893 Stop after the preprocessing stage; do not run the compiler proper. The
894 output is in the form of preprocessed source code, which is sent to the
897 Input files which don't require preprocessing are ignored.
899 @cindex output file option
902 Place output in file @var{file}. This applies regardless to whatever
903 sort of output is being produced, whether it be an executable file,
904 an object file, an assembler file or preprocessed C code.
906 If @option{-o} is not specified, the default is to put an executable
907 file in @file{a.out}, the object file for
908 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
909 assembler file in @file{@var{source}.s}, a precompiled header file in
910 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
915 Print (on standard error output) the commands executed to run the stages
916 of compilation. Also print the version number of the compiler driver
917 program and of the preprocessor and the compiler proper.
921 Like @option{-v} except the commands are not executed and all command
922 arguments are quoted. This is useful for shell scripts to capture the
923 driver-generated command lines.
927 Use pipes rather than temporary files for communication between the
928 various stages of compilation. This fails to work on some systems where
929 the assembler is unable to read from a pipe; but the GNU assembler has
934 If you are compiling multiple source files, this option tells the driver
935 to pass all the source files to the compiler at once (for those
936 languages for which the compiler can handle this). This will allow
937 intermodule analysis (IMA) to be performed by the compiler. Currently the only
938 language for which this is supported is C. If you pass source files for
939 multiple languages to the driver, using this option, the driver will invoke
940 the compiler(s) that support IMA once each, passing each compiler all the
941 source files appropriate for it. For those languages that do not support
942 IMA this option will be ignored, and the compiler will be invoked once for
943 each source file in that language. If you use this option in conjunction
944 with -save-temps, the compiler will generate multiple pre-processed files
945 (one for each source file), but only one (combined) .o or .s file.
949 Print (on the standard output) a description of the command line options
950 understood by @command{gcc}. If the @option{-v} option is also specified
951 then @option{--help} will also be passed on to the various processes
952 invoked by @command{gcc}, so that they can display the command line options
953 they accept. If the @option{-Wextra} option is also specified then command
954 line options which have no documentation associated with them will also
959 Print (on the standard output) a description of target specific command
960 line options for each tool.
964 Display the version number and copyrights of the invoked GCC.
968 @section Compiling C++ Programs
970 @cindex suffixes for C++ source
971 @cindex C++ source file suffixes
972 C++ source files conventionally use one of the suffixes @samp{.C},
973 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
974 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
975 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
976 files with these names and compiles them as C++ programs even if you
977 call the compiler the same way as for compiling C programs (usually
978 with the name @command{gcc}).
982 However, C++ programs often require class libraries as well as a
983 compiler that understands the C++ language---and under some
984 circumstances, you might want to compile programs or header files from
985 standard input, or otherwise without a suffix that flags them as C++
986 programs. You might also like to precompile a C header file with a
987 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
988 program that calls GCC with the default language set to C++, and
989 automatically specifies linking against the C++ library. On many
990 systems, @command{g++} is also installed with the name @command{c++}.
992 @cindex invoking @command{g++}
993 When you compile C++ programs, you may specify many of the same
994 command-line options that you use for compiling programs in any
995 language; or command-line options meaningful for C and related
996 languages; or options that are meaningful only for C++ programs.
997 @xref{C Dialect Options,,Options Controlling C Dialect}, for
998 explanations of options for languages related to C@.
999 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1000 explanations of options that are meaningful only for C++ programs.
1002 @node C Dialect Options
1003 @section Options Controlling C Dialect
1004 @cindex dialect options
1005 @cindex language dialect options
1006 @cindex options, dialect
1008 The following options control the dialect of C (or languages derived
1009 from C, such as C++ and Objective-C) that the compiler accepts:
1012 @cindex ANSI support
1016 In C mode, support all ISO C90 programs. In C++ mode,
1017 remove GNU extensions that conflict with ISO C++.
1019 This turns off certain features of GCC that are incompatible with ISO
1020 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1021 such as the @code{asm} and @code{typeof} keywords, and
1022 predefined macros such as @code{unix} and @code{vax} that identify the
1023 type of system you are using. It also enables the undesirable and
1024 rarely used ISO trigraph feature. For the C compiler,
1025 it disables recognition of C++ style @samp{//} comments as well as
1026 the @code{inline} keyword.
1028 The alternate keywords @code{__asm__}, @code{__extension__},
1029 @code{__inline__} and @code{__typeof__} continue to work despite
1030 @option{-ansi}. You would not want to use them in an ISO C program, of
1031 course, but it is useful to put them in header files that might be included
1032 in compilations done with @option{-ansi}. Alternate predefined macros
1033 such as @code{__unix__} and @code{__vax__} are also available, with or
1034 without @option{-ansi}.
1036 The @option{-ansi} option does not cause non-ISO programs to be
1037 rejected gratuitously. For that, @option{-pedantic} is required in
1038 addition to @option{-ansi}. @xref{Warning Options}.
1040 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1041 option is used. Some header files may notice this macro and refrain
1042 from declaring certain functions or defining certain macros that the
1043 ISO standard doesn't call for; this is to avoid interfering with any
1044 programs that might use these names for other things.
1046 Functions which would normally be built in but do not have semantics
1047 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1048 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1049 built-in functions provided by GCC}, for details of the functions
1054 Determine the language standard. This option is currently only
1055 supported when compiling C or C++. A value for this option must be
1056 provided; possible values are
1061 ISO C90 (same as @option{-ansi}).
1063 @item iso9899:199409
1064 ISO C90 as modified in amendment 1.
1070 ISO C99. Note that this standard is not yet fully supported; see
1071 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1072 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1075 Default, ISO C90 plus GNU extensions (including some C99 features).
1079 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1080 this will become the default. The name @samp{gnu9x} is deprecated.
1083 The 1998 ISO C++ standard plus amendments.
1086 The same as @option{-std=c++98} plus GNU extensions. This is the
1087 default for C++ code.
1090 Even when this option is not specified, you can still use some of the
1091 features of newer standards in so far as they do not conflict with
1092 previous C standards. For example, you may use @code{__restrict__} even
1093 when @option{-std=c99} is not specified.
1095 The @option{-std} options specifying some version of ISO C have the same
1096 effects as @option{-ansi}, except that features that were not in ISO C90
1097 but are in the specified version (for example, @samp{//} comments and
1098 the @code{inline} keyword in ISO C99) are not disabled.
1100 @xref{Standards,,Language Standards Supported by GCC}, for details of
1101 these standard versions.
1103 @item -aux-info @var{filename}
1105 Output to the given filename prototyped declarations for all functions
1106 declared and/or defined in a translation unit, including those in header
1107 files. This option is silently ignored in any language other than C@.
1109 Besides declarations, the file indicates, in comments, the origin of
1110 each declaration (source file and line), whether the declaration was
1111 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1112 @samp{O} for old, respectively, in the first character after the line
1113 number and the colon), and whether it came from a declaration or a
1114 definition (@samp{C} or @samp{F}, respectively, in the following
1115 character). In the case of function definitions, a K&R-style list of
1116 arguments followed by their declarations is also provided, inside
1117 comments, after the declaration.
1121 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1122 keyword, so that code can use these words as identifiers. You can use
1123 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1124 instead. @option{-ansi} implies @option{-fno-asm}.
1126 In C++, this switch only affects the @code{typeof} keyword, since
1127 @code{asm} and @code{inline} are standard keywords. You may want to
1128 use the @option{-fno-gnu-keywords} flag instead, which has the same
1129 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1130 switch only affects the @code{asm} and @code{typeof} keywords, since
1131 @code{inline} is a standard keyword in ISO C99.
1134 @itemx -fno-builtin-@var{function}
1135 @opindex fno-builtin
1136 @cindex built-in functions
1137 Don't recognize built-in functions that do not begin with
1138 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1139 functions provided by GCC}, for details of the functions affected,
1140 including those which are not built-in functions when @option{-ansi} or
1141 @option{-std} options for strict ISO C conformance are used because they
1142 do not have an ISO standard meaning.
1144 GCC normally generates special code to handle certain built-in functions
1145 more efficiently; for instance, calls to @code{alloca} may become single
1146 instructions that adjust the stack directly, and calls to @code{memcpy}
1147 may become inline copy loops. The resulting code is often both smaller
1148 and faster, but since the function calls no longer appear as such, you
1149 cannot set a breakpoint on those calls, nor can you change the behavior
1150 of the functions by linking with a different library.
1152 With the @option{-fno-builtin-@var{function}} option
1153 only the built-in function @var{function} is
1154 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1155 function is named this is not built-in in this version of GCC, this
1156 option is ignored. There is no corresponding
1157 @option{-fbuiltin-@var{function}} option; if you wish to enable
1158 built-in functions selectively when using @option{-fno-builtin} or
1159 @option{-ffreestanding}, you may define macros such as:
1162 #define abs(n) __builtin_abs ((n))
1163 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1168 @cindex hosted environment
1170 Assert that compilation takes place in a hosted environment. This implies
1171 @option{-fbuiltin}. A hosted environment is one in which the
1172 entire standard library is available, and in which @code{main} has a return
1173 type of @code{int}. Examples are nearly everything except a kernel.
1174 This is equivalent to @option{-fno-freestanding}.
1176 @item -ffreestanding
1177 @opindex ffreestanding
1178 @cindex hosted environment
1180 Assert that compilation takes place in a freestanding environment. This
1181 implies @option{-fno-builtin}. A freestanding environment
1182 is one in which the standard library may not exist, and program startup may
1183 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1184 This is equivalent to @option{-fno-hosted}.
1186 @xref{Standards,,Language Standards Supported by GCC}, for details of
1187 freestanding and hosted environments.
1189 @item -fms-extensions
1190 @opindex fms-extensions
1191 Accept some non-standard constructs used in Microsoft header files.
1195 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1196 options for strict ISO C conformance) implies @option{-trigraphs}.
1198 @item -no-integrated-cpp
1199 @opindex no-integrated-cpp
1200 Performs a compilation in two passes: preprocessing and compiling. This
1201 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1202 @option{-B} option. The user supplied compilation step can then add in
1203 an additional preprocessing step after normal preprocessing but before
1204 compiling. The default is to use the integrated cpp (internal cpp)
1206 The semantics of this option will change if "cc1", "cc1plus", and
1207 "cc1obj" are merged.
1209 @cindex traditional C language
1210 @cindex C language, traditional
1212 @itemx -traditional-cpp
1213 @opindex traditional-cpp
1214 @opindex traditional
1215 Formerly, these options caused GCC to attempt to emulate a pre-standard
1216 C compiler. They are now only supported with the @option{-E} switch.
1217 The preprocessor continues to support a pre-standard mode. See the GNU
1218 CPP manual for details.
1220 @item -fcond-mismatch
1221 @opindex fcond-mismatch
1222 Allow conditional expressions with mismatched types in the second and
1223 third arguments. The value of such an expression is void. This option
1224 is not supported for C++.
1226 @item -funsigned-char
1227 @opindex funsigned-char
1228 Let the type @code{char} be unsigned, like @code{unsigned char}.
1230 Each kind of machine has a default for what @code{char} should
1231 be. It is either like @code{unsigned char} by default or like
1232 @code{signed char} by default.
1234 Ideally, a portable program should always use @code{signed char} or
1235 @code{unsigned char} when it depends on the signedness of an object.
1236 But many programs have been written to use plain @code{char} and
1237 expect it to be signed, or expect it to be unsigned, depending on the
1238 machines they were written for. This option, and its inverse, let you
1239 make such a program work with the opposite default.
1241 The type @code{char} is always a distinct type from each of
1242 @code{signed char} or @code{unsigned char}, even though its behavior
1243 is always just like one of those two.
1246 @opindex fsigned-char
1247 Let the type @code{char} be signed, like @code{signed char}.
1249 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1250 the negative form of @option{-funsigned-char}. Likewise, the option
1251 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1253 @item -fsigned-bitfields
1254 @itemx -funsigned-bitfields
1255 @itemx -fno-signed-bitfields
1256 @itemx -fno-unsigned-bitfields
1257 @opindex fsigned-bitfields
1258 @opindex funsigned-bitfields
1259 @opindex fno-signed-bitfields
1260 @opindex fno-unsigned-bitfields
1261 These options control whether a bit-field is signed or unsigned, when the
1262 declaration does not use either @code{signed} or @code{unsigned}. By
1263 default, such a bit-field is signed, because this is consistent: the
1264 basic integer types such as @code{int} are signed types.
1267 @node C++ Dialect Options
1268 @section Options Controlling C++ Dialect
1270 @cindex compiler options, C++
1271 @cindex C++ options, command line
1272 @cindex options, C++
1273 This section describes the command-line options that are only meaningful
1274 for C++ programs; but you can also use most of the GNU compiler options
1275 regardless of what language your program is in. For example, you
1276 might compile a file @code{firstClass.C} like this:
1279 g++ -g -frepo -O -c firstClass.C
1283 In this example, only @option{-frepo} is an option meant
1284 only for C++ programs; you can use the other options with any
1285 language supported by GCC@.
1287 Here is a list of options that are @emph{only} for compiling C++ programs:
1291 @item -fabi-version=@var{n}
1292 @opindex fabi-version
1293 Use version @var{n} of the C++ ABI. Version 2 is the version of the
1294 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1295 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1296 the version that conforms most closely to the C++ ABI specification.
1297 Therefore, the ABI obtained using version 0 will change as ABI bugs
1300 The default is version 2.
1302 @item -fno-access-control
1303 @opindex fno-access-control
1304 Turn off all access checking. This switch is mainly useful for working
1305 around bugs in the access control code.
1309 Check that the pointer returned by @code{operator new} is non-null
1310 before attempting to modify the storage allocated. This check is
1311 normally unnecessary because the C++ standard specifies that
1312 @code{operator new} will only return @code{0} if it is declared
1313 @samp{throw()}, in which case the compiler will always check the
1314 return value even without this option. In all other cases, when
1315 @code{operator new} has a non-empty exception specification, memory
1316 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1317 @samp{new (nothrow)}.
1319 @item -fconserve-space
1320 @opindex fconserve-space
1321 Put uninitialized or runtime-initialized global variables into the
1322 common segment, as C does. This saves space in the executable at the
1323 cost of not diagnosing duplicate definitions. If you compile with this
1324 flag and your program mysteriously crashes after @code{main()} has
1325 completed, you may have an object that is being destroyed twice because
1326 two definitions were merged.
1328 This option is no longer useful on most targets, now that support has
1329 been added for putting variables into BSS without making them common.
1331 @item -fno-const-strings
1332 @opindex fno-const-strings
1333 Give string constants type @code{char *} instead of type @code{const
1334 char *}. By default, G++ uses type @code{const char *} as required by
1335 the standard. Even if you use @option{-fno-const-strings}, you cannot
1336 actually modify the value of a string constant.
1338 This option might be removed in a future release of G++. For maximum
1339 portability, you should structure your code so that it works with
1340 string constants that have type @code{const char *}.
1342 @item -fno-elide-constructors
1343 @opindex fno-elide-constructors
1344 The C++ standard allows an implementation to omit creating a temporary
1345 which is only used to initialize another object of the same type.
1346 Specifying this option disables that optimization, and forces G++ to
1347 call the copy constructor in all cases.
1349 @item -fno-enforce-eh-specs
1350 @opindex fno-enforce-eh-specs
1351 Don't check for violation of exception specifications at runtime. This
1352 option violates the C++ standard, but may be useful for reducing code
1353 size in production builds, much like defining @samp{NDEBUG}. The compiler
1354 will still optimize based on the exception specifications.
1357 @itemx -fno-for-scope
1359 @opindex fno-for-scope
1360 If @option{-ffor-scope} is specified, the scope of variables declared in
1361 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1362 as specified by the C++ standard.
1363 If @option{-fno-for-scope} is specified, the scope of variables declared in
1364 a @i{for-init-statement} extends to the end of the enclosing scope,
1365 as was the case in old versions of G++, and other (traditional)
1366 implementations of C++.
1368 The default if neither flag is given to follow the standard,
1369 but to allow and give a warning for old-style code that would
1370 otherwise be invalid, or have different behavior.
1372 @item -fno-gnu-keywords
1373 @opindex fno-gnu-keywords
1374 Do not recognize @code{typeof} as a keyword, so that code can use this
1375 word as an identifier. You can use the keyword @code{__typeof__} instead.
1376 @option{-ansi} implies @option{-fno-gnu-keywords}.
1378 @item -fno-implicit-templates
1379 @opindex fno-implicit-templates
1380 Never emit code for non-inline templates which are instantiated
1381 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1382 @xref{Template Instantiation}, for more information.
1384 @item -fno-implicit-inline-templates
1385 @opindex fno-implicit-inline-templates
1386 Don't emit code for implicit instantiations of inline templates, either.
1387 The default is to handle inlines differently so that compiles with and
1388 without optimization will need the same set of explicit instantiations.
1390 @item -fno-implement-inlines
1391 @opindex fno-implement-inlines
1392 To save space, do not emit out-of-line copies of inline functions
1393 controlled by @samp{#pragma implementation}. This will cause linker
1394 errors if these functions are not inlined everywhere they are called.
1396 @item -fms-extensions
1397 @opindex fms-extensions
1398 Disable pedantic warnings about constructs used in MFC, such as implicit
1399 int and getting a pointer to member function via non-standard syntax.
1401 @item -fno-nonansi-builtins
1402 @opindex fno-nonansi-builtins
1403 Disable built-in declarations of functions that are not mandated by
1404 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1405 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1407 @item -fno-operator-names
1408 @opindex fno-operator-names
1409 Do not treat the operator name keywords @code{and}, @code{bitand},
1410 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1411 synonyms as keywords.
1413 @item -fno-optional-diags
1414 @opindex fno-optional-diags
1415 Disable diagnostics that the standard says a compiler does not need to
1416 issue. Currently, the only such diagnostic issued by G++ is the one for
1417 a name having multiple meanings within a class.
1420 @opindex fpermissive
1421 Downgrade some diagnostics about nonconformant code from errors to
1422 warnings. Thus, using @option{-fpermissive} will allow some
1423 nonconforming code to compile.
1427 Enable automatic template instantiation at link time. This option also
1428 implies @option{-fno-implicit-templates}. @xref{Template
1429 Instantiation}, for more information.
1433 Disable generation of information about every class with virtual
1434 functions for use by the C++ runtime type identification features
1435 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1436 of the language, you can save some space by using this flag. Note that
1437 exception handling uses the same information, but it will generate it as
1442 Emit statistics about front-end processing at the end of the compilation.
1443 This information is generally only useful to the G++ development team.
1445 @item -ftemplate-depth-@var{n}
1446 @opindex ftemplate-depth
1447 Set the maximum instantiation depth for template classes to @var{n}.
1448 A limit on the template instantiation depth is needed to detect
1449 endless recursions during template class instantiation. ANSI/ISO C++
1450 conforming programs must not rely on a maximum depth greater than 17.
1452 @item -fuse-cxa-atexit
1453 @opindex fuse-cxa-atexit
1454 Register destructors for objects with static storage duration with the
1455 @code{__cxa_atexit} function rather than the @code{atexit} function.
1456 This option is required for fully standards-compliant handling of static
1457 destructors, but will only work if your C library supports
1458 @code{__cxa_atexit}.
1462 Do not use weak symbol support, even if it is provided by the linker.
1463 By default, G++ will use weak symbols if they are available. This
1464 option exists only for testing, and should not be used by end-users;
1465 it will result in inferior code and has no benefits. This option may
1466 be removed in a future release of G++.
1470 Do not search for header files in the standard directories specific to
1471 C++, but do still search the other standard directories. (This option
1472 is used when building the C++ library.)
1475 In addition, these optimization, warning, and code generation options
1476 have meanings only for C++ programs:
1479 @item -fno-default-inline
1480 @opindex fno-default-inline
1481 Do not assume @samp{inline} for functions defined inside a class scope.
1482 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1483 functions will have linkage like inline functions; they just won't be
1486 @item -Wabi @r{(C++ only)}
1488 Warn when G++ generates code that is probably not compatible with the
1489 vendor-neutral C++ ABI. Although an effort has been made to warn about
1490 all such cases, there are probably some cases that are not warned about,
1491 even though G++ is generating incompatible code. There may also be
1492 cases where warnings are emitted even though the code that is generated
1495 You should rewrite your code to avoid these warnings if you are
1496 concerned about the fact that code generated by G++ may not be binary
1497 compatible with code generated by other compilers.
1499 The known incompatibilities at this point include:
1504 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1505 pack data into the same byte as a base class. For example:
1508 struct A @{ virtual void f(); int f1 : 1; @};
1509 struct B : public A @{ int f2 : 1; @};
1513 In this case, G++ will place @code{B::f2} into the same byte
1514 as@code{A::f1}; other compilers will not. You can avoid this problem
1515 by explicitly padding @code{A} so that its size is a multiple of the
1516 byte size on your platform; that will cause G++ and other compilers to
1517 layout @code{B} identically.
1520 Incorrect handling of tail-padding for virtual bases. G++ does not use
1521 tail padding when laying out virtual bases. For example:
1524 struct A @{ virtual void f(); char c1; @};
1525 struct B @{ B(); char c2; @};
1526 struct C : public A, public virtual B @{@};
1530 In this case, G++ will not place @code{B} into the tail-padding for
1531 @code{A}; other compilers will. You can avoid this problem by
1532 explicitly padding @code{A} so that its size is a multiple of its
1533 alignment (ignoring virtual base classes); that will cause G++ and other
1534 compilers to layout @code{C} identically.
1537 Incorrect handling of bit-fields with declared widths greater than that
1538 of their underlying types, when the bit-fields appear in a union. For
1542 union U @{ int i : 4096; @};
1546 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1547 union too small by the number of bits in an @code{int}.
1550 Empty classes can be placed at incorrect offsets. For example:
1560 struct C : public B, public A @{@};
1564 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1565 it should be placed at offset zero. G++ mistakenly believes that the
1566 @code{A} data member of @code{B} is already at offset zero.
1569 Names of template functions whose types involve @code{typename} or
1570 template template parameters can be mangled incorrectly.
1573 template <typename Q>
1574 void f(typename Q::X) @{@}
1576 template <template <typename> class Q>
1577 void f(typename Q<int>::X) @{@}
1581 Instantiations of these templates may be mangled incorrectly.
1585 @item -Wctor-dtor-privacy @r{(C++ only)}
1586 @opindex Wctor-dtor-privacy
1587 Warn when a class seems unusable because all the constructors or
1588 destructors in that class are private, and it has neither friends nor
1589 public static member functions.
1591 @item -Wnon-virtual-dtor @r{(C++ only)}
1592 @opindex Wnon-virtual-dtor
1593 Warn when a class appears to be polymorphic, thereby requiring a virtual
1594 destructor, yet it declares a non-virtual one.
1595 This warning is enabled by @option{-Wall}.
1597 @item -Wreorder @r{(C++ only)}
1599 @cindex reordering, warning
1600 @cindex warning for reordering of member initializers
1601 Warn when the order of member initializers given in the code does not
1602 match the order in which they must be executed. For instance:
1608 A(): j (0), i (1) @{ @}
1612 The compiler will rearrange the member initializers for @samp{i}
1613 and @samp{j} to match the declaration order of the members, emitting
1614 a warning to that effect. This warning is enabled by @option{-Wall}.
1617 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1620 @item -Weffc++ @r{(C++ only)}
1622 Warn about violations of the following style guidelines from Scott Meyers'
1623 @cite{Effective C++} book:
1627 Item 11: Define a copy constructor and an assignment operator for classes
1628 with dynamically allocated memory.
1631 Item 12: Prefer initialization to assignment in constructors.
1634 Item 14: Make destructors virtual in base classes.
1637 Item 15: Have @code{operator=} return a reference to @code{*this}.
1640 Item 23: Don't try to return a reference when you must return an object.
1644 Also warn about violations of the following style guidelines from
1645 Scott Meyers' @cite{More Effective C++} book:
1649 Item 6: Distinguish between prefix and postfix forms of increment and
1650 decrement operators.
1653 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1657 When selecting this option, be aware that the standard library
1658 headers do not obey all of these guidelines; use @samp{grep -v}
1659 to filter out those warnings.
1661 @item -Wno-deprecated @r{(C++ only)}
1662 @opindex Wno-deprecated
1663 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1665 @item -Wno-non-template-friend @r{(C++ only)}
1666 @opindex Wno-non-template-friend
1667 Disable warnings when non-templatized friend functions are declared
1668 within a template. Since the advent of explicit template specification
1669 support in G++, if the name of the friend is an unqualified-id (i.e.,
1670 @samp{friend foo(int)}), the C++ language specification demands that the
1671 friend declare or define an ordinary, nontemplate function. (Section
1672 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1673 could be interpreted as a particular specialization of a templatized
1674 function. Because this non-conforming behavior is no longer the default
1675 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1676 check existing code for potential trouble spots and is on by default.
1677 This new compiler behavior can be turned off with
1678 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1679 but disables the helpful warning.
1681 @item -Wold-style-cast @r{(C++ only)}
1682 @opindex Wold-style-cast
1683 Warn if an old-style (C-style) cast to a non-void type is used within
1684 a C++ program. The new-style casts (@samp{static_cast},
1685 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1686 unintended effects and much easier to search for.
1688 @item -Woverloaded-virtual @r{(C++ only)}
1689 @opindex Woverloaded-virtual
1690 @cindex overloaded virtual fn, warning
1691 @cindex warning for overloaded virtual fn
1692 Warn when a function declaration hides virtual functions from a
1693 base class. For example, in:
1700 struct B: public A @{
1705 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1713 will fail to compile.
1715 @item -Wno-pmf-conversions @r{(C++ only)}
1716 @opindex Wno-pmf-conversions
1717 Disable the diagnostic for converting a bound pointer to member function
1720 @item -Wsign-promo @r{(C++ only)}
1721 @opindex Wsign-promo
1722 Warn when overload resolution chooses a promotion from unsigned or
1723 enumerated type to a signed type, over a conversion to an unsigned type of
1724 the same size. Previous versions of G++ would try to preserve
1725 unsignedness, but the standard mandates the current behavior.
1727 @item -Wsynth @r{(C++ only)}
1729 @cindex warning for synthesized methods
1730 @cindex synthesized methods, warning
1731 Warn when G++'s synthesis behavior does not match that of cfront. For
1737 A& operator = (int);
1747 In this example, G++ will synthesize a default @samp{A& operator =
1748 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1751 @node Objective-C Dialect Options
1752 @section Options Controlling Objective-C Dialect
1754 @cindex compiler options, Objective-C
1755 @cindex Objective-C options, command line
1756 @cindex options, Objective-C
1757 (NOTE: This manual does not describe the Objective-C language itself. See
1758 @w{@uref{http://gcc.gnu.org/readings.html}} for references.)
1760 This section describes the command-line options that are only meaningful
1761 for Objective-C programs, but you can also use most of the GNU compiler
1762 options regardless of what language your program is in. For example,
1763 you might compile a file @code{some_class.m} like this:
1766 gcc -g -fgnu-runtime -O -c some_class.m
1770 In this example, @option{-fgnu-runtime} is an option meant only for
1771 Objective-C programs; you can use the other options with any language
1774 Here is a list of options that are @emph{only} for compiling Objective-C
1778 @item -fconstant-string-class=@var{class-name}
1779 @opindex fconstant-string-class
1780 Use @var{class-name} as the name of the class to instantiate for each
1781 literal string specified with the syntax @code{@@"@dots{}"}. The default
1782 class name is @code{NXConstantString} if the GNU runtime is being used, and
1783 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1784 @option{-fconstant-cfstrings} option, if also present, will override the
1785 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1786 to be laid out as constant CoreFoundation strings.
1789 @opindex fgnu-runtime
1790 Generate object code compatible with the standard GNU Objective-C
1791 runtime. This is the default for most types of systems.
1793 @item -fnext-runtime
1794 @opindex fnext-runtime
1795 Generate output compatible with the NeXT runtime. This is the default
1796 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1797 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1800 @item -fno-nil-receivers
1801 @opindex fno-nil-receivers
1802 Assume that all Objective-C message dispatches (e.g.,
1803 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1804 is not @code{nil}. This allows for more efficient entry points in the runtime to be
1805 used. Currently, this option is only available in conjunction with
1806 the NeXT runtime on Mac OS X 10.3 and later.
1808 @item -fobjc-exceptions
1809 @opindex fobjc-exceptions
1810 Enable syntactic support for structured exception handling in Objective-C,
1811 similar to what is offered by C++ and Java. Currently, this option is only
1812 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1820 @@catch (AnObjCClass *exc) @{
1827 @@catch (AnotherClass *exc) @{
1830 @@catch (id allOthers) @{
1840 The @code{@@throw} statement may appear anywhere in an Objective-C or
1841 Objective-C++ program; when used inside of a @code{@@catch} block, the
1842 @code{@@throw} may appear without an argument (as shown above), in which case
1843 the object caught by the @code{@@catch} will be rethrown.
1845 Note that only (pointers to) Objective-C objects may be thrown and
1846 caught using this scheme. When an object is thrown, it will be caught
1847 by the nearest @code{@@catch} clause capable of handling objects of that type,
1848 analogously to how @code{catch} blocks work in C++ and Java. A
1849 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1850 any and all Objective-C exceptions not caught by previous @code{@@catch}
1853 The @code{@@finally} clause, if present, will be executed upon exit from the
1854 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1855 regardless of whether any exceptions are thrown, caught or rethrown
1856 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1857 of the @code{finally} clause in Java.
1859 There are several caveats to using the new exception mechanism:
1863 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1864 idioms provided by the @code{NSException} class, the new
1865 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1866 systems, due to additional functionality needed in the (NeXT) Objective-C
1870 As mentioned above, the new exceptions do not support handling
1871 types other than Objective-C objects. Furthermore, when used from
1872 Objective-C++, the Objective-C exception model does not interoperate with C++
1873 exceptions at this time. This means you cannot @code{@@throw} an exception
1874 from Objective-C and @code{catch} it in C++, or vice versa
1875 (i.e., @code{throw @dots{} @@catch}).
1878 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1879 blocks for thread-safe execution:
1882 @@synchronized (ObjCClass *guard) @{
1887 Upon entering the @code{@@synchronized} block, a thread of execution shall
1888 first check whether a lock has been placed on the corresponding @code{guard}
1889 object by another thread. If it has, the current thread shall wait until
1890 the other thread relinquishes its lock. Once @code{guard} becomes available,
1891 the current thread will place its own lock on it, execute the code contained in
1892 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1893 making @code{guard} available to other threads).
1895 Unlike Java, Objective-C does not allow for entire methods to be marked
1896 @code{@@synchronized}. Note that throwing exceptions out of
1897 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1898 to be unlocked properly.
1900 @item -freplace-objc-classes
1901 @opindex freplace-objc-classes
1902 Emit a special marker instructing @command{ld(1)} not to statically link in
1903 the resulting object file, and allow @command{dyld(1)} to load it in at
1904 run time instead. This is used in conjunction with the Fix-and-Continue
1905 debugging mode, where the object file in question may be recompiled and
1906 dynamically reloaded in the course of program execution, without the need
1907 to restart the program itself. Currently, Fix-and-Continue functionality
1908 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1913 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1914 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1915 compile time) with static class references that get initialized at load time,
1916 which improves run-time performance. Specifying the @option{-fzero-link} flag
1917 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1918 to be retained. This is useful in Zero-Link debugging mode, since it allows
1919 for individual class implementations to be modified during program execution.
1923 Dump interface declarations for all classes seen in the source file to a
1924 file named @file{@var{sourcename}.decl}.
1927 @opindex Wno-protocol
1928 If a class is declared to implement a protocol, a warning is issued for
1929 every method in the protocol that is not implemented by the class. The
1930 default behavior is to issue a warning for every method not explicitly
1931 implemented in the class, even if a method implementation is inherited
1932 from the superclass. If you use the @code{-Wno-protocol} option, then
1933 methods inherited from the superclass are considered to be implemented,
1934 and no warning is issued for them.
1938 Warn if multiple methods of different types for the same selector are
1939 found during compilation. The check is performed on the list of methods
1940 in the final stage of compilation. Additionally, a check is performed
1941 for each selector appearing in a @code{@@selector(@dots{})}
1942 expression, and a corresponding method for that selector has been found
1943 during compilation. Because these checks scan the method table only at
1944 the end of compilation, these warnings are not produced if the final
1945 stage of compilation is not reached, for example because an error is
1946 found during compilation, or because the @code{-fsyntax-only} option is
1949 @item -Wundeclared-selector
1950 @opindex Wundeclared-selector
1951 Warn if a @code{@@selector(@dots{})} expression referring to an
1952 undeclared selector is found. A selector is considered undeclared if no
1953 method with that name has been declared before the
1954 @code{@@selector(@dots{})} expression, either explicitly in an
1955 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
1956 an @code{@@implementation} section. This option always performs its
1957 checks as soon as a @code{@@selector(@dots{})} expression is found,
1958 while @code{-Wselector} only performs its checks in the final stage of
1959 compilation. This also enforces the coding style convention
1960 that methods and selectors must be declared before being used.
1962 @item -print-objc-runtime-info
1963 @opindex print-objc-runtime-info
1964 Generate C header describing the largest structure that is passed by
1969 @node Language Independent Options
1970 @section Options to Control Diagnostic Messages Formatting
1971 @cindex options to control diagnostics formatting
1972 @cindex diagnostic messages
1973 @cindex message formatting
1975 Traditionally, diagnostic messages have been formatted irrespective of
1976 the output device's aspect (e.g.@: its width, @dots{}). The options described
1977 below can be used to control the diagnostic messages formatting
1978 algorithm, e.g.@: how many characters per line, how often source location
1979 information should be reported. Right now, only the C++ front end can
1980 honor these options. However it is expected, in the near future, that
1981 the remaining front ends would be able to digest them correctly.
1984 @item -fmessage-length=@var{n}
1985 @opindex fmessage-length
1986 Try to format error messages so that they fit on lines of about @var{n}
1987 characters. The default is 72 characters for @command{g++} and 0 for the rest of
1988 the front ends supported by GCC@. If @var{n} is zero, then no
1989 line-wrapping will be done; each error message will appear on a single
1992 @opindex fdiagnostics-show-location
1993 @item -fdiagnostics-show-location=once
1994 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
1995 reporter to emit @emph{once} source location information; that is, in
1996 case the message is too long to fit on a single physical line and has to
1997 be wrapped, the source location won't be emitted (as prefix) again,
1998 over and over, in subsequent continuation lines. This is the default
2001 @item -fdiagnostics-show-location=every-line
2002 Only meaningful in line-wrapping mode. Instructs the diagnostic
2003 messages reporter to emit the same source location information (as
2004 prefix) for physical lines that result from the process of breaking
2005 a message which is too long to fit on a single line.
2009 @node Warning Options
2010 @section Options to Request or Suppress Warnings
2011 @cindex options to control warnings
2012 @cindex warning messages
2013 @cindex messages, warning
2014 @cindex suppressing warnings
2016 Warnings are diagnostic messages that report constructions which
2017 are not inherently erroneous but which are risky or suggest there
2018 may have been an error.
2020 You can request many specific warnings with options beginning @samp{-W},
2021 for example @option{-Wimplicit} to request warnings on implicit
2022 declarations. Each of these specific warning options also has a
2023 negative form beginning @samp{-Wno-} to turn off warnings;
2024 for example, @option{-Wno-implicit}. This manual lists only one of the
2025 two forms, whichever is not the default.
2027 The following options control the amount and kinds of warnings produced
2028 by GCC; for further, language-specific options also refer to
2029 @ref{C++ Dialect Options} and @ref{Objective-C Dialect Options}.
2032 @cindex syntax checking
2034 @opindex fsyntax-only
2035 Check the code for syntax errors, but don't do anything beyond that.
2039 Issue all the warnings demanded by strict ISO C and ISO C++;
2040 reject all programs that use forbidden extensions, and some other
2041 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2042 version of the ISO C standard specified by any @option{-std} option used.
2044 Valid ISO C and ISO C++ programs should compile properly with or without
2045 this option (though a rare few will require @option{-ansi} or a
2046 @option{-std} option specifying the required version of ISO C)@. However,
2047 without this option, certain GNU extensions and traditional C and C++
2048 features are supported as well. With this option, they are rejected.
2050 @option{-pedantic} does not cause warning messages for use of the
2051 alternate keywords whose names begin and end with @samp{__}. Pedantic
2052 warnings are also disabled in the expression that follows
2053 @code{__extension__}. However, only system header files should use
2054 these escape routes; application programs should avoid them.
2055 @xref{Alternate Keywords}.
2057 Some users try to use @option{-pedantic} to check programs for strict ISO
2058 C conformance. They soon find that it does not do quite what they want:
2059 it finds some non-ISO practices, but not all---only those for which
2060 ISO C @emph{requires} a diagnostic, and some others for which
2061 diagnostics have been added.
2063 A feature to report any failure to conform to ISO C might be useful in
2064 some instances, but would require considerable additional work and would
2065 be quite different from @option{-pedantic}. We don't have plans to
2066 support such a feature in the near future.
2068 Where the standard specified with @option{-std} represents a GNU
2069 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2070 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2071 extended dialect is based. Warnings from @option{-pedantic} are given
2072 where they are required by the base standard. (It would not make sense
2073 for such warnings to be given only for features not in the specified GNU
2074 C dialect, since by definition the GNU dialects of C include all
2075 features the compiler supports with the given option, and there would be
2076 nothing to warn about.)
2078 @item -pedantic-errors
2079 @opindex pedantic-errors
2080 Like @option{-pedantic}, except that errors are produced rather than
2085 Inhibit all warning messages.
2089 Inhibit warning messages about the use of @samp{#import}.
2091 @item -Wchar-subscripts
2092 @opindex Wchar-subscripts
2093 Warn if an array subscript has type @code{char}. This is a common cause
2094 of error, as programmers often forget that this type is signed on some
2099 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2100 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2102 @item -Wfatal-errors
2103 @opindex Wfatal-errors
2104 This option causes the compiler to abort compilation on the first error
2105 occurred rather than trying to keep going and printing further error
2110 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2111 the arguments supplied have types appropriate to the format string
2112 specified, and that the conversions specified in the format string make
2113 sense. This includes standard functions, and others specified by format
2114 attributes (@pxref{Function Attributes}), in the @code{printf},
2115 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2116 not in the C standard) families.
2118 The formats are checked against the format features supported by GNU
2119 libc version 2.2. These include all ISO C90 and C99 features, as well
2120 as features from the Single Unix Specification and some BSD and GNU
2121 extensions. Other library implementations may not support all these
2122 features; GCC does not support warning about features that go beyond a
2123 particular library's limitations. However, if @option{-pedantic} is used
2124 with @option{-Wformat}, warnings will be given about format features not
2125 in the selected standard version (but not for @code{strfmon} formats,
2126 since those are not in any version of the C standard). @xref{C Dialect
2127 Options,,Options Controlling C Dialect}.
2129 Since @option{-Wformat} also checks for null format arguments for
2130 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2132 @option{-Wformat} is included in @option{-Wall}. For more control over some
2133 aspects of format checking, the options @option{-Wformat-y2k},
2134 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2135 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2136 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2139 @opindex Wformat-y2k
2140 If @option{-Wformat} is specified, also warn about @code{strftime}
2141 formats which may yield only a two-digit year.
2143 @item -Wno-format-extra-args
2144 @opindex Wno-format-extra-args
2145 If @option{-Wformat} is specified, do not warn about excess arguments to a
2146 @code{printf} or @code{scanf} format function. The C standard specifies
2147 that such arguments are ignored.
2149 Where the unused arguments lie between used arguments that are
2150 specified with @samp{$} operand number specifications, normally
2151 warnings are still given, since the implementation could not know what
2152 type to pass to @code{va_arg} to skip the unused arguments. However,
2153 in the case of @code{scanf} formats, this option will suppress the
2154 warning if the unused arguments are all pointers, since the Single
2155 Unix Specification says that such unused arguments are allowed.
2157 @item -Wno-format-zero-length
2158 @opindex Wno-format-zero-length
2159 If @option{-Wformat} is specified, do not warn about zero-length formats.
2160 The C standard specifies that zero-length formats are allowed.
2162 @item -Wformat-nonliteral
2163 @opindex Wformat-nonliteral
2164 If @option{-Wformat} is specified, also warn if the format string is not a
2165 string literal and so cannot be checked, unless the format function
2166 takes its format arguments as a @code{va_list}.
2168 @item -Wformat-security
2169 @opindex Wformat-security
2170 If @option{-Wformat} is specified, also warn about uses of format
2171 functions that represent possible security problems. At present, this
2172 warns about calls to @code{printf} and @code{scanf} functions where the
2173 format string is not a string literal and there are no format arguments,
2174 as in @code{printf (foo);}. This may be a security hole if the format
2175 string came from untrusted input and contains @samp{%n}. (This is
2176 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2177 in future warnings may be added to @option{-Wformat-security} that are not
2178 included in @option{-Wformat-nonliteral}.)
2182 Enable @option{-Wformat} plus format checks not included in
2183 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2184 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2188 Warn about passing a null pointer for arguments marked as
2189 requiring a non-null value by the @code{nonnull} function attribute.
2191 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2192 can be disabled with the @option{-Wno-nonnull} option.
2194 @item -Winit-self @r{(C, C++, and Objective-C only)}
2196 Warn about uninitialized variables which are initialized with themselves.
2197 Note this option can only be used with the @option{-Wuninitialized} option,
2198 which in turn only works with @option{-O1} and above.
2200 For example, GCC will warn about @code{i} being uninitialized in the
2201 following snippet only when @option{-Winit-self} has been specified:
2212 @item -Wimplicit-int
2213 @opindex Wimplicit-int
2214 Warn when a declaration does not specify a type.
2216 @item -Wimplicit-function-declaration
2217 @itemx -Werror-implicit-function-declaration
2218 @opindex Wimplicit-function-declaration
2219 @opindex Werror-implicit-function-declaration
2220 Give a warning (or error) whenever a function is used before being
2225 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2229 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2230 function with external linkage, returning int, taking either zero
2231 arguments, two, or three arguments of appropriate types.
2233 @item -Wmissing-braces
2234 @opindex Wmissing-braces
2235 Warn if an aggregate or union initializer is not fully bracketed. In
2236 the following example, the initializer for @samp{a} is not fully
2237 bracketed, but that for @samp{b} is fully bracketed.
2240 int a[2][2] = @{ 0, 1, 2, 3 @};
2241 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2244 @item -Wmissing-include-dirs @r{(C, C++, and Objective-C only)}
2245 @opindex Wmissing-include-dirs
2246 Warn if a user-supplied include directory does not exist.
2249 @opindex Wparentheses
2250 Warn if parentheses are omitted in certain contexts, such
2251 as when there is an assignment in a context where a truth value
2252 is expected, or when operators are nested whose precedence people
2253 often get confused about. Only the warning for an assignment used as
2254 a truth value is supported when compiling C++; the other warnings are
2255 only supported when compiling C@.
2257 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2258 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2259 interpretation from that of ordinary mathematical notation.
2261 Also warn about constructions where there may be confusion to which
2262 @code{if} statement an @code{else} branch belongs. Here is an example of
2277 In C, every @code{else} branch belongs to the innermost possible @code{if}
2278 statement, which in this example is @code{if (b)}. This is often not
2279 what the programmer expected, as illustrated in the above example by
2280 indentation the programmer chose. When there is the potential for this
2281 confusion, GCC will issue a warning when this flag is specified.
2282 To eliminate the warning, add explicit braces around the innermost
2283 @code{if} statement so there is no way the @code{else} could belong to
2284 the enclosing @code{if}. The resulting code would look like this:
2300 @item -Wsequence-point
2301 @opindex Wsequence-point
2302 Warn about code that may have undefined semantics because of violations
2303 of sequence point rules in the C standard.
2305 The C standard defines the order in which expressions in a C program are
2306 evaluated in terms of @dfn{sequence points}, which represent a partial
2307 ordering between the execution of parts of the program: those executed
2308 before the sequence point, and those executed after it. These occur
2309 after the evaluation of a full expression (one which is not part of a
2310 larger expression), after the evaluation of the first operand of a
2311 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2312 function is called (but after the evaluation of its arguments and the
2313 expression denoting the called function), and in certain other places.
2314 Other than as expressed by the sequence point rules, the order of
2315 evaluation of subexpressions of an expression is not specified. All
2316 these rules describe only a partial order rather than a total order,
2317 since, for example, if two functions are called within one expression
2318 with no sequence point between them, the order in which the functions
2319 are called is not specified. However, the standards committee have
2320 ruled that function calls do not overlap.
2322 It is not specified when between sequence points modifications to the
2323 values of objects take effect. Programs whose behavior depends on this
2324 have undefined behavior; the C standard specifies that ``Between the
2325 previous and next sequence point an object shall have its stored value
2326 modified at most once by the evaluation of an expression. Furthermore,
2327 the prior value shall be read only to determine the value to be
2328 stored.''. If a program breaks these rules, the results on any
2329 particular implementation are entirely unpredictable.
2331 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2332 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2333 diagnosed by this option, and it may give an occasional false positive
2334 result, but in general it has been found fairly effective at detecting
2335 this sort of problem in programs.
2337 The present implementation of this option only works for C programs. A
2338 future implementation may also work for C++ programs.
2340 The C standard is worded confusingly, therefore there is some debate
2341 over the precise meaning of the sequence point rules in subtle cases.
2342 Links to discussions of the problem, including proposed formal
2343 definitions, may be found on the GCC readings page, at
2344 @w{@uref{http://gcc.gnu.org/readings.html}}.
2347 @opindex Wreturn-type
2348 Warn whenever a function is defined with a return-type that defaults to
2349 @code{int}. Also warn about any @code{return} statement with no
2350 return-value in a function whose return-type is not @code{void}.
2352 For C++, a function without return type always produces a diagnostic
2353 message, even when @option{-Wno-return-type} is specified. The only
2354 exceptions are @samp{main} and functions defined in system headers.
2358 Warn whenever a @code{switch} statement has an index of enumerated type
2359 and lacks a @code{case} for one or more of the named codes of that
2360 enumeration. (The presence of a @code{default} label prevents this
2361 warning.) @code{case} labels outside the enumeration range also
2362 provoke warnings when this option is used.
2364 @item -Wswitch-default
2365 @opindex Wswitch-switch
2366 Warn whenever a @code{switch} statement does not have a @code{default}
2370 @opindex Wswitch-enum
2371 Warn whenever a @code{switch} statement has an index of enumerated type
2372 and lacks a @code{case} for one or more of the named codes of that
2373 enumeration. @code{case} labels outside the enumeration range also
2374 provoke warnings when this option is used.
2378 Warn if any trigraphs are encountered that might change the meaning of
2379 the program (trigraphs within comments are not warned about).
2381 @item -Wunused-function
2382 @opindex Wunused-function
2383 Warn whenever a static function is declared but not defined or a
2384 non\-inline static function is unused.
2386 @item -Wunused-label
2387 @opindex Wunused-label
2388 Warn whenever a label is declared but not used.
2390 To suppress this warning use the @samp{unused} attribute
2391 (@pxref{Variable Attributes}).
2393 @item -Wunused-parameter
2394 @opindex Wunused-parameter
2395 Warn whenever a function parameter is unused aside from its declaration.
2397 To suppress this warning use the @samp{unused} attribute
2398 (@pxref{Variable Attributes}).
2400 @item -Wunused-variable
2401 @opindex Wunused-variable
2402 Warn whenever a local variable or non-constant static variable is unused
2403 aside from its declaration
2405 To suppress this warning use the @samp{unused} attribute
2406 (@pxref{Variable Attributes}).
2408 @item -Wunused-value
2409 @opindex Wunused-value
2410 Warn whenever a statement computes a result that is explicitly not used.
2412 To suppress this warning cast the expression to @samp{void}.
2416 All the above @option{-Wunused} options combined.
2418 In order to get a warning about an unused function parameter, you must
2419 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2420 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2422 @item -Wuninitialized
2423 @opindex Wuninitialized
2424 Warn if an automatic variable is used without first being initialized or
2425 if a variable may be clobbered by a @code{setjmp} call.
2427 These warnings are possible only in optimizing compilation,
2428 because they require data flow information that is computed only
2429 when optimizing. If you don't specify @option{-O}, you simply won't
2432 If you want to warn about code which uses the uninitialized value of the
2433 variable in its own initializer, use the @option{-Winit-self} option.
2435 These warnings occur only for variables that are candidates for
2436 register allocation. Therefore, they do not occur for a variable that
2437 is declared @code{volatile}, or whose address is taken, or whose size
2438 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2439 structures, unions or arrays, even when they are in registers.
2441 Note that there may be no warning about a variable that is used only
2442 to compute a value that itself is never used, because such
2443 computations may be deleted by data flow analysis before the warnings
2446 These warnings are made optional because GCC is not smart
2447 enough to see all the reasons why the code might be correct
2448 despite appearing to have an error. Here is one example of how
2469 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2470 always initialized, but GCC doesn't know this. Here is
2471 another common case:
2476 if (change_y) save_y = y, y = new_y;
2478 if (change_y) y = save_y;
2483 This has no bug because @code{save_y} is used only if it is set.
2485 @cindex @code{longjmp} warnings
2486 This option also warns when a non-volatile automatic variable might be
2487 changed by a call to @code{longjmp}. These warnings as well are possible
2488 only in optimizing compilation.
2490 The compiler sees only the calls to @code{setjmp}. It cannot know
2491 where @code{longjmp} will be called; in fact, a signal handler could
2492 call it at any point in the code. As a result, you may get a warning
2493 even when there is in fact no problem because @code{longjmp} cannot
2494 in fact be called at the place which would cause a problem.
2496 Some spurious warnings can be avoided if you declare all the functions
2497 you use that never return as @code{noreturn}. @xref{Function
2500 @item -Wunknown-pragmas
2501 @opindex Wunknown-pragmas
2502 @cindex warning for unknown pragmas
2503 @cindex unknown pragmas, warning
2504 @cindex pragmas, warning of unknown
2505 Warn when a #pragma directive is encountered which is not understood by
2506 GCC@. If this command line option is used, warnings will even be issued
2507 for unknown pragmas in system header files. This is not the case if
2508 the warnings were only enabled by the @option{-Wall} command line option.
2510 @item -Wstrict-aliasing
2511 @opindex Wstrict-aliasing
2512 This option is only active when @option{-fstrict-aliasing} is active.
2513 It warns about code which might break the strict aliasing rules that the
2514 compiler is using for optimization. The warning does not catch all
2515 cases, but does attempt to catch the more common pitfalls. It is
2516 included in @option{-Wall}.
2518 @item -Wstrict-aliasing=2
2519 @opindex Wstrict-aliasing=2
2520 This option is only active when @option{-fstrict-aliasing} is active.
2521 It warns about all code which might break the strict aliasing rules that the
2522 compiler is using for optimization. This warning catches all cases, but
2523 it will also give a warning for some ambiguous cases that are safe.
2527 All of the above @samp{-W} options combined. This enables all the
2528 warnings about constructions that some users consider questionable, and
2529 that are easy to avoid (or modify to prevent the warning), even in
2530 conjunction with macros. This also enables some language-specific
2531 warnings described in @ref{C++ Dialect Options} and
2532 @ref{Objective-C Dialect Options}.
2535 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2536 Some of them warn about constructions that users generally do not
2537 consider questionable, but which occasionally you might wish to check
2538 for; others warn about constructions that are necessary or hard to avoid
2539 in some cases, and there is no simple way to modify the code to suppress
2546 (This option used to be called @option{-W}. The older name is still
2547 supported, but the newer name is more descriptive.) Print extra warning
2548 messages for these events:
2552 A function can return either with or without a value. (Falling
2553 off the end of the function body is considered returning without
2554 a value.) For example, this function would evoke such a
2568 An expression-statement or the left-hand side of a comma expression
2569 contains no side effects.
2570 To suppress the warning, cast the unused expression to void.
2571 For example, an expression such as @samp{x[i,j]} will cause a warning,
2572 but @samp{x[(void)i,j]} will not.
2575 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2578 Storage-class specifiers like @code{static} are not the first things in
2579 a declaration. According to the C Standard, this usage is obsolescent.
2582 The return type of a function has a type qualifier such as @code{const}.
2583 Such a type qualifier has no effect, since the value returned by a
2584 function is not an lvalue. (But don't warn about the GNU extension of
2585 @code{volatile void} return types. That extension will be warned about
2586 if @option{-pedantic} is specified.)
2589 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2593 A comparison between signed and unsigned values could produce an
2594 incorrect result when the signed value is converted to unsigned.
2595 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2598 An aggregate has an initializer which does not initialize all members.
2599 For example, the following code would cause such a warning, because
2600 @code{x.h} would be implicitly initialized to zero:
2603 struct s @{ int f, g, h; @};
2604 struct s x = @{ 3, 4 @};
2608 A function parameter is declared without a type specifier in K&R-style
2616 An empty body occurs in an @samp{if} or @samp{else} statement.
2619 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2620 @samp{>}, or @samp{>=}.
2623 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2626 Any of several floating-point events that often indicate errors, such as
2627 overflow, underflow, loss of precision, etc.
2629 @item @r{(C++ only)}
2630 An enumerator and a non-enumerator both appear in a conditional expression.
2632 @item @r{(C++ only)}
2633 A non-static reference or non-static @samp{const} member appears in a
2634 class without constructors.
2636 @item @r{(C++ only)}
2637 Ambiguous virtual bases.
2639 @item @r{(C++ only)}
2640 Subscripting an array which has been declared @samp{register}.
2642 @item @r{(C++ only)}
2643 Taking the address of a variable which has been declared @samp{register}.
2645 @item @r{(C++ only)}
2646 A base class is not initialized in a derived class' copy constructor.
2649 @item -Wno-div-by-zero
2650 @opindex Wno-div-by-zero
2651 @opindex Wdiv-by-zero
2652 Do not warn about compile-time integer division by zero. Floating point
2653 division by zero is not warned about, as it can be a legitimate way of
2654 obtaining infinities and NaNs.
2656 @item -Wsystem-headers
2657 @opindex Wsystem-headers
2658 @cindex warnings from system headers
2659 @cindex system headers, warnings from
2660 Print warning messages for constructs found in system header files.
2661 Warnings from system headers are normally suppressed, on the assumption
2662 that they usually do not indicate real problems and would only make the
2663 compiler output harder to read. Using this command line option tells
2664 GCC to emit warnings from system headers as if they occurred in user
2665 code. However, note that using @option{-Wall} in conjunction with this
2666 option will @emph{not} warn about unknown pragmas in system
2667 headers---for that, @option{-Wunknown-pragmas} must also be used.
2670 @opindex Wfloat-equal
2671 Warn if floating point values are used in equality comparisons.
2673 The idea behind this is that sometimes it is convenient (for the
2674 programmer) to consider floating-point values as approximations to
2675 infinitely precise real numbers. If you are doing this, then you need
2676 to compute (by analyzing the code, or in some other way) the maximum or
2677 likely maximum error that the computation introduces, and allow for it
2678 when performing comparisons (and when producing output, but that's a
2679 different problem). In particular, instead of testing for equality, you
2680 would check to see whether the two values have ranges that overlap; and
2681 this is done with the relational operators, so equality comparisons are
2684 @item -Wtraditional @r{(C only)}
2685 @opindex Wtraditional
2686 Warn about certain constructs that behave differently in traditional and
2687 ISO C@. Also warn about ISO C constructs that have no traditional C
2688 equivalent, and/or problematic constructs which should be avoided.
2692 Macro parameters that appear within string literals in the macro body.
2693 In traditional C macro replacement takes place within string literals,
2694 but does not in ISO C@.
2697 In traditional C, some preprocessor directives did not exist.
2698 Traditional preprocessors would only consider a line to be a directive
2699 if the @samp{#} appeared in column 1 on the line. Therefore
2700 @option{-Wtraditional} warns about directives that traditional C
2701 understands but would ignore because the @samp{#} does not appear as the
2702 first character on the line. It also suggests you hide directives like
2703 @samp{#pragma} not understood by traditional C by indenting them. Some
2704 traditional implementations would not recognize @samp{#elif}, so it
2705 suggests avoiding it altogether.
2708 A function-like macro that appears without arguments.
2711 The unary plus operator.
2714 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2715 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2716 constants.) Note, these suffixes appear in macros defined in the system
2717 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2718 Use of these macros in user code might normally lead to spurious
2719 warnings, however GCC's integrated preprocessor has enough context to
2720 avoid warning in these cases.
2723 A function declared external in one block and then used after the end of
2727 A @code{switch} statement has an operand of type @code{long}.
2730 A non-@code{static} function declaration follows a @code{static} one.
2731 This construct is not accepted by some traditional C compilers.
2734 The ISO type of an integer constant has a different width or
2735 signedness from its traditional type. This warning is only issued if
2736 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2737 typically represent bit patterns, are not warned about.
2740 Usage of ISO string concatenation is detected.
2743 Initialization of automatic aggregates.
2746 Identifier conflicts with labels. Traditional C lacks a separate
2747 namespace for labels.
2750 Initialization of unions. If the initializer is zero, the warning is
2751 omitted. This is done under the assumption that the zero initializer in
2752 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2753 initializer warnings and relies on default initialization to zero in the
2757 Conversions by prototypes between fixed/floating point values and vice
2758 versa. The absence of these prototypes when compiling with traditional
2759 C would cause serious problems. This is a subset of the possible
2760 conversion warnings, for the full set use @option{-Wconversion}.
2763 Use of ISO C style function definitions. This warning intentionally is
2764 @emph{not} issued for prototype declarations or variadic functions
2765 because these ISO C features will appear in your code when using
2766 libiberty's traditional C compatibility macros, @code{PARAMS} and
2767 @code{VPARAMS}. This warning is also bypassed for nested functions
2768 because that feature is already a GCC extension and thus not relevant to
2769 traditional C compatibility.
2772 @item -Wdeclaration-after-statement @r{(C only)}
2773 @opindex Wdeclaration-after-statement
2774 Warn when a declaration is found after a statement in a block. This
2775 construct, known from C++, was introduced with ISO C99 and is by default
2776 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2777 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2781 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2783 @item -Wendif-labels
2784 @opindex Wendif-labels
2785 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2789 Warn whenever a local variable shadows another local variable, parameter or
2790 global variable or whenever a built-in function is shadowed.
2792 @item -Wlarger-than-@var{len}
2793 @opindex Wlarger-than
2794 Warn whenever an object of larger than @var{len} bytes is defined.
2796 @item -Wpointer-arith
2797 @opindex Wpointer-arith
2798 Warn about anything that depends on the ``size of'' a function type or
2799 of @code{void}. GNU C assigns these types a size of 1, for
2800 convenience in calculations with @code{void *} pointers and pointers
2803 @item -Wbad-function-cast @r{(C only)}
2804 @opindex Wbad-function-cast
2805 Warn whenever a function call is cast to a non-matching type.
2806 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2810 Warn whenever a pointer is cast so as to remove a type qualifier from
2811 the target type. For example, warn if a @code{const char *} is cast
2812 to an ordinary @code{char *}.
2815 @opindex Wcast-align
2816 Warn whenever a pointer is cast such that the required alignment of the
2817 target is increased. For example, warn if a @code{char *} is cast to
2818 an @code{int *} on machines where integers can only be accessed at
2819 two- or four-byte boundaries.
2821 @item -Wwrite-strings
2822 @opindex Wwrite-strings
2823 When compiling C, give string constants the type @code{const
2824 char[@var{length}]} so that
2825 copying the address of one into a non-@code{const} @code{char *}
2826 pointer will get a warning; when compiling C++, warn about the
2827 deprecated conversion from string constants to @code{char *}.
2828 These warnings will help you find at
2829 compile time code that can try to write into a string constant, but
2830 only if you have been very careful about using @code{const} in
2831 declarations and prototypes. Otherwise, it will just be a nuisance;
2832 this is why we did not make @option{-Wall} request these warnings.
2835 @opindex Wconversion
2836 Warn if a prototype causes a type conversion that is different from what
2837 would happen to the same argument in the absence of a prototype. This
2838 includes conversions of fixed point to floating and vice versa, and
2839 conversions changing the width or signedness of a fixed point argument
2840 except when the same as the default promotion.
2842 Also, warn if a negative integer constant expression is implicitly
2843 converted to an unsigned type. For example, warn about the assignment
2844 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2845 casts like @code{(unsigned) -1}.
2847 @item -Wsign-compare
2848 @opindex Wsign-compare
2849 @cindex warning for comparison of signed and unsigned values
2850 @cindex comparison of signed and unsigned values, warning
2851 @cindex signed and unsigned values, comparison warning
2852 Warn when a comparison between signed and unsigned values could produce
2853 an incorrect result when the signed value is converted to unsigned.
2854 This warning is also enabled by @option{-Wextra}; to get the other warnings
2855 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2857 @item -Waggregate-return
2858 @opindex Waggregate-return
2859 Warn if any functions that return structures or unions are defined or
2860 called. (In languages where you can return an array, this also elicits
2863 @item -Wstrict-prototypes @r{(C only)}
2864 @opindex Wstrict-prototypes
2865 Warn if a function is declared or defined without specifying the
2866 argument types. (An old-style function definition is permitted without
2867 a warning if preceded by a declaration which specifies the argument
2870 @item -Wold-style-definition @r{(C only)}
2871 @opindex Wold-style-definition
2872 Warn if an old-style function definition is used. A warning is given
2873 even if there is a previous prototype.
2875 @item -Wmissing-prototypes @r{(C only)}
2876 @opindex Wmissing-prototypes
2877 Warn if a global function is defined without a previous prototype
2878 declaration. This warning is issued even if the definition itself
2879 provides a prototype. The aim is to detect global functions that fail
2880 to be declared in header files.
2882 @item -Wmissing-declarations @r{(C only)}
2883 @opindex Wmissing-declarations
2884 Warn if a global function is defined without a previous declaration.
2885 Do so even if the definition itself provides a prototype.
2886 Use this option to detect global functions that are not declared in
2889 @item -Wmissing-noreturn
2890 @opindex Wmissing-noreturn
2891 Warn about functions which might be candidates for attribute @code{noreturn}.
2892 Note these are only possible candidates, not absolute ones. Care should
2893 be taken to manually verify functions actually do not ever return before
2894 adding the @code{noreturn} attribute, otherwise subtle code generation
2895 bugs could be introduced. You will not get a warning for @code{main} in
2896 hosted C environments.
2898 @item -Wmissing-format-attribute
2899 @opindex Wmissing-format-attribute
2901 If @option{-Wformat} is enabled, also warn about functions which might be
2902 candidates for @code{format} attributes. Note these are only possible
2903 candidates, not absolute ones. GCC will guess that @code{format}
2904 attributes might be appropriate for any function that calls a function
2905 like @code{vprintf} or @code{vscanf}, but this might not always be the
2906 case, and some functions for which @code{format} attributes are
2907 appropriate may not be detected. This option has no effect unless
2908 @option{-Wformat} is enabled (possibly by @option{-Wall}).
2910 @item -Wno-multichar
2911 @opindex Wno-multichar
2913 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
2914 Usually they indicate a typo in the user's code, as they have
2915 implementation-defined values, and should not be used in portable code.
2917 @item -Wno-deprecated-declarations
2918 @opindex Wno-deprecated-declarations
2919 Do not warn about uses of functions, variables, and types marked as
2920 deprecated by using the @code{deprecated} attribute.
2921 (@pxref{Function Attributes}, @pxref{Variable Attributes},
2922 @pxref{Type Attributes}.)
2926 Warn if a structure is given the packed attribute, but the packed
2927 attribute has no effect on the layout or size of the structure.
2928 Such structures may be mis-aligned for little benefit. For
2929 instance, in this code, the variable @code{f.x} in @code{struct bar}
2930 will be misaligned even though @code{struct bar} does not itself
2931 have the packed attribute:
2938 @} __attribute__((packed));
2948 Warn if padding is included in a structure, either to align an element
2949 of the structure or to align the whole structure. Sometimes when this
2950 happens it is possible to rearrange the fields of the structure to
2951 reduce the padding and so make the structure smaller.
2953 @item -Wredundant-decls
2954 @opindex Wredundant-decls
2955 Warn if anything is declared more than once in the same scope, even in
2956 cases where multiple declaration is valid and changes nothing.
2958 @item -Wnested-externs @r{(C only)}
2959 @opindex Wnested-externs
2960 Warn if an @code{extern} declaration is encountered within a function.
2962 @item -Wunreachable-code
2963 @opindex Wunreachable-code
2964 Warn if the compiler detects that code will never be executed.
2966 This option is intended to warn when the compiler detects that at
2967 least a whole line of source code will never be executed, because
2968 some condition is never satisfied or because it is after a
2969 procedure that never returns.
2971 It is possible for this option to produce a warning even though there
2972 are circumstances under which part of the affected line can be executed,
2973 so care should be taken when removing apparently-unreachable code.
2975 For instance, when a function is inlined, a warning may mean that the
2976 line is unreachable in only one inlined copy of the function.
2978 This option is not made part of @option{-Wall} because in a debugging
2979 version of a program there is often substantial code which checks
2980 correct functioning of the program and is, hopefully, unreachable
2981 because the program does work. Another common use of unreachable
2982 code is to provide behavior which is selectable at compile-time.
2986 Warn if a function can not be inlined and it was declared as inline.
2987 Even with this option, the compiler will not warn about failures to
2988 inline functions declared in system headers.
2990 The compiler uses a variety of heuristics to determine whether or not
2991 to inline a function. For example, the compiler takes into account
2992 the size of the function being inlined and the the amount of inlining
2993 that has already been done in the current function. Therefore,
2994 seemingly insignificant changes in the source program can cause the
2995 warnings produced by @option{-Winline} to appear or disappear.
2997 @item -Wno-invalid-offsetof @r{(C++ only)}
2998 @opindex Wno-invalid-offsetof
2999 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3000 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3001 to a non-POD type is undefined. In existing C++ implementations,
3002 however, @samp{offsetof} typically gives meaningful results even when
3003 applied to certain kinds of non-POD types. (Such as a simple
3004 @samp{struct} that fails to be a POD type only by virtue of having a
3005 constructor.) This flag is for users who are aware that they are
3006 writing nonportable code and who have deliberately chosen to ignore the
3009 The restrictions on @samp{offsetof} may be relaxed in a future version
3010 of the C++ standard.
3013 @opindex Winvalid-pch
3014 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3015 the search path but can't be used.
3019 @opindex Wno-long-long
3020 Warn if @samp{long long} type is used. This is default. To inhibit
3021 the warning messages, use @option{-Wno-long-long}. Flags
3022 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3023 only when @option{-pedantic} flag is used.
3025 @item -Wvariadic-macros
3026 @opindex Wvariadic-macros
3027 @opindex Wno-variadic-macros
3028 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3029 alternate syntax when in pedantic ISO C99 mode. This is default.
3030 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3032 @item -Wdisabled-optimization
3033 @opindex Wdisabled-optimization
3034 Warn if a requested optimization pass is disabled. This warning does
3035 not generally indicate that there is anything wrong with your code; it
3036 merely indicates that GCC's optimizers were unable to handle the code
3037 effectively. Often, the problem is that your code is too big or too
3038 complex; GCC will refuse to optimize programs when the optimization
3039 itself is likely to take inordinate amounts of time.
3043 Make all warnings into errors.
3046 @node Debugging Options
3047 @section Options for Debugging Your Program or GCC
3048 @cindex options, debugging
3049 @cindex debugging information options
3051 GCC has various special options that are used for debugging
3052 either your program or GCC:
3057 Produce debugging information in the operating system's native format
3058 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
3061 On most systems that use stabs format, @option{-g} enables use of extra
3062 debugging information that only GDB can use; this extra information
3063 makes debugging work better in GDB but will probably make other debuggers
3065 refuse to read the program. If you want to control for certain whether
3066 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3067 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3069 Unlike most other C compilers, GCC allows you to use @option{-g} with
3070 @option{-O}. The shortcuts taken by optimized code may occasionally
3071 produce surprising results: some variables you declared may not exist
3072 at all; flow of control may briefly move where you did not expect it;
3073 some statements may not be executed because they compute constant
3074 results or their values were already at hand; some statements may
3075 execute in different places because they were moved out of loops.
3077 Nevertheless it proves possible to debug optimized output. This makes
3078 it reasonable to use the optimizer for programs that might have bugs.
3080 The following options are useful when GCC is generated with the
3081 capability for more than one debugging format.
3085 Produce debugging information for use by GDB@. This means to use the
3086 most expressive format available (DWARF 2, stabs, or the native format
3087 if neither of those are supported), including GDB extensions if at all
3092 Produce debugging information in stabs format (if that is supported),
3093 without GDB extensions. This is the format used by DBX on most BSD
3094 systems. On MIPS, Alpha and System V Release 4 systems this option
3095 produces stabs debugging output which is not understood by DBX or SDB@.
3096 On System V Release 4 systems this option requires the GNU assembler.
3098 @item -feliminate-unused-debug-symbols
3099 @opindex feliminate-unused-debug-symbols
3100 Produce debugging information in stabs format (if that is supported),
3101 for only symbols that are actually used.
3105 Produce debugging information in stabs format (if that is supported),
3106 using GNU extensions understood only by the GNU debugger (GDB)@. The
3107 use of these extensions is likely to make other debuggers crash or
3108 refuse to read the program.
3112 Produce debugging information in COFF format (if that is supported).
3113 This is the format used by SDB on most System V systems prior to
3118 Produce debugging information in XCOFF format (if that is supported).
3119 This is the format used by the DBX debugger on IBM RS/6000 systems.
3123 Produce debugging information in XCOFF format (if that is supported),
3124 using GNU extensions understood only by the GNU debugger (GDB)@. The
3125 use of these extensions is likely to make other debuggers crash or
3126 refuse to read the program, and may cause assemblers other than the GNU
3127 assembler (GAS) to fail with an error.
3131 Produce debugging information in DWARF version 2 format (if that is
3132 supported). This is the format used by DBX on IRIX 6.
3136 Produce debugging information in VMS debug format (if that is
3137 supported). This is the format used by DEBUG on VMS systems.
3140 @itemx -ggdb@var{level}
3141 @itemx -gstabs@var{level}
3142 @itemx -gcoff@var{level}
3143 @itemx -gxcoff@var{level}
3144 @itemx -gvms@var{level}
3145 Request debugging information and also use @var{level} to specify how
3146 much information. The default level is 2.
3148 Level 1 produces minimal information, enough for making backtraces in
3149 parts of the program that you don't plan to debug. This includes
3150 descriptions of functions and external variables, but no information
3151 about local variables and no line numbers.
3153 Level 3 includes extra information, such as all the macro definitions
3154 present in the program. Some debuggers support macro expansion when
3155 you use @option{-g3}.
3157 Note that in order to avoid confusion between DWARF1 debug level 2,
3158 and DWARF2 @option{-gdwarf-2} does not accept a concatenated debug
3159 level. Instead use an additional @option{-g@var{level}} option to
3160 change the debug level for DWARF2.
3162 @item -feliminate-dwarf2-dups
3163 @opindex feliminate-dwarf2-dups
3164 Compress DWARF2 debugging information by eliminating duplicated
3165 information about each symbol. This option only makes sense when
3166 generating DWARF2 debugging information with @option{-gdwarf-2}.
3168 @cindex @command{prof}
3171 Generate extra code to write profile information suitable for the
3172 analysis program @command{prof}. You must use this option when compiling
3173 the source files you want data about, and you must also use it when
3176 @cindex @command{gprof}
3179 Generate extra code to write profile information suitable for the
3180 analysis program @command{gprof}. You must use this option when compiling
3181 the source files you want data about, and you must also use it when
3186 Makes the compiler print out each function name as it is compiled, and
3187 print some statistics about each pass when it finishes.
3190 @opindex ftime-report
3191 Makes the compiler print some statistics about the time consumed by each
3192 pass when it finishes.
3195 @opindex fmem-report
3196 Makes the compiler print some statistics about permanent memory
3197 allocation when it finishes.
3199 @item -fprofile-arcs
3200 @opindex fprofile-arcs
3201 Add code so that program flow @dfn{arcs} are instrumented. During
3202 execution the program records how many times each branch and call is
3203 executed and how many times it is taken or returns. When the compiled
3204 program exits it saves this data to a file called
3205 @file{@var{auxname}.gcda} for each source file. The data may be used for
3206 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3207 test coverage analysis (@option{-ftest-coverage}). Each object file's
3208 @var{auxname} is generated from the name of the output file, if
3209 explicitly specified and it is not the final executable, otherwise it is
3210 the basename of the source file. In both cases any suffix is removed
3211 (e.g. @file{foo.gcda} for input file @file{dir/foo.c}, or
3212 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3217 Compile the source files with @option{-fprofile-arcs} plus optimization
3218 and code generation options. For test coverage analysis, use the
3219 additional @option{-ftest-coverage} option. You do not need to profile
3220 every source file in a program.
3223 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3224 (the latter implies the former).
3227 Run the program on a representative workload to generate the arc profile
3228 information. This may be repeated any number of times. You can run
3229 concurrent instances of your program, and provided that the file system
3230 supports locking, the data files will be correctly updated. Also
3231 @code{fork} calls are detected and correctly handled (double counting
3235 For profile-directed optimizations, compile the source files again with
3236 the same optimization and code generation options plus
3237 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3238 Control Optimization}).
3241 For test coverage analysis, use @command{gcov} to produce human readable
3242 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3243 @command{gcov} documentation for further information.
3247 With @option{-fprofile-arcs}, for each function of your program GCC
3248 creates a program flow graph, then finds a spanning tree for the graph.
3249 Only arcs that are not on the spanning tree have to be instrumented: the
3250 compiler adds code to count the number of times that these arcs are
3251 executed. When an arc is the only exit or only entrance to a block, the
3252 instrumentation code can be added to the block; otherwise, a new basic
3253 block must be created to hold the instrumentation code.
3255 @item -ftree-based-profiling
3256 @opindex ftree-based-profiling
3257 This option is used in addition to @option{-fprofile-arcs} or
3258 @option{-fbranch-probabilities} to control whether those optimizations
3259 are performed on a tree-based or rtl-based internal representation.
3260 If you use this option when compiling with @option{-fprofile-arcs},
3261 you must also use it when compiling later with @option{-fbranch-probabilities}.
3262 Currently the tree-based optimization is in an early stage of
3263 development, and this option is recommended only for those people
3264 working on improving it.
3267 @item -ftest-coverage
3268 @opindex ftest-coverage
3269 Produce a notes file that the @command{gcov} code-coverage utility
3270 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3271 show program coverage. Each source file's note file is called
3272 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3273 above for a description of @var{auxname} and instructions on how to
3274 generate test coverage data. Coverage data will match the source files
3275 more closely, if you do not optimize.
3277 @item -d@var{letters}
3279 Says to make debugging dumps during compilation at times specified by
3280 @var{letters}. This is used for debugging the compiler. The file names
3281 for most of the dumps are made by appending a pass number and a word to
3282 the @var{dumpname}. @var{dumpname} is generated from the name of the
3283 output file, if explicitly specified and it is not an executable,
3284 otherwise it is the basename of the source file. In both cases any
3285 suffix is removed (e.g. @file{foo.01.rtl} or @file{foo.02.sibling}).
3286 Here are the possible letters for use in @var{letters}, and their
3292 Annotate the assembler output with miscellaneous debugging information.
3295 Dump after computing branch probabilities, to @file{@var{file}.12.bp}.
3298 Dump after block reordering, to @file{@var{file}.32.bbro}.
3301 Dump after instruction combination, to the file @file{@var{file}.20.combine}.
3304 Dump after the first if conversion, to the file @file{@var{file}.14.ce1}.
3305 Also dump after the second if conversion, to the file @file{@var{file}.21.ce2}.
3308 Dump after branch target load optimization, to to @file{@var{file}.33.btl}.
3309 Also dump after delayed branch scheduling, to @file{@var{file}.37.dbr}.
3312 Dump all macro definitions, at the end of preprocessing, in addition to
3316 Dump after the third if conversion, to @file{@var{file}.31.ce3}.
3319 Dump after control and data flow analysis, to @file{@var{file}.11.cfg}.
3320 Also dump after life analysis, to @file{@var{file}.19.life}.
3323 Dump after global register allocation, to @file{@var{file}.26.greg}.
3326 Dump after GCSE, to @file{@var{file}.08.gcse}.
3327 Also dump after jump bypassing and control flow optimizations, to
3328 @file{@var{file}.10.bypass}.
3331 Dump after finalization of EH handling code, to @file{@var{file}.03.eh}.
3334 Dump after sibling call optimizations, to @file{@var{file}.02.sibling}.
3337 Dump after the first jump optimization, to @file{@var{file}.04.jump}.
3340 Dump after conversion from registers to stack, to @file{@var{file}.35.stack}.
3343 Dump after local register allocation, to @file{@var{file}.25.lreg}.
3346 Dump after loop optimization passes, to @file{@var{file}.09.loop} and
3347 @file{@var{file}.16.loop2}.
3350 Dump after modulo scheduling, to @file{@var{file}.23.sms}.
3353 Dump after performing the machine dependent reorganization pass, to
3354 @file{@var{file}.36.mach}.
3357 Dump after register renumbering, to @file{@var{file}.30.rnreg}.
3360 Dump after the register move pass, to @file{@var{file}.22.regmove}.
3363 Dump after post-reload optimizations, to @file{@var{file}.27.postreload}.
3366 Dump after RTL generation, to @file{@var{file}.01.rtl}.
3369 Dump after the second scheduling pass, to @file{@var{file}.34.sched2}.
3372 Dump after CSE (including the jump optimization that sometimes follows
3373 CSE), to @file{@var{file}.06.cse}.
3376 Dump after the first scheduling pass, to @file{@var{file}.24.sched}.
3379 Dump after the second CSE pass (including the jump optimization that
3380 sometimes follows CSE), to @file{@var{file}.18.cse2}.
3383 Dump after running tracer, to @file{@var{file}.15.tracer}.
3386 Dump after null pointer elimination pass to @file{@var{file}.05.null}.
3389 Dump callgraph and unit-at-a-time optimization @file{@var{file}.00.unit}.
3392 Dump after the value profile transformations, to @file{@var{file}.13.vpt}.
3393 Also dump after variable tracking, to @file{@var{file}.35.vartrack}.
3396 Dump after the second flow pass, to @file{@var{file}.28.flow2}.
3399 Dump after the peephole pass, to @file{@var{file}.29.peephole2}.
3402 Dump after constructing the web, to @file{@var{file}.17.web}.
3405 Produce all the dumps listed above.
3408 Produce a core dump whenever an error occurs.
3411 Print statistics on memory usage, at the end of the run, to
3415 Annotate the assembler output with a comment indicating which
3416 pattern and alternative was used. The length of each instruction is
3420 Dump the RTL in the assembler output as a comment before each instruction.
3421 Also turns on @option{-dp} annotation.
3424 For each of the other indicated dump files (except for
3425 @file{@var{file}.01.rtl}), dump a representation of the control flow graph
3426 suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3429 Just generate RTL for a function instead of compiling it. Usually used
3433 Dump debugging information during parsing, to standard error.
3436 @item -fdump-unnumbered
3437 @opindex fdump-unnumbered
3438 When doing debugging dumps (see @option{-d} option above), suppress instruction
3439 numbers and line number note output. This makes it more feasible to
3440 use diff on debugging dumps for compiler invocations with different
3441 options, in particular with and without @option{-g}.
3443 @item -fdump-translation-unit @r{(C and C++ only)}
3444 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3445 @opindex fdump-translation-unit
3446 Dump a representation of the tree structure for the entire translation
3447 unit to a file. The file name is made by appending @file{.tu} to the
3448 source file name. If the @samp{-@var{options}} form is used, @var{options}
3449 controls the details of the dump as described for the
3450 @option{-fdump-tree} options.
3452 @item -fdump-class-hierarchy @r{(C++ only)}
3453 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3454 @opindex fdump-class-hierarchy
3455 Dump a representation of each class's hierarchy and virtual function
3456 table layout to a file. The file name is made by appending @file{.class}
3457 to the source file name. If the @samp{-@var{options}} form is used,
3458 @var{options} controls the details of the dump as described for the
3459 @option{-fdump-tree} options.
3461 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3462 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3464 Control the dumping at various stages of processing the intermediate
3465 language tree to a file. The file name is generated by appending a switch
3466 specific suffix to the source file name. If the @samp{-@var{options}}
3467 form is used, @var{options} is a list of @samp{-} separated options that
3468 control the details of the dump. Not all options are applicable to all
3469 dumps, those which are not meaningful will be ignored. The following
3470 options are available
3474 Print the address of each node. Usually this is not meaningful as it
3475 changes according to the environment and source file. Its primary use
3476 is for tying up a dump file with a debug environment.
3478 Inhibit dumping of members of a scope or body of a function merely
3479 because that scope has been reached. Only dump such items when they
3480 are directly reachable by some other path. When dumping pretty-printed
3481 trees, this option inhibits dumping the bodies of control structures.
3483 Print a raw representation of the tree. By default, trees are
3484 pretty-printed into a C-like representation.
3486 Enable more detailed dumps (not honored by every dump option).
3488 Enable dumping various statistics about the pass (not honored by every dump
3491 Enable showing basic block boundaries (disabled in raw dumps).
3493 Enable showing virtual operands for every statement.
3495 Enable showing line numbers for statements.
3497 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3499 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3502 The following tree dumps are possible:
3506 Dump before any tree based optimization, to @file{@var{file}.original}.
3509 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3512 Dump after function inlining, to @file{@var{file}.inlined}.
3515 @opindex fdump-tree-gimple
3516 Dump each function before and after the gimplification pass to a file. The
3517 file name is made by appending @file{.gimple} to the source file name.
3520 @opindex fdump-tree-cfg
3521 Dump the control flow graph of each function to a file. The file name is
3522 made by appending @file{.cfg} to the source file name.
3525 @opindex fdump-tree-vcg
3526 Dump the control flow graph of each function to a file in VCG format. The
3527 file name is made by appending @file{.vcg} to the source file name. Note
3528 that if the file contains more than one function, the generated file cannot
3529 be used directly by VCG. You will need to cut and paste each function's
3530 graph into its own separate file first.
3533 @opindex fdump-tree-ch
3534 Dump each function after copying loop headers. The file name is made by
3535 appending @file{.ch} to the source file name.
3538 @opindex fdump-tree-ssa
3539 Dump SSA related information to a file. The file name is made by appending
3540 @file{.ssa} to the source file name.
3543 @opindex fdump-tree-alias
3544 Dump aliasing information for each function. The file name is made by
3545 appending @file{.alias} to the source file name.
3548 @opindex fdump-tree-ccp
3549 Dump each function after CCP. The file name is made by appending
3550 @file{.ccp} to the source file name.
3553 @opindex fdump-tree-pre
3554 Dump trees after partial redundancy elimination. The file name is made
3555 by appending @file{.pre} to the source file name.
3558 @opindex fdump-tree-fre
3559 Dump trees after full redundancy elimination. The file name is made
3560 by appending @file{.fre} to the source file name.
3563 @opindex fdump-tree-dce
3564 Dump each function after dead code elimination. The file name is made by
3565 appending @file{.dce} to the source file name.
3568 @opindex fdump-tree-mudflap
3569 Dump each function after adding mudflap instrumentation. The file name is
3570 made by appending @file{.mudflap} to the source file name.
3573 @opindex fdump-tree-sra
3574 Dump each function after performing scalar replacement of aggregates. The
3575 file name is made by appending @file{.sra} to the source file name.
3578 @opindex fdump-tree-dom
3579 Dump each function after applying dominator tree optimizations. The file
3580 name is made by appending @file{.dom} to the source file name.
3583 @opindex fdump-tree-dse
3584 Dump each function after applying dead store elimination. The file
3585 name is made by appending @file{.dse} to the source file name.
3588 @opindex fdump-tree-phiopt
3589 Dump each function after optimizing PHI nodes into straightline code. The file
3590 name is made by appending @file{.phiopt} to the source file name.
3593 @opindex fdump-tree-forwprop
3594 Dump each function after forward propagating single use variables. The file
3595 name is made by appending @file{.forwprop} to the source file name.
3598 @opindex fdump-tree-copyrename
3599 Dump each function after applying the copy rename optimization. The file
3600 name is made by appending @file{.copyrename} to the source file name.
3603 @opindex fdump-tree-nrv
3604 Dump each function after applying the named return value optimization on
3605 generic trees. The file name is made by appending @file{.nrv} to the source
3609 @opindex fdump-tree-all
3610 Enable all the available tree dumps with the flags provided in this option.
3613 @item -frandom-seed=@var{string}
3614 @opindex frandom-string
3615 This option provides a seed that GCC uses when it would otherwise use
3616 random numbers. It is used to generate certain symbol names
3617 that have to be different in every compiled file. It is also used to
3618 place unique stamps in coverage data files and the object files that
3619 produce them. You can use the @option{-frandom-seed} option to produce
3620 reproducibly identical object files.
3622 The @var{string} should be different for every file you compile.
3624 @item -fsched-verbose=@var{n}
3625 @opindex fsched-verbose
3626 On targets that use instruction scheduling, this option controls the
3627 amount of debugging output the scheduler prints. This information is
3628 written to standard error, unless @option{-dS} or @option{-dR} is
3629 specified, in which case it is output to the usual dump
3630 listing file, @file{.sched} or @file{.sched2} respectively. However
3631 for @var{n} greater than nine, the output is always printed to standard
3634 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3635 same information as @option{-dRS}. For @var{n} greater than one, it
3636 also output basic block probabilities, detailed ready list information
3637 and unit/insn info. For @var{n} greater than two, it includes RTL
3638 at abort point, control-flow and regions info. And for @var{n} over
3639 four, @option{-fsched-verbose} also includes dependence info.
3643 Store the usual ``temporary'' intermediate files permanently; place them
3644 in the current directory and name them based on the source file. Thus,
3645 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3646 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3647 preprocessed @file{foo.i} output file even though the compiler now
3648 normally uses an integrated preprocessor.
3652 Report the CPU time taken by each subprocess in the compilation
3653 sequence. For C source files, this is the compiler proper and assembler
3654 (plus the linker if linking is done). The output looks like this:
3661 The first number on each line is the ``user time,'' that is time spent
3662 executing the program itself. The second number is ``system time,''
3663 time spent executing operating system routines on behalf of the program.
3664 Both numbers are in seconds.
3666 @item -fvar-tracking
3667 @opindex fvar-tracking
3668 Run variable tracking pass. It computes where variables are stored at each
3669 position in code. Better debugging information is then generated
3670 (if the debugging information format supports this information).
3672 It is enabled by default when compiling with optimization (@option{-Os},
3673 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
3674 the debug info format supports it.
3676 @item -print-file-name=@var{library}
3677 @opindex print-file-name
3678 Print the full absolute name of the library file @var{library} that
3679 would be used when linking---and don't do anything else. With this
3680 option, GCC does not compile or link anything; it just prints the
3683 @item -print-multi-directory
3684 @opindex print-multi-directory
3685 Print the directory name corresponding to the multilib selected by any
3686 other switches present in the command line. This directory is supposed
3687 to exist in @env{GCC_EXEC_PREFIX}.
3689 @item -print-multi-lib
3690 @opindex print-multi-lib
3691 Print the mapping from multilib directory names to compiler switches
3692 that enable them. The directory name is separated from the switches by
3693 @samp{;}, and each switch starts with an @samp{@@} instead of the
3694 @samp{-}, without spaces between multiple switches. This is supposed to
3695 ease shell-processing.
3697 @item -print-prog-name=@var{program}
3698 @opindex print-prog-name
3699 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3701 @item -print-libgcc-file-name
3702 @opindex print-libgcc-file-name
3703 Same as @option{-print-file-name=libgcc.a}.
3705 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3706 but you do want to link with @file{libgcc.a}. You can do
3709 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3712 @item -print-search-dirs
3713 @opindex print-search-dirs
3714 Print the name of the configured installation directory and a list of
3715 program and library directories @command{gcc} will search---and don't do anything else.
3717 This is useful when @command{gcc} prints the error message
3718 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3719 To resolve this you either need to put @file{cpp0} and the other compiler
3720 components where @command{gcc} expects to find them, or you can set the environment
3721 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3722 Don't forget the trailing '/'.
3723 @xref{Environment Variables}.
3726 @opindex dumpmachine
3727 Print the compiler's target machine (for example,
3728 @samp{i686-pc-linux-gnu})---and don't do anything else.
3731 @opindex dumpversion
3732 Print the compiler version (for example, @samp{3.0})---and don't do
3737 Print the compiler's built-in specs---and don't do anything else. (This
3738 is used when GCC itself is being built.) @xref{Spec Files}.
3740 @item -feliminate-unused-debug-types
3741 @opindex feliminate-unused-debug-types
3742 Normally, when producing DWARF2 output, GCC will emit debugging
3743 information for all types declared in a compilation
3744 unit, regardless of whether or not they are actually used
3745 in that compilation unit. Sometimes this is useful, such as
3746 if, in the debugger, you want to cast a value to a type that is
3747 not actually used in your program (but is declared). More often,
3748 however, this results in a significant amount of wasted space.
3749 With this option, GCC will avoid producing debug symbol output
3750 for types that are nowhere used in the source file being compiled.
3753 @node Optimize Options
3754 @section Options That Control Optimization
3755 @cindex optimize options
3756 @cindex options, optimization
3758 These options control various sorts of optimizations.
3760 Without any optimization option, the compiler's goal is to reduce the
3761 cost of compilation and to make debugging produce the expected
3762 results. Statements are independent: if you stop the program with a
3763 breakpoint between statements, you can then assign a new value to any
3764 variable or change the program counter to any other statement in the
3765 function and get exactly the results you would expect from the source
3768 Turning on optimization flags makes the compiler attempt to improve
3769 the performance and/or code size at the expense of compilation time
3770 and possibly the ability to debug the program.
3772 The compiler performs optimization based on the knowledge it has of
3773 the program. Optimization levels @option{-O2} and above, in
3774 particular, enable @emph{unit-at-a-time} mode, which allows the
3775 compiler to consider information gained from later functions in
3776 the file when compiling a function. Compiling multiple files at
3777 once to a single output file in @emph{unit-at-a-time} mode allows
3778 the compiler to use information gained from all of the files when
3779 compiling each of them.
3781 Not all optimizations are controlled directly by a flag. Only
3782 optimizations that have a flag are listed.
3789 Optimize. Optimizing compilation takes somewhat more time, and a lot
3790 more memory for a large function.
3792 With @option{-O}, the compiler tries to reduce code size and execution
3793 time, without performing any optimizations that take a great deal of
3796 @option{-O} turns on the following optimization flags:
3797 @gccoptlist{-fdefer-pop @gol
3798 -fmerge-constants @gol
3800 -floop-optimize @gol
3801 -fif-conversion @gol
3802 -fif-conversion2 @gol
3803 -fdelayed-branch @gol
3804 -fguess-branch-probability @gol
3807 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
3808 where doing so does not interfere with debugging.
3812 Optimize even more. GCC performs nearly all supported optimizations
3813 that do not involve a space-speed tradeoff. The compiler does not
3814 perform loop unrolling or function inlining when you specify @option{-O2}.
3815 As compared to @option{-O}, this option increases both compilation time
3816 and the performance of the generated code.
3818 @option{-O2} turns on all optimization flags specified by @option{-O}. It
3819 also turns on the following optimization flags:
3820 @gccoptlist{-fforce-mem @gol
3821 -foptimize-sibling-calls @gol
3822 -fstrength-reduce @gol
3823 -fcse-follow-jumps -fcse-skip-blocks @gol
3824 -frerun-cse-after-loop -frerun-loop-opt @gol
3825 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
3826 -fdelete-null-pointer-checks @gol
3827 -fexpensive-optimizations @gol
3829 -fschedule-insns -fschedule-insns2 @gol
3830 -fsched-interblock -fsched-spec @gol
3833 -freorder-blocks -freorder-functions @gol
3834 -fstrict-aliasing @gol
3835 -funit-at-a-time @gol
3836 -falign-functions -falign-jumps @gol
3837 -falign-loops -falign-labels @gol
3840 Please note the warning under @option{-fgcse} about
3841 invoking @option{-O2} on programs that use computed gotos.
3845 Optimize yet more. @option{-O3} turns on all optimizations specified by
3846 @option{-O2} and also turns on the @option{-finline-functions},
3847 @option{-fweb} and @option{-fgcse-after-reload} options.
3851 Do not optimize. This is the default.
3855 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
3856 do not typically increase code size. It also performs further
3857 optimizations designed to reduce code size.
3859 @option{-Os} disables the following optimization flags:
3860 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
3861 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
3863 If you use multiple @option{-O} options, with or without level numbers,
3864 the last such option is the one that is effective.
3867 Options of the form @option{-f@var{flag}} specify machine-independent
3868 flags. Most flags have both positive and negative forms; the negative
3869 form of @option{-ffoo} would be @option{-fno-foo}. In the table
3870 below, only one of the forms is listed---the one you typically will
3871 use. You can figure out the other form by either removing @samp{no-}
3874 The following options control specific optimizations. They are either
3875 activated by @option{-O} options or are related to ones that are. You
3876 can use the following flags in the rare cases when ``fine-tuning'' of
3877 optimizations to be performed is desired.
3880 @item -fno-default-inline
3881 @opindex fno-default-inline
3882 Do not make member functions inline by default merely because they are
3883 defined inside the class scope (C++ only). Otherwise, when you specify
3884 @w{@option{-O}}, member functions defined inside class scope are compiled
3885 inline by default; i.e., you don't need to add @samp{inline} in front of
3886 the member function name.
3888 @item -fno-defer-pop
3889 @opindex fno-defer-pop
3890 Always pop the arguments to each function call as soon as that function
3891 returns. For machines which must pop arguments after a function call,
3892 the compiler normally lets arguments accumulate on the stack for several
3893 function calls and pops them all at once.
3895 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3899 Force memory operands to be copied into registers before doing
3900 arithmetic on them. This produces better code by making all memory
3901 references potential common subexpressions. When they are not common
3902 subexpressions, instruction combination should eliminate the separate
3905 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3908 @opindex fforce-addr
3909 Force memory address constants to be copied into registers before
3910 doing arithmetic on them. This may produce better code just as
3911 @option{-fforce-mem} may.
3913 @item -fomit-frame-pointer
3914 @opindex fomit-frame-pointer
3915 Don't keep the frame pointer in a register for functions that
3916 don't need one. This avoids the instructions to save, set up and
3917 restore frame pointers; it also makes an extra register available
3918 in many functions. @strong{It also makes debugging impossible on
3921 On some machines, such as the VAX, this flag has no effect, because
3922 the standard calling sequence automatically handles the frame pointer
3923 and nothing is saved by pretending it doesn't exist. The
3924 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
3925 whether a target machine supports this flag. @xref{Registers,,Register
3926 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
3928 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
3930 @item -foptimize-sibling-calls
3931 @opindex foptimize-sibling-calls
3932 Optimize sibling and tail recursive calls.
3934 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
3938 Don't pay attention to the @code{inline} keyword. Normally this option
3939 is used to keep the compiler from expanding any functions inline.
3940 Note that if you are not optimizing, no functions can be expanded inline.
3942 @item -finline-functions
3943 @opindex finline-functions
3944 Integrate all simple functions into their callers. The compiler
3945 heuristically decides which functions are simple enough to be worth
3946 integrating in this way.
3948 If all calls to a given function are integrated, and the function is
3949 declared @code{static}, then the function is normally not output as
3950 assembler code in its own right.
3952 Enabled at level @option{-O3}.
3954 @item -finline-limit=@var{n}
3955 @opindex finline-limit
3956 By default, GCC limits the size of functions that can be inlined. This flag
3957 allows the control of this limit for functions that are explicitly marked as
3958 inline (i.e., marked with the inline keyword or defined within the class
3959 definition in c++). @var{n} is the size of functions that can be inlined in
3960 number of pseudo instructions (not counting parameter handling). The default
3961 value of @var{n} is 600.
3962 Increasing this value can result in more inlined code at
3963 the cost of compilation time and memory consumption. Decreasing usually makes
3964 the compilation faster and less code will be inlined (which presumably
3965 means slower programs). This option is particularly useful for programs that
3966 use inlining heavily such as those based on recursive templates with C++.
3968 Inlining is actually controlled by a number of parameters, which may be
3969 specified individually by using @option{--param @var{name}=@var{value}}.
3970 The @option{-finline-limit=@var{n}} option sets some of these parameters
3974 @item max-inline-insns-single
3975 is set to @var{n}/2.
3976 @item max-inline-insns-auto
3977 is set to @var{n}/2.
3978 @item min-inline-insns
3979 is set to 130 or @var{n}/4, whichever is smaller.
3980 @item max-inline-insns-rtl
3984 See below for a documentation of the individual
3985 parameters controlling inlining.
3987 @emph{Note:} pseudo instruction represents, in this particular context, an
3988 abstract measurement of function's size. In no way, it represents a count
3989 of assembly instructions and as such its exact meaning might change from one
3990 release to an another.
3992 @item -fkeep-inline-functions
3993 @opindex fkeep-inline-functions
3994 Even if all calls to a given function are integrated, and the function
3995 is declared @code{static}, nevertheless output a separate run-time
3996 callable version of the function. This switch does not affect
3997 @code{extern inline} functions.
3999 @item -fkeep-static-consts
4000 @opindex fkeep-static-consts
4001 Emit variables declared @code{static const} when optimization isn't turned
4002 on, even if the variables aren't referenced.
4004 GCC enables this option by default. If you want to force the compiler to
4005 check if the variable was referenced, regardless of whether or not
4006 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4008 @item -fmerge-constants
4009 Attempt to merge identical constants (string constants and floating point
4010 constants) across compilation units.
4012 This option is the default for optimized compilation if the assembler and
4013 linker support it. Use @option{-fno-merge-constants} to inhibit this
4016 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4018 @item -fmerge-all-constants
4019 Attempt to merge identical constants and identical variables.
4021 This option implies @option{-fmerge-constants}. In addition to
4022 @option{-fmerge-constants} this considers e.g. even constant initialized
4023 arrays or initialized constant variables with integral or floating point
4024 types. Languages like C or C++ require each non-automatic variable to
4025 have distinct location, so using this option will result in non-conforming
4028 @item -fmodulo-sched
4029 @opindex fmodulo-sched
4030 Perform swing modulo scheduling immediately before the first scheduling
4031 pass. This pass looks at innermost loops and reorders their
4032 instructions by overlapping different iterations.
4036 Use a graph coloring register allocator. Currently this option is meant
4037 only for testing. Users should not specify this option, since it is not
4038 yet ready for production use.
4040 @item -fno-branch-count-reg
4041 @opindex fno-branch-count-reg
4042 Do not use ``decrement and branch'' instructions on a count register,
4043 but instead generate a sequence of instructions that decrement a
4044 register, compare it against zero, then branch based upon the result.
4045 This option is only meaningful on architectures that support such
4046 instructions, which include x86, PowerPC, IA-64 and S/390.
4048 The default is @option{-fbranch-count-reg}, enabled when
4049 @option{-fstrength-reduce} is enabled.
4051 @item -fno-function-cse
4052 @opindex fno-function-cse
4053 Do not put function addresses in registers; make each instruction that
4054 calls a constant function contain the function's address explicitly.
4056 This option results in less efficient code, but some strange hacks
4057 that alter the assembler output may be confused by the optimizations
4058 performed when this option is not used.
4060 The default is @option{-ffunction-cse}
4062 @item -fno-zero-initialized-in-bss
4063 @opindex fno-zero-initialized-in-bss
4064 If the target supports a BSS section, GCC by default puts variables that
4065 are initialized to zero into BSS@. This can save space in the resulting
4068 This option turns off this behavior because some programs explicitly
4069 rely on variables going to the data section. E.g., so that the
4070 resulting executable can find the beginning of that section and/or make
4071 assumptions based on that.
4073 The default is @option{-fzero-initialized-in-bss}.
4075 @item -fbounds-check
4076 @opindex fbounds-check
4077 For front-ends that support it, generate additional code to check that
4078 indices used to access arrays are within the declared range. This is
4079 currently only supported by the Java and Fortran front-ends, where
4080 this option defaults to true and false respectively.
4082 @item -fmudflap -fmudflapth -fmudflapir
4086 @cindex bounds checking
4088 For front-ends that support it (C and C++), instrument all risky
4089 pointer/array dereferencing operations, some standard library
4090 string/heap functions, and some other associated constructs with
4091 range/validity tests. Modules so instrumented should be immune to
4092 buffer overflows, invalid heap use, and some other classes of C/C++
4093 programming errors. The instrumentation relies on a separate runtime
4094 library (@file{libmudflap}), which will be linked into a program if
4095 @option{-fmudflap} is given at link time. Run-time behavior of the
4096 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4097 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4100 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4101 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4102 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4103 instrumentation should ignore pointer reads. This produces less
4104 instrumentation (and therefore faster execution) and still provides
4105 some protection against outright memory corrupting writes, but allows
4106 erroneously read data to propagate within a program.
4108 @item -fstrength-reduce
4109 @opindex fstrength-reduce
4110 Perform the optimizations of loop strength reduction and
4111 elimination of iteration variables.
4113 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4115 @item -fthread-jumps
4116 @opindex fthread-jumps
4117 Perform optimizations where we check to see if a jump branches to a
4118 location where another comparison subsumed by the first is found. If
4119 so, the first branch is redirected to either the destination of the
4120 second branch or a point immediately following it, depending on whether
4121 the condition is known to be true or false.
4123 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4125 @item -fcse-follow-jumps
4126 @opindex fcse-follow-jumps
4127 In common subexpression elimination, scan through jump instructions
4128 when the target of the jump is not reached by any other path. For
4129 example, when CSE encounters an @code{if} statement with an
4130 @code{else} clause, CSE will follow the jump when the condition
4133 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4135 @item -fcse-skip-blocks
4136 @opindex fcse-skip-blocks
4137 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4138 follow jumps which conditionally skip over blocks. When CSE
4139 encounters a simple @code{if} statement with no else clause,
4140 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4141 body of the @code{if}.
4143 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4145 @item -frerun-cse-after-loop
4146 @opindex frerun-cse-after-loop
4147 Re-run common subexpression elimination after loop optimizations has been
4150 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4152 @item -frerun-loop-opt
4153 @opindex frerun-loop-opt
4154 Run the loop optimizer twice.
4156 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4160 Perform a global common subexpression elimination pass.
4161 This pass also performs global constant and copy propagation.
4163 @emph{Note:} When compiling a program using computed gotos, a GCC
4164 extension, you may get better runtime performance if you disable
4165 the global common subexpression elimination pass by adding
4166 @option{-fno-gcse} to the command line.
4168 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4172 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4173 attempt to move loads which are only killed by stores into themselves. This
4174 allows a loop containing a load/store sequence to be changed to a load outside
4175 the loop, and a copy/store within the loop.
4177 Enabled by default when gcse is enabled.
4181 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4182 global common subexpression elimination. This pass will attempt to move
4183 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4184 loops containing a load/store sequence can be changed to a load before
4185 the loop and a store after the loop.
4187 Enabled by default when gcse is enabled.
4191 When @option{-fgcse-las} is enabled, the global common subexpression
4192 elimination pass eliminates redundant loads that come after stores to the
4193 same memory location (both partial and full redundancies).
4195 Enabled by default when gcse is enabled.
4197 @item -fgcse-after-reload
4198 @opindex fgcse-after-reload
4199 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4200 pass is performed after reload. The purpose of this pass is to cleanup
4203 @item -floop-optimize
4204 @opindex floop-optimize
4205 Perform loop optimizations: move constant expressions out of loops, simplify
4206 exit test conditions and optionally do strength-reduction and loop unrolling as
4209 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4211 @item -floop-optimize2
4212 @opindex floop-optimize2
4213 Perform loop optimizations using the new loop optimizer. The optimizations
4214 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4217 @item -fcrossjumping
4218 @opindex crossjumping
4219 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4220 resulting code may or may not perform better than without cross-jumping.
4222 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4224 @item -fif-conversion
4225 @opindex if-conversion
4226 Attempt to transform conditional jumps into branch-less equivalents. This
4227 include use of conditional moves, min, max, set flags and abs instructions, and
4228 some tricks doable by standard arithmetics. The use of conditional execution
4229 on chips where it is available is controlled by @code{if-conversion2}.
4231 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4233 @item -fif-conversion2
4234 @opindex if-conversion2
4235 Use conditional execution (where available) to transform conditional jumps into
4236 branch-less equivalents.
4238 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4240 @item -fdelete-null-pointer-checks
4241 @opindex fdelete-null-pointer-checks
4242 Use global dataflow analysis to identify and eliminate useless checks
4243 for null pointers. The compiler assumes that dereferencing a null
4244 pointer would have halted the program. If a pointer is checked after
4245 it has already been dereferenced, it cannot be null.
4247 In some environments, this assumption is not true, and programs can
4248 safely dereference null pointers. Use
4249 @option{-fno-delete-null-pointer-checks} to disable this optimization
4250 for programs which depend on that behavior.
4252 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4254 @item -fexpensive-optimizations
4255 @opindex fexpensive-optimizations
4256 Perform a number of minor optimizations that are relatively expensive.
4258 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4260 @item -foptimize-register-move
4262 @opindex foptimize-register-move
4264 Attempt to reassign register numbers in move instructions and as
4265 operands of other simple instructions in order to maximize the amount of
4266 register tying. This is especially helpful on machines with two-operand
4269 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4272 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4274 @item -fdelayed-branch
4275 @opindex fdelayed-branch
4276 If supported for the target machine, attempt to reorder instructions
4277 to exploit instruction slots available after delayed branch
4280 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4282 @item -fschedule-insns
4283 @opindex fschedule-insns
4284 If supported for the target machine, attempt to reorder instructions to
4285 eliminate execution stalls due to required data being unavailable. This
4286 helps machines that have slow floating point or memory load instructions
4287 by allowing other instructions to be issued until the result of the load
4288 or floating point instruction is required.
4290 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4292 @item -fschedule-insns2
4293 @opindex fschedule-insns2
4294 Similar to @option{-fschedule-insns}, but requests an additional pass of
4295 instruction scheduling after register allocation has been done. This is
4296 especially useful on machines with a relatively small number of
4297 registers and where memory load instructions take more than one cycle.
4299 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4301 @item -fno-sched-interblock
4302 @opindex fno-sched-interblock
4303 Don't schedule instructions across basic blocks. This is normally
4304 enabled by default when scheduling before register allocation, i.e.@:
4305 with @option{-fschedule-insns} or at @option{-O2} or higher.
4307 @item -fno-sched-spec
4308 @opindex fno-sched-spec
4309 Don't allow speculative motion of non-load instructions. This is normally
4310 enabled by default when scheduling before register allocation, i.e.@:
4311 with @option{-fschedule-insns} or at @option{-O2} or higher.
4313 @item -fsched-spec-load
4314 @opindex fsched-spec-load
4315 Allow speculative motion of some load instructions. This only makes
4316 sense when scheduling before register allocation, i.e.@: with
4317 @option{-fschedule-insns} or at @option{-O2} or higher.
4319 @item -fsched-spec-load-dangerous
4320 @opindex fsched-spec-load-dangerous
4321 Allow speculative motion of more load instructions. This only makes
4322 sense when scheduling before register allocation, i.e.@: with
4323 @option{-fschedule-insns} or at @option{-O2} or higher.
4325 @item -fsched-stalled-insns=@var{n}
4326 @opindex fsched-stalled-insns
4327 Define how many insns (if any) can be moved prematurely from the queue
4328 of stalled insns into the ready list, during the second scheduling pass.
4330 @item -fsched-stalled-insns-dep=@var{n}
4331 @opindex fsched-stalled-insns-dep
4332 Define how many insn groups (cycles) will be examined for a dependency
4333 on a stalled insn that is candidate for premature removal from the queue
4334 of stalled insns. Has an effect only during the second scheduling pass,
4335 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4337 @item -fsched2-use-superblocks
4338 @opindex fsched2-use-superblocks
4339 When scheduling after register allocation, do use superblock scheduling
4340 algorithm. Superblock scheduling allows motion across basic block boundaries
4341 resulting on faster schedules. This option is experimental, as not all machine
4342 descriptions used by GCC model the CPU closely enough to avoid unreliable
4343 results from the algorithm.
4345 This only makes sense when scheduling after register allocation, i.e.@: with
4346 @option{-fschedule-insns2} or at @option{-O2} or higher.
4348 @item -fsched2-use-traces
4349 @opindex fsched2-use-traces
4350 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4351 allocation and additionally perform code duplication in order to increase the
4352 size of superblocks using tracer pass. See @option{-ftracer} for details on
4355 This mode should produce faster but significantly longer programs. Also
4356 without @code{-fbranch-probabilities} the traces constructed may not match the
4357 reality and hurt the performance. This only makes
4358 sense when scheduling after register allocation, i.e.@: with
4359 @option{-fschedule-insns2} or at @option{-O2} or higher.
4361 @item -fcaller-saves
4362 @opindex fcaller-saves
4363 Enable values to be allocated in registers that will be clobbered by
4364 function calls, by emitting extra instructions to save and restore the
4365 registers around such calls. Such allocation is done only when it
4366 seems to result in better code than would otherwise be produced.
4368 This option is always enabled by default on certain machines, usually
4369 those which have no call-preserved registers to use instead.
4371 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4374 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4375 enabled by default at -O and higher.
4378 Perform Full Redundancy Elimination (FRE) on trees. The difference
4379 between FRE and PRE is that FRE only considers expressions
4380 that are computed on all paths leading to the redundant computation.
4381 This analysis faster than PRE, though it exposes fewer redundancies.
4382 This flag is enabled by default at -O and higher.
4385 Perform sparse conditional constant propagation (CCP) on trees. This flag
4386 is enabled by default at -O and higher.
4389 Perform dead code elimination (DCE) on trees. This flag is enabled by
4390 default at -O and higher.
4392 @item -ftree-dominator-opts
4393 Perform dead code elimination (DCE) on trees. This flag is enabled by
4394 default at -O and higher.
4397 Perform loop header copying on trees. This is beneficial since it increases
4398 effectivity of code motion optimizations. It also saves one jump. This flag
4399 is enabled by default at -O and higher. It is not enabled for -Os, since it
4400 usually increases code size.
4402 @item -ftree-loop-optimize
4403 Perform loop optimizations on trees. This flag is enabled by default at -O
4407 Perform loop invariant motion on trees. This pass moves only invartiants that
4408 would be hard to handle on rtl level (function calls, operations that expand to
4409 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4410 operands of conditions that are invariant out of the loop, so that we can use
4411 just trivial invariantness analysis in loop unswitching. The pass also includes
4415 Perform scalar replacement of aggregates. This pass replaces structure
4416 references with scalars to prevent committing structures to memory too
4417 early. This flag is enabled by default at -O and higher.
4419 @item -ftree-copyrename
4420 Perform copy renaming on trees. This pass attempts to rename compiler
4421 temporaries to other variables at copy locations, usually resulting in
4422 variable names which more closely resemble the original variables. This flag
4423 is enabled by default at -O and higher.
4426 Perform temporary expression replacement during the SSA->normal phase. Single
4427 use/single def temporaries are replaced at their use location with their
4428 defining expression. This results in non-GIMPLE code, but gives the expanders
4429 much more complex trees to work on resulting in better RTL generation. This is
4430 enabled by default at -O and higher.
4433 Perform live range splitting during the SSA->normal phase. Distinct live
4434 ranges of a variable are split into unique variables, allowing for better
4435 optimization later. This is enabled by default at -O and higher.
4439 Perform tail duplication to enlarge superblock size. This transformation
4440 simplifies the control flow of the function allowing other optimizations to do
4443 @item -funroll-loops
4444 @opindex funroll-loops
4445 Unroll loops whose number of iterations can be determined at compile
4446 time or upon entry to the loop. @option{-funroll-loops} implies both
4447 @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 -funroll-all-loops
4451 @opindex funroll-all-loops
4452 Unroll all loops, even if their number of iterations is uncertain when
4453 the loop is entered. This usually makes programs run more slowly.
4454 @option{-funroll-all-loops} implies the same options as
4455 @option{-funroll-loops},
4457 @item -fprefetch-loop-arrays
4458 @opindex fprefetch-loop-arrays
4459 If supported by the target machine, generate instructions to prefetch
4460 memory to improve the performance of loops that access large arrays.
4462 @item -fmove-all-movables
4463 @opindex fmove-all-movables
4464 Forces all invariant computations in loops to be moved
4467 @item -freduce-all-givs
4468 @opindex freduce-all-givs
4469 Forces all general-induction variables in loops to be
4472 @emph{Note:} When compiling programs written in Fortran,
4473 @option{-fmove-all-movables} and @option{-freduce-all-givs} are enabled
4474 by default when you use the optimizer.
4476 These options may generate better or worse code; results are highly
4477 dependent on the structure of loops within the source code.
4479 These two options are intended to be removed someday, once
4480 they have helped determine the efficacy of various
4481 approaches to improving loop optimizations.
4483 Please contact @w{@email{gcc@@gcc.gnu.org}}, and describe how use of
4484 these options affects the performance of your production code.
4485 Examples of code that runs @emph{slower} when these options are
4486 @emph{enabled} are very valuable.
4489 @itemx -fno-peephole2
4490 @opindex fno-peephole
4491 @opindex fno-peephole2
4492 Disable any machine-specific peephole optimizations. The difference
4493 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4494 are implemented in the compiler; some targets use one, some use the
4495 other, a few use both.
4497 @option{-fpeephole} is enabled by default.
4498 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4500 @item -fno-guess-branch-probability
4501 @opindex fno-guess-branch-probability
4502 Do not guess branch probabilities using a randomized model.
4504 Sometimes GCC will opt to use a randomized model to guess branch
4505 probabilities, when none are available from either profiling feedback
4506 (@option{-fprofile-arcs}) or @samp{__builtin_expect}. This means that
4507 different runs of the compiler on the same program may produce different
4510 In a hard real-time system, people don't want different runs of the
4511 compiler to produce code that has different behavior; minimizing
4512 non-determinism is of paramount import. This switch allows users to
4513 reduce non-determinism, possibly at the expense of inferior
4516 The default is @option{-fguess-branch-probability} at levels
4517 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4519 @item -freorder-blocks
4520 @opindex freorder-blocks
4521 Reorder basic blocks in the compiled function in order to reduce number of
4522 taken branches and improve code locality.
4524 Enabled at levels @option{-O2}, @option{-O3}.
4526 @item -freorder-blocks-and-partition
4527 @opindex freorder-blocks-and-partition
4528 In addition to reordering basic blocks in the compiled function, in order
4529 to reduce number of taken branches, partitions hot and cold basic blocks
4530 into separate sections of the assembly and .o files, to improve
4531 paging and cache locality performance.
4533 @item -freorder-functions
4534 @opindex freorder-functions
4535 Reorder basic blocks in the compiled function in order to reduce number of
4536 taken branches and improve code locality. This is implemented by using special
4537 subsections @code{.text.hot} for most frequently executed functions and
4538 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4539 the linker so object file format must support named sections and linker must
4540 place them in a reasonable way.
4542 Also profile feedback must be available in to make this option effective. See
4543 @option{-fprofile-arcs} for details.
4545 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4547 @item -fstrict-aliasing
4548 @opindex fstrict-aliasing
4549 Allows the compiler to assume the strictest aliasing rules applicable to
4550 the language being compiled. For C (and C++), this activates
4551 optimizations based on the type of expressions. In particular, an
4552 object of one type is assumed never to reside at the same address as an
4553 object of a different type, unless the types are almost the same. For
4554 example, an @code{unsigned int} can alias an @code{int}, but not a
4555 @code{void*} or a @code{double}. A character type may alias any other
4558 Pay special attention to code like this:
4571 The practice of reading from a different union member than the one most
4572 recently written to (called ``type-punning'') is common. Even with
4573 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4574 is accessed through the union type. So, the code above will work as
4575 expected. However, this code might not:
4586 Every language that wishes to perform language-specific alias analysis
4587 should define a function that computes, given an @code{tree}
4588 node, an alias set for the node. Nodes in different alias sets are not
4589 allowed to alias. For an example, see the C front-end function
4590 @code{c_get_alias_set}.
4592 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4594 @item -falign-functions
4595 @itemx -falign-functions=@var{n}
4596 @opindex falign-functions
4597 Align the start of functions to the next power-of-two greater than
4598 @var{n}, skipping up to @var{n} bytes. For instance,
4599 @option{-falign-functions=32} aligns functions to the next 32-byte
4600 boundary, but @option{-falign-functions=24} would align to the next
4601 32-byte boundary only if this can be done by skipping 23 bytes or less.
4603 @option{-fno-align-functions} and @option{-falign-functions=1} are
4604 equivalent and mean that functions will not be aligned.
4606 Some assemblers only support this flag when @var{n} is a power of two;
4607 in that case, it is rounded up.
4609 If @var{n} is not specified or is zero, use a machine-dependent default.
4611 Enabled at levels @option{-O2}, @option{-O3}.
4613 @item -falign-labels
4614 @itemx -falign-labels=@var{n}
4615 @opindex falign-labels
4616 Align all branch targets to a power-of-two boundary, skipping up to
4617 @var{n} bytes like @option{-falign-functions}. This option can easily
4618 make code slower, because it must insert dummy operations for when the
4619 branch target is reached in the usual flow of the code.
4621 @option{-fno-align-labels} and @option{-falign-labels=1} are
4622 equivalent and mean that labels will not be aligned.
4624 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4625 are greater than this value, then their values are used instead.
4627 If @var{n} is not specified or is zero, use a machine-dependent default
4628 which is very likely to be @samp{1}, meaning no alignment.
4630 Enabled at levels @option{-O2}, @option{-O3}.
4633 @itemx -falign-loops=@var{n}
4634 @opindex falign-loops
4635 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4636 like @option{-falign-functions}. The hope is that the loop will be
4637 executed many times, which will make up for any execution of the dummy
4640 @option{-fno-align-loops} and @option{-falign-loops=1} are
4641 equivalent and mean that loops will not be aligned.
4643 If @var{n} is not specified or is zero, use a machine-dependent default.
4645 Enabled at levels @option{-O2}, @option{-O3}.
4648 @itemx -falign-jumps=@var{n}
4649 @opindex falign-jumps
4650 Align branch targets to a power-of-two boundary, for branch targets
4651 where the targets can only be reached by jumping, skipping up to @var{n}
4652 bytes like @option{-falign-functions}. In this case, no dummy operations
4655 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4656 equivalent and mean that loops will not be aligned.
4658 If @var{n} is not specified or is zero, use a machine-dependent default.
4660 Enabled at levels @option{-O2}, @option{-O3}.
4662 @item -funit-at-a-time
4663 @opindex funit-at-a-time
4664 Parse the whole compilation unit before starting to produce code.
4665 This allows some extra optimizations to take place but consumes
4666 more memory (in general). There are some compatibility issues
4667 with @emph{unit-at-at-time} mode:
4670 enabling @emph{unit-at-a-time} mode may change the order
4671 in which functions, variables, and top-level @code{asm} statements
4672 are emitted, and will likely break code relying on some particular
4673 ordering. The majority of such top-level @code{asm} statements,
4674 though, can be replaced by @code{section} attributes.
4677 @emph{unit-at-a-time} mode removes unreferenced static variables
4678 and functions are removed. This may result in undefined references
4679 when an @code{asm} statement refers directly to variables or functions
4680 that are otherwise unused. In that case either the variable/function
4681 shall be listed as an operand of the @code{asm} statement operand or,
4682 in the case of top-level @code{asm} statements the attribute @code{used}
4683 shall be used on the declaration.
4686 Static functions now can use non-standard passing conventions that
4687 may break @code{asm} statements calling functions directly. Again,
4688 attribute @code{used} will prevent this behavior.
4691 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
4692 but this scheme may not be supported by future releases of GCC.
4694 Enabled at levels @option{-O2}, @option{-O3}.
4698 Constructs webs as commonly used for register allocation purposes and assign
4699 each web individual pseudo register. This allows the register allocation pass
4700 to operate on pseudos directly, but also strengthens several other optimization
4701 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4702 however, make debugging impossible, since variables will no longer stay in a
4705 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
4706 on targets where the default format for debugging information supports
4709 @item -fno-cprop-registers
4710 @opindex fno-cprop-registers
4711 After register allocation and post-register allocation instruction splitting,
4712 we perform a copy-propagation pass to try to reduce scheduling dependencies
4713 and occasionally eliminate the copy.
4715 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4717 @item -fprofile-generate
4718 @opindex fprofile-generate
4720 Enable options usually used for instrumenting application to produce
4721 profile useful for later recompilation with profile feedback based
4722 optimization. You must use @code{-fprofile-generate} both when
4723 compiling and when linking your program.
4725 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
4728 @opindex fprofile-use
4729 Enable profile feedback directed optimizations, and optimizations
4730 generally profitable only with profile feedback available.
4732 The following options are enabled: @code{-fbranch-probabilities},
4733 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
4737 The following options control compiler behavior regarding floating
4738 point arithmetic. These options trade off between speed and
4739 correctness. All must be specifically enabled.
4743 @opindex ffloat-store
4744 Do not store floating point variables in registers, and inhibit other
4745 options that might change whether a floating point value is taken from a
4748 @cindex floating point precision
4749 This option prevents undesirable excess precision on machines such as
4750 the 68000 where the floating registers (of the 68881) keep more
4751 precision than a @code{double} is supposed to have. Similarly for the
4752 x86 architecture. For most programs, the excess precision does only
4753 good, but a few programs rely on the precise definition of IEEE floating
4754 point. Use @option{-ffloat-store} for such programs, after modifying
4755 them to store all pertinent intermediate computations into variables.
4759 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
4760 @option{-fno-trapping-math}, @option{-ffinite-math-only},
4761 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
4763 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
4765 This option should never be turned on by any @option{-O} option since
4766 it can result in incorrect output for programs which depend on
4767 an exact implementation of IEEE or ISO rules/specifications for
4770 @item -fno-math-errno
4771 @opindex fno-math-errno
4772 Do not set ERRNO after calling math functions that are executed
4773 with a single instruction, e.g., sqrt. A program that relies on
4774 IEEE exceptions for math error handling may want to use this flag
4775 for speed while maintaining IEEE arithmetic compatibility.
4777 This option should never be turned on by any @option{-O} option since
4778 it can result in incorrect output for programs which depend on
4779 an exact implementation of IEEE or ISO rules/specifications for
4782 The default is @option{-fmath-errno}.
4784 @item -funsafe-math-optimizations
4785 @opindex funsafe-math-optimizations
4786 Allow optimizations for floating-point arithmetic that (a) assume
4787 that arguments and results are valid and (b) may violate IEEE or
4788 ANSI standards. When used at link-time, it may include libraries
4789 or startup files that change the default FPU control word or other
4790 similar optimizations.
4792 This option should never be turned on by any @option{-O} option since
4793 it can result in incorrect output for programs which depend on
4794 an exact implementation of IEEE or ISO rules/specifications for
4797 The default is @option{-fno-unsafe-math-optimizations}.
4799 @item -ffinite-math-only
4800 @opindex ffinite-math-only
4801 Allow optimizations for floating-point arithmetic that assume
4802 that arguments and results are not NaNs or +-Infs.
4804 This option should never be turned on by any @option{-O} option since
4805 it can result in incorrect output for programs which depend on
4806 an exact implementation of IEEE or ISO rules/specifications.
4808 The default is @option{-fno-finite-math-only}.
4810 @item -fno-trapping-math
4811 @opindex fno-trapping-math
4812 Compile code assuming that floating-point operations cannot generate
4813 user-visible traps. These traps include division by zero, overflow,
4814 underflow, inexact result and invalid operation. This option implies
4815 @option{-fno-signaling-nans}. Setting this option may allow faster
4816 code if one relies on ``non-stop'' IEEE arithmetic, for example.
4818 This option should never be turned on by any @option{-O} option since
4819 it can result in incorrect output for programs which depend on
4820 an exact implementation of IEEE or ISO rules/specifications for
4823 The default is @option{-ftrapping-math}.
4825 @item -frounding-math
4826 @opindex frounding-math
4827 Disable transformations and optimizations that assume default floating
4828 point rounding behavior. This is round-to-zero for all floating point
4829 to integer conversions, and round-to-nearest for all other arithmetic
4830 truncations. This option should be specified for programs that change
4831 the FP rounding mode dynamically, or that may be executed with a
4832 non-default rounding mode. This option disables constant folding of
4833 floating point expressions at compile-time (which may be affected by
4834 rounding mode) and arithmetic transformations that are unsafe in the
4835 presence of sign-dependent rounding modes.
4837 The default is @option{-fno-rounding-math}.
4839 This option is experimental and does not currently guarantee to
4840 disable all GCC optimizations that are affected by rounding mode.
4841 Future versions of GCC may provide finer control of this setting
4842 using C99's @code{FENV_ACCESS} pragma. This command line option
4843 will be used to specify the default state for @code{FENV_ACCESS}.
4845 @item -fsignaling-nans
4846 @opindex fsignaling-nans
4847 Compile code assuming that IEEE signaling NaNs may generate user-visible
4848 traps during floating-point operations. Setting this option disables
4849 optimizations that may change the number of exceptions visible with
4850 signaling NaNs. This option implies @option{-ftrapping-math}.
4852 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
4855 The default is @option{-fno-signaling-nans}.
4857 This option is experimental and does not currently guarantee to
4858 disable all GCC optimizations that affect signaling NaN behavior.
4860 @item -fsingle-precision-constant
4861 @opindex fsingle-precision-constant
4862 Treat floating point constant as single precision constant instead of
4863 implicitly converting it to double precision constant.
4868 The following options control optimizations that may improve
4869 performance, but are not enabled by any @option{-O} options. This
4870 section includes experimental options that may produce broken code.
4873 @item -fbranch-probabilities
4874 @opindex fbranch-probabilities
4875 After running a program compiled with @option{-fprofile-arcs}
4876 (@pxref{Debugging Options,, Options for Debugging Your Program or
4877 @command{gcc}}), you can compile it a second time using
4878 @option{-fbranch-probabilities}, to improve optimizations based on
4879 the number of times each branch was taken. When the program
4880 compiled with @option{-fprofile-arcs} exits it saves arc execution
4881 counts to a file called @file{@var{sourcename}.gcda} for each source
4882 file The information in this data file is very dependent on the
4883 structure of the generated code, so you must use the same source code
4884 and the same optimization options for both compilations.
4886 With @option{-fbranch-probabilities}, GCC puts a
4887 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
4888 These can be used to improve optimization. Currently, they are only
4889 used in one place: in @file{reorg.c}, instead of guessing which path a
4890 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
4891 exactly determine which path is taken more often.
4893 @item -fprofile-values
4894 @opindex fprofile-values
4895 If combined with @option{-fprofile-arcs}, it adds code so that some
4896 data about values of expressions in the program is gathered.
4898 With @option{-fbranch-probabilities}, it reads back the data gathered
4899 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
4900 notes to instructions for their later usage in optimizations.
4902 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4906 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
4907 a code to gather information about values of expressions.
4909 With @option{-fbranch-probabilities}, it reads back the data gathered
4910 and actually performs the optimizations based on them.
4911 Currently the optimizations include specialization of division operation
4912 using the knowledge about the value of the denominator.
4914 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
4916 @item -frename-registers
4917 @opindex frename-registers
4918 Attempt to avoid false dependencies in scheduled code by making use
4919 of registers left over after register allocation. This optimization
4920 will most benefit processors with lots of registers. Depending on the
4921 debug information format adopted by the target, however, it can
4922 make debugging impossible, since variables will no longer stay in
4923 a ``home register''.
4925 Not enabled by default at any level because it has known bugs.
4929 Use a graph coloring register allocator. Currently this option is meant
4930 for testing, so we are interested to hear about miscompilations with
4935 Perform tail duplication to enlarge superblock size. This transformation
4936 simplifies the control flow of the function allowing other optimizations to do
4939 Enabled with @option{-fprofile-use}.
4941 @item -funroll-loops
4942 @opindex funroll-loops
4943 Unroll loops whose number of iterations can be determined at compile time or
4944 upon entry to the loop. @option{-funroll-loops} implies
4945 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
4946 (i.e. complete removal of loops with small constant number of iterations).
4947 This option makes code larger, and may or may not make it run faster.
4949 Enabled with @option{-fprofile-use}.
4951 @item -funroll-all-loops
4952 @opindex funroll-all-loops
4953 Unroll all loops, even if their number of iterations is uncertain when
4954 the loop is entered. This usually makes programs run more slowly.
4955 @option{-funroll-all-loops} implies the same options as
4956 @option{-funroll-loops}.
4959 @opindex fpeel-loops
4960 Peels the loops for that there is enough information that they do not
4961 roll much (from profile feedback). It also turns on complete loop peeling
4962 (i.e. complete removal of loops with small constant number of iterations).
4964 Enabled with @option{-fprofile-use}.
4966 @item -fmove-loop-invariants
4967 @opindex fmove-loop-invariants
4968 Enables the loop invariant motion pass in the new loop optimizer. Enabled
4969 at level @option{-O1}
4971 @item -funswitch-loops
4972 @opindex funswitch-loops
4973 Move branches with loop invariant conditions out of the loop, with duplicates
4974 of the loop on both branches (modified according to result of the condition).
4976 @item -fold-unroll-loops
4977 @opindex fold-unroll-loops
4978 Unroll loops whose number of iterations can be determined at compile
4979 time or upon entry to the loop, using the old loop unroller whose loop
4980 recognition is based on notes from frontend. @option{-fold-unroll-loops} implies
4981 both @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4982 option makes code larger, and may or may not make it run faster.
4984 @item -fold-unroll-all-loops
4985 @opindex fold-unroll-all-loops
4986 Unroll all loops, even if their number of iterations is uncertain when
4987 the loop is entered. This is done using the old loop unroller whose loop
4988 recognition is based on notes from frontend. This usually makes programs run more slowly.
4989 @option{-fold-unroll-all-loops} implies the same options as
4990 @option{-fold-unroll-loops}.
4992 @item -fprefetch-loop-arrays
4993 @opindex fprefetch-loop-arrays
4994 If supported by the target machine, generate instructions to prefetch
4995 memory to improve the performance of loops that access large arrays.
4997 Disabled at level @option{-Os}.
4999 @item -ffunction-sections
5000 @itemx -fdata-sections
5001 @opindex ffunction-sections
5002 @opindex fdata-sections
5003 Place each function or data item into its own section in the output
5004 file if the target supports arbitrary sections. The name of the
5005 function or the name of the data item determines the section's name
5008 Use these options on systems where the linker can perform optimizations
5009 to improve locality of reference in the instruction space. Most systems
5010 using the ELF object format and SPARC processors running Solaris 2 have
5011 linkers with such optimizations. AIX may have these optimizations in
5014 Only use these options when there are significant benefits from doing
5015 so. When you specify these options, the assembler and linker will
5016 create larger object and executable files and will also be slower.
5017 You will not be able to use @code{gprof} on all systems if you
5018 specify this option and you may have problems with debugging if
5019 you specify both this option and @option{-g}.
5021 @item -fbranch-target-load-optimize
5022 @opindex fbranch-target-load-optimize
5023 Perform branch target register load optimization before prologue / epilogue
5025 The use of target registers can typically be exposed only during reload,
5026 thus hoisting loads out of loops and doing inter-block scheduling needs
5027 a separate optimization pass.
5029 @item -fbranch-target-load-optimize2
5030 @opindex fbranch-target-load-optimize2
5031 Perform branch target register load optimization after prologue / epilogue
5034 @item -fbtr-bb-exclusive
5035 @opindex fbtr-bb-exclusive
5036 When performing branch target register load optimization, don't reuse
5037 branch target registers in within any basic block.
5039 @item --param @var{name}=@var{value}
5041 In some places, GCC uses various constants to control the amount of
5042 optimization that is done. For example, GCC will not inline functions
5043 that contain more that a certain number of instructions. You can
5044 control some of these constants on the command-line using the
5045 @option{--param} option.
5047 The names of specific parameters, and the meaning of the values, are
5048 tied to the internals of the compiler, and are subject to change
5049 without notice in future releases.
5051 In each case, the @var{value} is an integer. The allowable choices for
5052 @var{name} are given in the following table:
5055 @item max-crossjump-edges
5056 The maximum number of incoming edges to consider for crossjumping.
5057 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5058 the number of edges incoming to each block. Increasing values mean
5059 more aggressive optimization, making the compile time increase with
5060 probably small improvement in executable size.
5062 @item max-delay-slot-insn-search
5063 The maximum number of instructions to consider when looking for an
5064 instruction to fill a delay slot. If more than this arbitrary number of
5065 instructions is searched, the time savings from filling the delay slot
5066 will be minimal so stop searching. Increasing values mean more
5067 aggressive optimization, making the compile time increase with probably
5068 small improvement in executable run time.
5070 @item max-delay-slot-live-search
5071 When trying to fill delay slots, the maximum number of instructions to
5072 consider when searching for a block with valid live register
5073 information. Increasing this arbitrarily chosen value means more
5074 aggressive optimization, increasing the compile time. This parameter
5075 should be removed when the delay slot code is rewritten to maintain the
5078 @item max-gcse-memory
5079 The approximate maximum amount of memory that will be allocated in
5080 order to perform the global common subexpression elimination
5081 optimization. If more memory than specified is required, the
5082 optimization will not be done.
5084 @item max-gcse-passes
5085 The maximum number of passes of GCSE to run. The default is 1.
5087 @item max-pending-list-length
5088 The maximum number of pending dependencies scheduling will allow
5089 before flushing the current state and starting over. Large functions
5090 with few branches or calls can create excessively large lists which
5091 needlessly consume memory and resources.
5093 @item max-inline-insns-single
5094 Several parameters control the tree inliner used in gcc.
5095 This number sets the maximum number of instructions (counted in GCC's
5096 internal representation) in a single function that the tree inliner
5097 will consider for inlining. This only affects functions declared
5098 inline and methods implemented in a class declaration (C++).
5099 The default value is 500.
5101 @item max-inline-insns-auto
5102 When you use @option{-finline-functions} (included in @option{-O3}),
5103 a lot of functions that would otherwise not be considered for inlining
5104 by the compiler will be investigated. To those functions, a different
5105 (more restrictive) limit compared to functions declared inline can
5107 The default value is 120.
5109 @item large-function-insns
5110 The limit specifying really large functions. For functions greater than this
5111 limit inlining is constrained by @option{--param large-function-growth}.
5112 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
5113 algorithms used by the backend.
5114 This parameter is ignored when @option{-funit-at-a-time} is not used.
5115 The default value is 3000.
5117 @item large-function-growth
5118 Specifies maximal growth of large function caused by inlining in percents.
5119 This parameter is ignored when @option{-funit-at-a-time} is not used.
5120 The default value is 200.
5122 @item inline-unit-growth
5123 Specifies maximal overall growth of the compilation unit caused by inlining.
5124 This parameter is ignored when @option{-funit-at-a-time} is not used.
5125 The default value is 150.
5127 @item max-inline-insns-recursive
5128 @itemx max-inline-insns-recursive-auto
5129 Specifies maximum number of instructions out-of-line copy of self recursive inline
5130 function can grow into by performing recursive inlining.
5132 For functions declared inline @option{--param max-inline-insns-recursive} is
5133 taken into acount. For function not declared inline, recursive inlining
5134 happens only when @option{-finline-functions} (included in @option{-O3}) is
5135 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5136 default value is 500.
5138 @item max-inline-recursive-depth
5139 @itemx max-inline-recursive-depth-auto
5140 Specifies maximum recursion depth used by the recursive inlining.
5142 For functions declared inline @option{--param max-inline-recursive-depth} is
5143 taken into acount. For function not declared inline, recursive inlining
5144 happens only when @option{-finline-functions} (included in @option{-O3}) is
5145 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5146 default value is 500.
5148 @item max-inline-insns-rtl
5149 For languages that use the RTL inliner (this happens at a later stage
5150 than tree inlining), you can set the maximum allowable size (counted
5151 in RTL instructions) for the RTL inliner with this parameter.
5152 The default value is 600.
5154 @item max-unrolled-insns
5155 The maximum number of instructions that a loop should have if that loop
5156 is unrolled, and if the loop is unrolled, it determines how many times
5157 the loop code is unrolled.
5159 @item max-average-unrolled-insns
5160 The maximum number of instructions biased by probabilities of their execution
5161 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5162 it determines how many times the loop code is unrolled.
5164 @item max-unroll-times
5165 The maximum number of unrollings of a single loop.
5167 @item max-peeled-insns
5168 The maximum number of instructions that a loop should have if that loop
5169 is peeled, and if the loop is peeled, it determines how many times
5170 the loop code is peeled.
5172 @item max-peel-times
5173 The maximum number of peelings of a single loop.
5175 @item max-completely-peeled-insns
5176 The maximum number of insns of a completely peeled loop.
5178 @item max-completely-peel-times
5179 The maximum number of iterations of a loop to be suitable for complete peeling.
5181 @item max-unswitch-insns
5182 The maximum number of insns of an unswitched loop.
5184 @item max-unswitch-level
5185 The maximum number of branches unswitched in a single loop.
5188 The minimum cost of an expensive expression in the loop invariant motion.
5190 @item max-iterations-to-track
5192 The maximum number of iterations of a loop the brute force algorithm
5193 for analysis of # of iterations of the loop tries to evaluate.
5195 @item hot-bb-count-fraction
5196 Select fraction of the maximal count of repetitions of basic block in program
5197 given basic block needs to have to be considered hot.
5199 @item hot-bb-frequency-fraction
5200 Select fraction of the maximal frequency of executions of basic block in
5201 function given basic block needs to have to be considered hot
5203 @item tracer-dynamic-coverage
5204 @itemx tracer-dynamic-coverage-feedback
5206 This value is used to limit superblock formation once the given percentage of
5207 executed instructions is covered. This limits unnecessary code size
5210 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5211 feedback is available. The real profiles (as opposed to statically estimated
5212 ones) are much less balanced allowing the threshold to be larger value.
5214 @item tracer-max-code-growth
5215 Stop tail duplication once code growth has reached given percentage. This is
5216 rather hokey argument, as most of the duplicates will be eliminated later in
5217 cross jumping, so it may be set to much higher values than is the desired code
5220 @item tracer-min-branch-ratio
5222 Stop reverse growth when the reverse probability of best edge is less than this
5223 threshold (in percent).
5225 @item tracer-min-branch-ratio
5226 @itemx tracer-min-branch-ratio-feedback
5228 Stop forward growth if the best edge do have probability lower than this
5231 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5232 compilation for profile feedback and one for compilation without. The value
5233 for compilation with profile feedback needs to be more conservative (higher) in
5234 order to make tracer effective.
5236 @item max-cse-path-length
5238 Maximum number of basic blocks on path that cse considers. The default is 10.
5240 @item global-var-threshold
5242 Counts the number of function calls (N) and the number of
5243 call-clobbered variables (V). If NxV is larger than this limit, a
5244 single artificial variable will be created to represent all the
5245 call-clobbered variables at function call sites. This artificial
5246 variable will then be made to alias every call-clobbered variable.
5247 (done as int * size_t on the host machine; beware overflow).
5249 @item max-aliased-vops
5251 Maxiumum number of virtual operands allowed to represent aliases
5252 before triggering the alias grouping heuristic. Alias grouping
5253 reduces compile times and memory consumption needed for aliasing at
5254 the expense of precision loss in alias information.
5256 @item ggc-min-expand
5258 GCC uses a garbage collector to manage its own memory allocation. This
5259 parameter specifies the minimum percentage by which the garbage
5260 collector's heap should be allowed to expand between collections.
5261 Tuning this may improve compilation speed; it has no effect on code
5264 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5265 RAM >= 1GB. If @code{getrlimit} is available, the notion of "RAM" is
5266 the smallest of actual RAM, RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If
5267 GCC is not able to calculate RAM on a particular platform, the lower
5268 bound of 30% is used. Setting this parameter and
5269 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5270 every opportunity. This is extremely slow, but can be useful for
5273 @item ggc-min-heapsize
5275 Minimum size of the garbage collector's heap before it begins bothering
5276 to collect garbage. The first collection occurs after the heap expands
5277 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5278 tuning this may improve compilation speed, and has no effect on code
5281 The default is RAM/8, with a lower bound of 4096 (four megabytes) and an
5282 upper bound of 131072 (128 megabytes). If @code{getrlimit} is
5283 available, the notion of "RAM" is the smallest of actual RAM,
5284 RLIMIT_RSS, RLIMIT_DATA and RLIMIT_AS. If GCC is not able to calculate
5285 RAM on a particular platform, the lower bound is used. Setting this
5286 parameter very large effectively disables garbage collection. Setting
5287 this parameter and @option{ggc-min-expand} to zero causes a full
5288 collection to occur at every opportunity.
5290 @item max-reload-search-insns
5291 The maximum number of instruction reload should look backward for equivalent
5292 register. Increasing values mean more aggressive optimization, making the
5293 compile time increase with probably slightly better performance. The default
5296 @item max-cselib-memory-location
5297 The maximum number of memory locations cselib should take into acount.
5298 Increasing values mean more aggressive optimization, making the compile time
5299 increase with probably slightly better performance. The default value is 500.
5301 @item reorder-blocks-duplicate
5302 @itemx reorder-blocks-duplicate-feedback
5304 Used by basic block reordering pass to decide whether to use unconditional
5305 branch or duplicate the code on its destination. Code is duplicated when its
5306 estimated size is smaller than this value multiplied by the estimated size of
5307 unconditional jump in the hot spots of the program.
5309 The @option{reorder-block-duplicate-feedback} is used only when profile
5310 feedback is available and may be set to higher values than
5311 @option{reorder-block-duplicate} since information about the hot spots is more
5314 @item max-sched-region-blocks
5315 The maximum number of blocks in a region to be considered for
5316 interblock scheduling. The default value is 10.
5318 @item max-sched-region-insns
5319 The maximum number of insns in a region to be considered for
5320 interblock scheduling. The default value is 100.
5324 @node Preprocessor Options
5325 @section Options Controlling the Preprocessor
5326 @cindex preprocessor options
5327 @cindex options, preprocessor
5329 These options control the C preprocessor, which is run on each C source
5330 file before actual compilation.
5332 If you use the @option{-E} option, nothing is done except preprocessing.
5333 Some of these options make sense only together with @option{-E} because
5334 they cause the preprocessor output to be unsuitable for actual
5339 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5340 and pass @var{option} directly through to the preprocessor. If
5341 @var{option} contains commas, it is split into multiple options at the
5342 commas. However, many options are modified, translated or interpreted
5343 by the compiler driver before being passed to the preprocessor, and
5344 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5345 interface is undocumented and subject to change, so whenever possible
5346 you should avoid using @option{-Wp} and let the driver handle the
5349 @item -Xpreprocessor @var{option}
5350 @opindex preprocessor
5351 Pass @var{option} as an option to the preprocessor. You can use this to
5352 supply system-specific preprocessor options which GCC does not know how to
5355 If you want to pass an option that takes an argument, you must use
5356 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5359 @include cppopts.texi
5361 @node Assembler Options
5362 @section Passing Options to the Assembler
5364 @c prevent bad page break with this line
5365 You can pass options to the assembler.
5368 @item -Wa,@var{option}
5370 Pass @var{option} as an option to the assembler. If @var{option}
5371 contains commas, it is split into multiple options at the commas.
5373 @item -Xassembler @var{option}
5375 Pass @var{option} as an option to the assembler. You can use this to
5376 supply system-specific assembler options which GCC does not know how to
5379 If you want to pass an option that takes an argument, you must use
5380 @option{-Xassembler} twice, once for the option and once for the argument.
5385 @section Options for Linking
5386 @cindex link options
5387 @cindex options, linking
5389 These options come into play when the compiler links object files into
5390 an executable output file. They are meaningless if the compiler is
5391 not doing a link step.
5395 @item @var{object-file-name}
5396 A file name that does not end in a special recognized suffix is
5397 considered to name an object file or library. (Object files are
5398 distinguished from libraries by the linker according to the file
5399 contents.) If linking is done, these object files are used as input
5408 If any of these options is used, then the linker is not run, and
5409 object file names should not be used as arguments. @xref{Overall
5413 @item -l@var{library}
5414 @itemx -l @var{library}
5416 Search the library named @var{library} when linking. (The second
5417 alternative with the library as a separate argument is only for
5418 POSIX compliance and is not recommended.)
5420 It makes a difference where in the command you write this option; the
5421 linker searches and processes libraries and object files in the order they
5422 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5423 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5424 to functions in @samp{z}, those functions may not be loaded.
5426 The linker searches a standard list of directories for the library,
5427 which is actually a file named @file{lib@var{library}.a}. The linker
5428 then uses this file as if it had been specified precisely by name.
5430 The directories searched include several standard system directories
5431 plus any that you specify with @option{-L}.
5433 Normally the files found this way are library files---archive files
5434 whose members are object files. The linker handles an archive file by
5435 scanning through it for members which define symbols that have so far
5436 been referenced but not defined. But if the file that is found is an
5437 ordinary object file, it is linked in the usual fashion. The only
5438 difference between using an @option{-l} option and specifying a file name
5439 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5440 and searches several directories.
5444 You need this special case of the @option{-l} option in order to
5445 link an Objective-C program.
5448 @opindex nostartfiles
5449 Do not use the standard system startup files when linking.
5450 The standard system libraries are used normally, unless @option{-nostdlib}
5451 or @option{-nodefaultlibs} is used.
5453 @item -nodefaultlibs
5454 @opindex nodefaultlibs
5455 Do not use the standard system libraries when linking.
5456 Only the libraries you specify will be passed to the linker.
5457 The standard startup files are used normally, unless @option{-nostartfiles}
5458 is used. The compiler may generate calls to @code{memcmp},
5459 @code{memset}, @code{memcpy} and @code{memmove}.
5460 These entries are usually resolved by entries in
5461 libc. These entry points should be supplied through some other
5462 mechanism when this option is specified.
5466 Do not use the standard system startup files or libraries when linking.
5467 No startup files and only the libraries you specify will be passed to
5468 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5469 @code{memcpy} and @code{memmove}.
5470 These entries are usually resolved by entries in
5471 libc. These entry points should be supplied through some other
5472 mechanism when this option is specified.
5474 @cindex @option{-lgcc}, use with @option{-nostdlib}
5475 @cindex @option{-nostdlib} and unresolved references
5476 @cindex unresolved references and @option{-nostdlib}
5477 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5478 @cindex @option{-nodefaultlibs} and unresolved references
5479 @cindex unresolved references and @option{-nodefaultlibs}
5480 One of the standard libraries bypassed by @option{-nostdlib} and
5481 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5482 that GCC uses to overcome shortcomings of particular machines, or special
5483 needs for some languages.
5484 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5485 Collection (GCC) Internals},
5486 for more discussion of @file{libgcc.a}.)
5487 In most cases, you need @file{libgcc.a} even when you want to avoid
5488 other standard libraries. In other words, when you specify @option{-nostdlib}
5489 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5490 This ensures that you have no unresolved references to internal GCC
5491 library subroutines. (For example, @samp{__main}, used to ensure C++
5492 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5493 GNU Compiler Collection (GCC) Internals}.)
5497 Produce a position independent executable on targets which support it.
5498 For predictable results, you must also specify the same set of options
5499 that were used to generate code (@option{-fpie}, @option{-fPIE},
5500 or model suboptions) when you specify this option.
5504 Remove all symbol table and relocation information from the executable.
5508 On systems that support dynamic linking, this prevents linking with the shared
5509 libraries. On other systems, this option has no effect.
5513 Produce a shared object which can then be linked with other objects to
5514 form an executable. Not all systems support this option. For predictable
5515 results, you must also specify the same set of options that were used to
5516 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5517 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5518 needs to build supplementary stub code for constructors to work. On
5519 multi-libbed systems, @samp{gcc -shared} must select the correct support
5520 libraries to link against. Failing to supply the correct flags may lead
5521 to subtle defects. Supplying them in cases where they are not necessary
5524 @item -shared-libgcc
5525 @itemx -static-libgcc
5526 @opindex shared-libgcc
5527 @opindex static-libgcc
5528 On systems that provide @file{libgcc} as a shared library, these options
5529 force the use of either the shared or static version respectively.
5530 If no shared version of @file{libgcc} was built when the compiler was
5531 configured, these options have no effect.
5533 There are several situations in which an application should use the
5534 shared @file{libgcc} instead of the static version. The most common
5535 of these is when the application wishes to throw and catch exceptions
5536 across different shared libraries. In that case, each of the libraries
5537 as well as the application itself should use the shared @file{libgcc}.
5539 Therefore, the G++ and GCJ drivers automatically add
5540 @option{-shared-libgcc} whenever you build a shared library or a main
5541 executable, because C++ and Java programs typically use exceptions, so
5542 this is the right thing to do.
5544 If, instead, you use the GCC driver to create shared libraries, you may
5545 find that they will not always be linked with the shared @file{libgcc}.
5546 If GCC finds, at its configuration time, that you have a non-GNU linker
5547 or a GNU linker that does not support option @option{--eh-frame-hdr},
5548 it will link the shared version of @file{libgcc} into shared libraries
5549 by default. Otherwise, it will take advantage of the linker and optimize
5550 away the linking with the shared version of @file{libgcc}, linking with
5551 the static version of libgcc by default. This allows exceptions to
5552 propagate through such shared libraries, without incurring relocation
5553 costs at library load time.
5555 However, if a library or main executable is supposed to throw or catch
5556 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5557 for the languages used in the program, or using the option
5558 @option{-shared-libgcc}, such that it is linked with the shared
5563 Bind references to global symbols when building a shared object. Warn
5564 about any unresolved references (unless overridden by the link editor
5565 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5568 @item -Xlinker @var{option}
5570 Pass @var{option} as an option to the linker. You can use this to
5571 supply system-specific linker options which GCC does not know how to
5574 If you want to pass an option that takes an argument, you must use
5575 @option{-Xlinker} twice, once for the option and once for the argument.
5576 For example, to pass @option{-assert definitions}, you must write
5577 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5578 @option{-Xlinker "-assert definitions"}, because this passes the entire
5579 string as a single argument, which is not what the linker expects.
5581 @item -Wl,@var{option}
5583 Pass @var{option} as an option to the linker. If @var{option} contains
5584 commas, it is split into multiple options at the commas.
5586 @item -u @var{symbol}
5588 Pretend the symbol @var{symbol} is undefined, to force linking of
5589 library modules to define it. You can use @option{-u} multiple times with
5590 different symbols to force loading of additional library modules.
5593 @node Directory Options
5594 @section Options for Directory Search
5595 @cindex directory options
5596 @cindex options, directory search
5599 These options specify directories to search for header files, for
5600 libraries and for parts of the compiler:
5605 Add the directory @var{dir} to the head of the list of directories to be
5606 searched for header files. This can be used to override a system header
5607 file, substituting your own version, since these directories are
5608 searched before the system header file directories. However, you should
5609 not use this option to add directories that contain vendor-supplied
5610 system header files (use @option{-isystem} for that). If you use more than
5611 one @option{-I} option, the directories are scanned in left-to-right
5612 order; the standard system directories come after.
5614 If a standard system include directory, or a directory specified with
5615 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5616 option will be ignored. The directory will still be searched but as a
5617 system directory at its normal position in the system include chain.
5618 This is to ensure that GCC's procedure to fix buggy system headers and
5619 the ordering for the include_next directive are not inadvertently changed.
5620 If you really need to change the search order for system directories,
5621 use the @option{-nostdinc} and/or @option{-isystem} options.
5623 @item -iquote@var{dir}
5625 Add the directory @var{dir} to the head of the list of directories to
5626 be searched for header files only for the case of @samp{#include
5627 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
5628 otherwise just like @option{-I}.
5632 Add directory @var{dir} to the list of directories to be searched
5635 @item -B@var{prefix}
5637 This option specifies where to find the executables, libraries,
5638 include files, and data files of the compiler itself.
5640 The compiler driver program runs one or more of the subprograms
5641 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5642 @var{prefix} as a prefix for each program it tries to run, both with and
5643 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5645 For each subprogram to be run, the compiler driver first tries the
5646 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5647 was not specified, the driver tries two standard prefixes, which are
5648 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5649 those results in a file name that is found, the unmodified program
5650 name is searched for using the directories specified in your
5651 @env{PATH} environment variable.
5653 The compiler will check to see if the path provided by the @option{-B}
5654 refers to a directory, and if necessary it will add a directory
5655 separator character at the end of the path.
5657 @option{-B} prefixes that effectively specify directory names also apply
5658 to libraries in the linker, because the compiler translates these
5659 options into @option{-L} options for the linker. They also apply to
5660 includes files in the preprocessor, because the compiler translates these
5661 options into @option{-isystem} options for the preprocessor. In this case,
5662 the compiler appends @samp{include} to the prefix.
5664 The run-time support file @file{libgcc.a} can also be searched for using
5665 the @option{-B} prefix, if needed. If it is not found there, the two
5666 standard prefixes above are tried, and that is all. The file is left
5667 out of the link if it is not found by those means.
5669 Another way to specify a prefix much like the @option{-B} prefix is to use
5670 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5673 As a special kludge, if the path provided by @option{-B} is
5674 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5675 9, then it will be replaced by @file{[dir/]include}. This is to help
5676 with boot-strapping the compiler.
5678 @item -specs=@var{file}
5680 Process @var{file} after the compiler reads in the standard @file{specs}
5681 file, in order to override the defaults that the @file{gcc} driver
5682 program uses when determining what switches to pass to @file{cc1},
5683 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5684 @option{-specs=@var{file}} can be specified on the command line, and they
5685 are processed in order, from left to right.
5689 This option has been deprecated. Please use @option{-iquote} instead for
5690 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
5691 Any directories you specify with @option{-I} options before the @option{-I-}
5692 option are searched only for the case of @samp{#include "@var{file}"};
5693 they are not searched for @samp{#include <@var{file}>}.
5695 If additional directories are specified with @option{-I} options after
5696 the @option{-I-}, these directories are searched for all @samp{#include}
5697 directives. (Ordinarily @emph{all} @option{-I} directories are used
5700 In addition, the @option{-I-} option inhibits the use of the current
5701 directory (where the current input file came from) as the first search
5702 directory for @samp{#include "@var{file}"}. There is no way to
5703 override this effect of @option{-I-}. With @option{-I.} you can specify
5704 searching the directory which was current when the compiler was
5705 invoked. That is not exactly the same as what the preprocessor does
5706 by default, but it is often satisfactory.
5708 @option{-I-} does not inhibit the use of the standard system directories
5709 for header files. Thus, @option{-I-} and @option{-nostdinc} are
5716 @section Specifying subprocesses and the switches to pass to them
5719 @command{gcc} is a driver program. It performs its job by invoking a
5720 sequence of other programs to do the work of compiling, assembling and
5721 linking. GCC interprets its command-line parameters and uses these to
5722 deduce which programs it should invoke, and which command-line options
5723 it ought to place on their command lines. This behavior is controlled
5724 by @dfn{spec strings}. In most cases there is one spec string for each
5725 program that GCC can invoke, but a few programs have multiple spec
5726 strings to control their behavior. The spec strings built into GCC can
5727 be overridden by using the @option{-specs=} command-line switch to specify
5730 @dfn{Spec files} are plaintext files that are used to construct spec
5731 strings. They consist of a sequence of directives separated by blank
5732 lines. The type of directive is determined by the first non-whitespace
5733 character on the line and it can be one of the following:
5736 @item %@var{command}
5737 Issues a @var{command} to the spec file processor. The commands that can
5741 @item %include <@var{file}>
5743 Search for @var{file} and insert its text at the current point in the
5746 @item %include_noerr <@var{file}>
5747 @cindex %include_noerr
5748 Just like @samp{%include}, but do not generate an error message if the include
5749 file cannot be found.
5751 @item %rename @var{old_name} @var{new_name}
5753 Rename the spec string @var{old_name} to @var{new_name}.
5757 @item *[@var{spec_name}]:
5758 This tells the compiler to create, override or delete the named spec
5759 string. All lines after this directive up to the next directive or
5760 blank line are considered to be the text for the spec string. If this
5761 results in an empty string then the spec will be deleted. (Or, if the
5762 spec did not exist, then nothing will happened.) Otherwise, if the spec
5763 does not currently exist a new spec will be created. If the spec does
5764 exist then its contents will be overridden by the text of this
5765 directive, unless the first character of that text is the @samp{+}
5766 character, in which case the text will be appended to the spec.
5768 @item [@var{suffix}]:
5769 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
5770 and up to the next directive or blank line are considered to make up the
5771 spec string for the indicated suffix. When the compiler encounters an
5772 input file with the named suffix, it will processes the spec string in
5773 order to work out how to compile that file. For example:
5780 This says that any input file whose name ends in @samp{.ZZ} should be
5781 passed to the program @samp{z-compile}, which should be invoked with the
5782 command-line switch @option{-input} and with the result of performing the
5783 @samp{%i} substitution. (See below.)
5785 As an alternative to providing a spec string, the text that follows a
5786 suffix directive can be one of the following:
5789 @item @@@var{language}
5790 This says that the suffix is an alias for a known @var{language}. This is
5791 similar to using the @option{-x} command-line switch to GCC to specify a
5792 language explicitly. For example:
5799 Says that .ZZ files are, in fact, C++ source files.
5802 This causes an error messages saying:
5805 @var{name} compiler not installed on this system.
5809 GCC already has an extensive list of suffixes built into it.
5810 This directive will add an entry to the end of the list of suffixes, but
5811 since the list is searched from the end backwards, it is effectively
5812 possible to override earlier entries using this technique.
5816 GCC has the following spec strings built into it. Spec files can
5817 override these strings or create their own. Note that individual
5818 targets can also add their own spec strings to this list.
5821 asm Options to pass to the assembler
5822 asm_final Options to pass to the assembler post-processor
5823 cpp Options to pass to the C preprocessor
5824 cc1 Options to pass to the C compiler
5825 cc1plus Options to pass to the C++ compiler
5826 endfile Object files to include at the end of the link
5827 link Options to pass to the linker
5828 lib Libraries to include on the command line to the linker
5829 libgcc Decides which GCC support library to pass to the linker
5830 linker Sets the name of the linker
5831 predefines Defines to be passed to the C preprocessor
5832 signed_char Defines to pass to CPP to say whether @code{char} is signed
5834 startfile Object files to include at the start of the link
5837 Here is a small example of a spec file:
5843 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
5846 This example renames the spec called @samp{lib} to @samp{old_lib} and
5847 then overrides the previous definition of @samp{lib} with a new one.
5848 The new definition adds in some extra command-line options before
5849 including the text of the old definition.
5851 @dfn{Spec strings} are a list of command-line options to be passed to their
5852 corresponding program. In addition, the spec strings can contain
5853 @samp{%}-prefixed sequences to substitute variable text or to
5854 conditionally insert text into the command line. Using these constructs
5855 it is possible to generate quite complex command lines.
5857 Here is a table of all defined @samp{%}-sequences for spec
5858 strings. Note that spaces are not generated automatically around the
5859 results of expanding these sequences. Therefore you can concatenate them
5860 together or combine them with constant text in a single argument.
5864 Substitute one @samp{%} into the program name or argument.
5867 Substitute the name of the input file being processed.
5870 Substitute the basename of the input file being processed.
5871 This is the substring up to (and not including) the last period
5872 and not including the directory.
5875 This is the same as @samp{%b}, but include the file suffix (text after
5879 Marks the argument containing or following the @samp{%d} as a
5880 temporary file name, so that that file will be deleted if GCC exits
5881 successfully. Unlike @samp{%g}, this contributes no text to the
5884 @item %g@var{suffix}
5885 Substitute a file name that has suffix @var{suffix} and is chosen
5886 once per compilation, and mark the argument in the same way as
5887 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
5888 name is now chosen in a way that is hard to predict even when previously
5889 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
5890 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
5891 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
5892 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
5893 was simply substituted with a file name chosen once per compilation,
5894 without regard to any appended suffix (which was therefore treated
5895 just like ordinary text), making such attacks more likely to succeed.
5897 @item %u@var{suffix}
5898 Like @samp{%g}, but generates a new temporary file name even if
5899 @samp{%u@var{suffix}} was already seen.
5901 @item %U@var{suffix}
5902 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
5903 new one if there is no such last file name. In the absence of any
5904 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
5905 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
5906 would involve the generation of two distinct file names, one
5907 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
5908 simply substituted with a file name chosen for the previous @samp{%u},
5909 without regard to any appended suffix.
5911 @item %j@var{suffix}
5912 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
5913 writable, and if save-temps is off; otherwise, substitute the name
5914 of a temporary file, just like @samp{%u}. This temporary file is not
5915 meant for communication between processes, but rather as a junk
5918 @item %|@var{suffix}
5919 @itemx %m@var{suffix}
5920 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
5921 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
5922 all. These are the two most common ways to instruct a program that it
5923 should read from standard input or write to standard output. If you
5924 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
5925 construct: see for example @file{f/lang-specs.h}.
5927 @item %.@var{SUFFIX}
5928 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
5929 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
5930 terminated by the next space or %.
5933 Marks the argument containing or following the @samp{%w} as the
5934 designated output file of this compilation. This puts the argument
5935 into the sequence of arguments that @samp{%o} will substitute later.
5938 Substitutes the names of all the output files, with spaces
5939 automatically placed around them. You should write spaces
5940 around the @samp{%o} as well or the results are undefined.
5941 @samp{%o} is for use in the specs for running the linker.
5942 Input files whose names have no recognized suffix are not compiled
5943 at all, but they are included among the output files, so they will
5947 Substitutes the suffix for object files. Note that this is
5948 handled specially when it immediately follows @samp{%g, %u, or %U},
5949 because of the need for those to form complete file names. The
5950 handling is such that @samp{%O} is treated exactly as if it had already
5951 been substituted, except that @samp{%g, %u, and %U} do not currently
5952 support additional @var{suffix} characters following @samp{%O} as they would
5953 following, for example, @samp{.o}.
5956 Substitutes the standard macro predefinitions for the
5957 current target machine. Use this when running @code{cpp}.
5960 Like @samp{%p}, but puts @samp{__} before and after the name of each
5961 predefined macro, except for macros that start with @samp{__} or with
5962 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
5966 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
5967 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
5968 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
5972 Current argument is the name of a library or startup file of some sort.
5973 Search for that file in a standard list of directories and substitute
5974 the full name found.
5977 Print @var{str} as an error message. @var{str} is terminated by a newline.
5978 Use this when inconsistent options are detected.
5981 Substitute the contents of spec string @var{name} at this point.
5984 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
5986 @item %x@{@var{option}@}
5987 Accumulate an option for @samp{%X}.
5990 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
5994 Output the accumulated assembler options specified by @option{-Wa}.
5997 Output the accumulated preprocessor options specified by @option{-Wp}.
6000 Process the @code{asm} spec. This is used to compute the
6001 switches to be passed to the assembler.
6004 Process the @code{asm_final} spec. This is a spec string for
6005 passing switches to an assembler post-processor, if such a program is
6009 Process the @code{link} spec. This is the spec for computing the
6010 command line passed to the linker. Typically it will make use of the
6011 @samp{%L %G %S %D and %E} sequences.
6014 Dump out a @option{-L} option for each directory that GCC believes might
6015 contain startup files. If the target supports multilibs then the
6016 current multilib directory will be prepended to each of these paths.
6019 Output the multilib directory with directory separators replaced with
6020 @samp{_}. If multilib directories are not set, or the multilib directory is
6021 @file{.} then this option emits nothing.
6024 Process the @code{lib} spec. This is a spec string for deciding which
6025 libraries should be included on the command line to the linker.
6028 Process the @code{libgcc} spec. This is a spec string for deciding
6029 which GCC support library should be included on the command line to the linker.
6032 Process the @code{startfile} spec. This is a spec for deciding which
6033 object files should be the first ones passed to the linker. Typically
6034 this might be a file named @file{crt0.o}.
6037 Process the @code{endfile} spec. This is a spec string that specifies
6038 the last object files that will be passed to the linker.
6041 Process the @code{cpp} spec. This is used to construct the arguments
6042 to be passed to the C preprocessor.
6045 Process the @code{cc1} spec. This is used to construct the options to be
6046 passed to the actual C compiler (@samp{cc1}).
6049 Process the @code{cc1plus} spec. This is used to construct the options to be
6050 passed to the actual C++ compiler (@samp{cc1plus}).
6053 Substitute the variable part of a matched option. See below.
6054 Note that each comma in the substituted string is replaced by
6058 Remove all occurrences of @code{-S} from the command line. Note---this
6059 command is position dependent. @samp{%} commands in the spec string
6060 before this one will see @code{-S}, @samp{%} commands in the spec string
6061 after this one will not.
6063 @item %:@var{function}(@var{args})
6064 Call the named function @var{function}, passing it @var{args}.
6065 @var{args} is first processed as a nested spec string, then split
6066 into an argument vector in the usual fashion. The function returns
6067 a string which is processed as if it had appeared literally as part
6068 of the current spec.
6070 The following built-in spec functions are provided:
6073 @item @code{if-exists}
6074 The @code{if-exists} spec function takes one argument, an absolute
6075 pathname to a file. If the file exists, @code{if-exists} returns the
6076 pathname. Here is a small example of its usage:
6080 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6083 @item @code{if-exists-else}
6084 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6085 spec function, except that it takes two arguments. The first argument is
6086 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6087 returns the pathname. If it does not exist, it returns the second argument.
6088 This way, @code{if-exists-else} can be used to select one file or another,
6089 based on the existence of the first. Here is a small example of its usage:
6093 crt0%O%s %:if-exists(crti%O%s) \
6094 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6099 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6100 If that switch was not specified, this substitutes nothing. Note that
6101 the leading dash is omitted when specifying this option, and it is
6102 automatically inserted if the substitution is performed. Thus the spec
6103 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6104 and would output the command line option @option{-foo}.
6106 @item %W@{@code{S}@}
6107 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6110 @item %@{@code{S}*@}
6111 Substitutes all the switches specified to GCC whose names start
6112 with @code{-S}, but which also take an argument. This is used for
6113 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6114 GCC considers @option{-o foo} as being
6115 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6116 text, including the space. Thus two arguments would be generated.
6118 @item %@{@code{S}*&@code{T}*@}
6119 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6120 (the order of @code{S} and @code{T} in the spec is not significant).
6121 There can be any number of ampersand-separated variables; for each the
6122 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6124 @item %@{@code{S}:@code{X}@}
6125 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6127 @item %@{!@code{S}:@code{X}@}
6128 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6130 @item %@{@code{S}*:@code{X}@}
6131 Substitutes @code{X} if one or more switches whose names start with
6132 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6133 once, no matter how many such switches appeared. However, if @code{%*}
6134 appears somewhere in @code{X}, then @code{X} will be substituted once
6135 for each matching switch, with the @code{%*} replaced by the part of
6136 that switch that matched the @code{*}.
6138 @item %@{.@code{S}:@code{X}@}
6139 Substitutes @code{X}, if processing a file with suffix @code{S}.
6141 @item %@{!.@code{S}:@code{X}@}
6142 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6144 @item %@{@code{S}|@code{P}:@code{X}@}
6145 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6146 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6147 although they have a stronger binding than the @samp{|}. If @code{%*}
6148 appears in @code{X}, all of the alternatives must be starred, and only
6149 the first matching alternative is substituted.
6151 For example, a spec string like this:
6154 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6157 will output the following command-line options from the following input
6158 command-line options:
6163 -d fred.c -foo -baz -boggle
6164 -d jim.d -bar -baz -boggle
6167 @item %@{S:X; T:Y; :D@}
6169 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6170 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6171 be as many clauses as you need. This may be combined with @code{.},
6172 @code{!}, @code{|}, and @code{*} as needed.
6177 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6178 construct may contain other nested @samp{%} constructs or spaces, or
6179 even newlines. They are processed as usual, as described above.
6180 Trailing white space in @code{X} is ignored. White space may also
6181 appear anywhere on the left side of the colon in these constructs,
6182 except between @code{.} or @code{*} and the corresponding word.
6184 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6185 handled specifically in these constructs. If another value of
6186 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6187 @option{-W} switch is found later in the command line, the earlier
6188 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6189 just one letter, which passes all matching options.
6191 The character @samp{|} at the beginning of the predicate text is used to
6192 indicate that a command should be piped to the following command, but
6193 only if @option{-pipe} is specified.
6195 It is built into GCC which switches take arguments and which do not.
6196 (You might think it would be useful to generalize this to allow each
6197 compiler's spec to say which switches take arguments. But this cannot
6198 be done in a consistent fashion. GCC cannot even decide which input
6199 files have been specified without knowing which switches take arguments,
6200 and it must know which input files to compile in order to tell which
6203 GCC also knows implicitly that arguments starting in @option{-l} are to be
6204 treated as compiler output files, and passed to the linker in their
6205 proper position among the other output files.
6207 @c man begin OPTIONS
6209 @node Target Options
6210 @section Specifying Target Machine and Compiler Version
6211 @cindex target options
6212 @cindex cross compiling
6213 @cindex specifying machine version
6214 @cindex specifying compiler version and target machine
6215 @cindex compiler version, specifying
6216 @cindex target machine, specifying
6218 The usual way to run GCC is to run the executable called @file{gcc}, or
6219 @file{<machine>-gcc} when cross-compiling, or
6220 @file{<machine>-gcc-<version>} to run a version other than the one that
6221 was installed last. Sometimes this is inconvenient, so GCC provides
6222 options that will switch to another cross-compiler or version.
6225 @item -b @var{machine}
6227 The argument @var{machine} specifies the target machine for compilation.
6229 The value to use for @var{machine} is the same as was specified as the
6230 machine type when configuring GCC as a cross-compiler. For
6231 example, if a cross-compiler was configured with @samp{configure
6232 i386v}, meaning to compile for an 80386 running System V, then you
6233 would specify @option{-b i386v} to run that cross compiler.
6235 @item -V @var{version}
6237 The argument @var{version} specifies which version of GCC to run.
6238 This is useful when multiple versions are installed. For example,
6239 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6242 The @option{-V} and @option{-b} options work by running the
6243 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6244 use them if you can just run that directly.
6246 @node Submodel Options
6247 @section Hardware Models and Configurations
6248 @cindex submodel options
6249 @cindex specifying hardware config
6250 @cindex hardware models and configurations, specifying
6251 @cindex machine dependent options
6253 Earlier we discussed the standard option @option{-b} which chooses among
6254 different installed compilers for completely different target
6255 machines, such as VAX vs.@: 68000 vs.@: 80386.
6257 In addition, each of these target machine types can have its own
6258 special options, starting with @samp{-m}, to choose among various
6259 hardware models or configurations---for example, 68010 vs 68020,
6260 floating coprocessor or none. A single installed version of the
6261 compiler can compile for any model or configuration, according to the
6264 Some configurations of the compiler also support additional special
6265 options, usually for compatibility with other compilers on the same
6268 These options are defined by the macro @code{TARGET_SWITCHES} in the
6269 machine description. The default for the options is also defined by
6270 that macro, which enables you to change the defaults.
6272 @c This list is ordered alphanumerically by subsection name.
6273 @c It should be the same order and spelling as these options are listed
6274 @c in Machine Dependent Options
6282 * DEC Alpha Options::
6283 * DEC Alpha/VMS Options::
6287 * i386 and x86-64 Options::
6299 * RS/6000 and PowerPC Options::
6300 * S/390 and zSeries Options::
6303 * System V Options::
6304 * TMS320C3x/C4x Options::
6308 * Xstormy16 Options::
6314 @subsection ARC Options
6317 These options are defined for ARC implementations:
6322 Compile code for little endian mode. This is the default.
6326 Compile code for big endian mode.
6329 @opindex mmangle-cpu
6330 Prepend the name of the cpu to all public symbol names.
6331 In multiple-processor systems, there are many ARC variants with different
6332 instruction and register set characteristics. This flag prevents code
6333 compiled for one cpu to be linked with code compiled for another.
6334 No facility exists for handling variants that are ``almost identical''.
6335 This is an all or nothing option.
6337 @item -mcpu=@var{cpu}
6339 Compile code for ARC variant @var{cpu}.
6340 Which variants are supported depend on the configuration.
6341 All variants support @option{-mcpu=base}, this is the default.
6343 @item -mtext=@var{text-section}
6344 @itemx -mdata=@var{data-section}
6345 @itemx -mrodata=@var{readonly-data-section}
6349 Put functions, data, and readonly data in @var{text-section},
6350 @var{data-section}, and @var{readonly-data-section} respectively
6351 by default. This can be overridden with the @code{section} attribute.
6352 @xref{Variable Attributes}.
6357 @subsection ARM Options
6360 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6364 @item -mabi=@var{name}
6366 Generate code for the specified ABI. Permissible values are: @samp{apcs-gnu},
6367 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6370 @opindex mapcs-frame
6371 Generate a stack frame that is compliant with the ARM Procedure Call
6372 Standard for all functions, even if this is not strictly necessary for
6373 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6374 with this option will cause the stack frames not to be generated for
6375 leaf functions. The default is @option{-mno-apcs-frame}.
6379 This is a synonym for @option{-mapcs-frame}.
6382 @c not currently implemented
6383 @item -mapcs-stack-check
6384 @opindex mapcs-stack-check
6385 Generate code to check the amount of stack space available upon entry to
6386 every function (that actually uses some stack space). If there is
6387 insufficient space available then either the function
6388 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6389 called, depending upon the amount of stack space required. The run time
6390 system is required to provide these functions. The default is
6391 @option{-mno-apcs-stack-check}, since this produces smaller code.
6393 @c not currently implemented
6395 @opindex mapcs-float
6396 Pass floating point arguments using the float point registers. This is
6397 one of the variants of the APCS@. This option is recommended if the
6398 target hardware has a floating point unit or if a lot of floating point
6399 arithmetic is going to be performed by the code. The default is
6400 @option{-mno-apcs-float}, since integer only code is slightly increased in
6401 size if @option{-mapcs-float} is used.
6403 @c not currently implemented
6404 @item -mapcs-reentrant
6405 @opindex mapcs-reentrant
6406 Generate reentrant, position independent code. The default is
6407 @option{-mno-apcs-reentrant}.
6410 @item -mthumb-interwork
6411 @opindex mthumb-interwork
6412 Generate code which supports calling between the ARM and Thumb
6413 instruction sets. Without this option the two instruction sets cannot
6414 be reliably used inside one program. The default is
6415 @option{-mno-thumb-interwork}, since slightly larger code is generated
6416 when @option{-mthumb-interwork} is specified.
6418 @item -mno-sched-prolog
6419 @opindex mno-sched-prolog
6420 Prevent the reordering of instructions in the function prolog, or the
6421 merging of those instruction with the instructions in the function's
6422 body. This means that all functions will start with a recognizable set
6423 of instructions (or in fact one of a choice from a small set of
6424 different function prologues), and this information can be used to
6425 locate the start if functions inside an executable piece of code. The
6426 default is @option{-msched-prolog}.
6429 @opindex mhard-float
6430 Generate output containing floating point instructions. This is the
6434 @opindex msoft-float
6435 Generate output containing library calls for floating point.
6436 @strong{Warning:} the requisite libraries are not available for all ARM
6437 targets. Normally the facilities of the machine's usual C compiler are
6438 used, but this cannot be done directly in cross-compilation. You must make
6439 your own arrangements to provide suitable library functions for
6442 @option{-msoft-float} changes the calling convention in the output file;
6443 therefore, it is only useful if you compile @emph{all} of a program with
6444 this option. In particular, you need to compile @file{libgcc.a}, the
6445 library that comes with GCC, with @option{-msoft-float} in order for
6448 @item -mfloat-abi=@var{name}
6450 Specifies which ABI to use for floating point values. Permissible values
6451 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6453 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6454 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6455 of floating point instructions, but still uses the soft-float calling
6458 @item -mlittle-endian
6459 @opindex mlittle-endian
6460 Generate code for a processor running in little-endian mode. This is
6461 the default for all standard configurations.
6464 @opindex mbig-endian
6465 Generate code for a processor running in big-endian mode; the default is
6466 to compile code for a little-endian processor.
6468 @item -mwords-little-endian
6469 @opindex mwords-little-endian
6470 This option only applies when generating code for big-endian processors.
6471 Generate code for a little-endian word order but a big-endian byte
6472 order. That is, a byte order of the form @samp{32107654}. Note: this
6473 option should only be used if you require compatibility with code for
6474 big-endian ARM processors generated by versions of the compiler prior to
6477 @item -mcpu=@var{name}
6479 This specifies the name of the target ARM processor. GCC uses this name
6480 to determine what kind of instructions it can emit when generating
6481 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6482 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6483 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6484 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6485 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6486 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm8},
6487 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6488 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6489 @samp{arm920t}, @samp{arm926ejs}, @samp{arm940t}, @samp{arm9tdmi},
6490 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ejs},
6491 @samp{arm1136js}, @samp{arm1136jfs} ,@samp{xscale}, @samp{iwmmxt},
6494 @itemx -mtune=@var{name}
6496 This option is very similar to the @option{-mcpu=} option, except that
6497 instead of specifying the actual target processor type, and hence
6498 restricting which instructions can be used, it specifies that GCC should
6499 tune the performance of the code as if the target were of the type
6500 specified in this option, but still choosing the instructions that it
6501 will generate based on the cpu specified by a @option{-mcpu=} option.
6502 For some ARM implementations better performance can be obtained by using
6505 @item -march=@var{name}
6507 This specifies the name of the target ARM architecture. GCC uses this
6508 name to determine what kind of instructions it can emit when generating
6509 assembly code. This option can be used in conjunction with or instead
6510 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6511 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6512 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6513 @samp{iwmmxt}, @samp{ep9312}.
6515 @item -mfpu=@var{name}
6516 @itemx -mfpe=@var{number}
6517 @itemx -mfp=@var{number}
6521 This specifies what floating point hardware (or hardware emulation) is
6522 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6523 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6524 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6525 with older versions of GCC@.
6527 If @option{-msoft-float} is specified this specifies the format of
6528 floating point values.
6530 @item -mstructure-size-boundary=@var{n}
6531 @opindex mstructure-size-boundary
6532 The size of all structures and unions will be rounded up to a multiple
6533 of the number of bits set by this option. Permissible values are 8, 32
6534 and 64. The default value varies for different toolchains. For the COFF
6535 targeted toolchain the default value is 8. A value of 64 is only allowed
6536 if the underlying ABI supports it.
6538 Specifying the larger number can produce faster, more efficient code, but
6539 can also increase the size of the program. Different values are potentially
6540 incompatible. Code compiled with one value cannot necessarily expect to
6541 work with code or libraries compiled with another value, if they exchange
6542 information using structures or unions.
6544 @item -mabort-on-noreturn
6545 @opindex mabort-on-noreturn
6546 Generate a call to the function @code{abort} at the end of a
6547 @code{noreturn} function. It will be executed if the function tries to
6551 @itemx -mno-long-calls
6552 @opindex mlong-calls
6553 @opindex mno-long-calls
6554 Tells the compiler to perform function calls by first loading the
6555 address of the function into a register and then performing a subroutine
6556 call on this register. This switch is needed if the target function
6557 will lie outside of the 64 megabyte addressing range of the offset based
6558 version of subroutine call instruction.
6560 Even if this switch is enabled, not all function calls will be turned
6561 into long calls. The heuristic is that static functions, functions
6562 which have the @samp{short-call} attribute, functions that are inside
6563 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6564 definitions have already been compiled within the current compilation
6565 unit, will not be turned into long calls. The exception to this rule is
6566 that weak function definitions, functions with the @samp{long-call}
6567 attribute or the @samp{section} attribute, and functions that are within
6568 the scope of a @samp{#pragma long_calls} directive, will always be
6569 turned into long calls.
6571 This feature is not enabled by default. Specifying
6572 @option{-mno-long-calls} will restore the default behavior, as will
6573 placing the function calls within the scope of a @samp{#pragma
6574 long_calls_off} directive. Note these switches have no effect on how
6575 the compiler generates code to handle function calls via function
6578 @item -mnop-fun-dllimport
6579 @opindex mnop-fun-dllimport
6580 Disable support for the @code{dllimport} attribute.
6582 @item -msingle-pic-base
6583 @opindex msingle-pic-base
6584 Treat the register used for PIC addressing as read-only, rather than
6585 loading it in the prologue for each function. The run-time system is
6586 responsible for initializing this register with an appropriate value
6587 before execution begins.
6589 @item -mpic-register=@var{reg}
6590 @opindex mpic-register
6591 Specify the register to be used for PIC addressing. The default is R10
6592 unless stack-checking is enabled, when R9 is used.
6594 @item -mcirrus-fix-invalid-insns
6595 @opindex mcirrus-fix-invalid-insns
6596 @opindex mno-cirrus-fix-invalid-insns
6597 Insert NOPs into the instruction stream to in order to work around
6598 problems with invalid Maverick instruction combinations. This option
6599 is only valid if the @option{-mcpu=ep9312} option has been used to
6600 enable generation of instructions for the Cirrus Maverick floating
6601 point co-processor. This option is not enabled by default, since the
6602 problem is only present in older Maverick implementations. The default
6603 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6606 @item -mpoke-function-name
6607 @opindex mpoke-function-name
6608 Write the name of each function into the text section, directly
6609 preceding the function prologue. The generated code is similar to this:
6613 .ascii "arm_poke_function_name", 0
6616 .word 0xff000000 + (t1 - t0)
6617 arm_poke_function_name
6619 stmfd sp!, @{fp, ip, lr, pc@}
6623 When performing a stack backtrace, code can inspect the value of
6624 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6625 location @code{pc - 12} and the top 8 bits are set, then we know that
6626 there is a function name embedded immediately preceding this location
6627 and has length @code{((pc[-3]) & 0xff000000)}.
6631 Generate code for the 16-bit Thumb instruction set. The default is to
6632 use the 32-bit ARM instruction set.
6635 @opindex mtpcs-frame
6636 Generate a stack frame that is compliant with the Thumb Procedure Call
6637 Standard for all non-leaf functions. (A leaf function is one that does
6638 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6640 @item -mtpcs-leaf-frame
6641 @opindex mtpcs-leaf-frame
6642 Generate a stack frame that is compliant with the Thumb Procedure Call
6643 Standard for all leaf functions. (A leaf function is one that does
6644 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6646 @item -mcallee-super-interworking
6647 @opindex mcallee-super-interworking
6648 Gives all externally visible functions in the file being compiled an ARM
6649 instruction set header which switches to Thumb mode before executing the
6650 rest of the function. This allows these functions to be called from
6651 non-interworking code.
6653 @item -mcaller-super-interworking
6654 @opindex mcaller-super-interworking
6655 Allows calls via function pointers (including virtual functions) to
6656 execute correctly regardless of whether the target code has been
6657 compiled for interworking or not. There is a small overhead in the cost
6658 of executing a function pointer if this option is enabled.
6663 @subsection AVR Options
6666 These options are defined for AVR implementations:
6669 @item -mmcu=@var{mcu}
6671 Specify ATMEL AVR instruction set or MCU type.
6673 Instruction set avr1 is for the minimal AVR core, not supported by the C
6674 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
6675 attiny11, attiny12, attiny15, attiny28).
6677 Instruction set avr2 (default) is for the classic AVR core with up to
6678 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
6679 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
6680 at90c8534, at90s8535).
6682 Instruction set avr3 is for the classic AVR core with up to 128K program
6683 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
6685 Instruction set avr4 is for the enhanced AVR core with up to 8K program
6686 memory space (MCU types: atmega8, atmega83, atmega85).
6688 Instruction set avr5 is for the enhanced AVR core with up to 128K program
6689 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
6690 atmega64, atmega128, at43usb355, at94k).
6694 Output instruction sizes to the asm file.
6696 @item -minit-stack=@var{N}
6697 @opindex minit-stack
6698 Specify the initial stack address, which may be a symbol or numeric value,
6699 @samp{__stack} is the default.
6701 @item -mno-interrupts
6702 @opindex mno-interrupts
6703 Generated code is not compatible with hardware interrupts.
6704 Code size will be smaller.
6706 @item -mcall-prologues
6707 @opindex mcall-prologues
6708 Functions prologues/epilogues expanded as call to appropriate
6709 subroutines. Code size will be smaller.
6711 @item -mno-tablejump
6712 @opindex mno-tablejump
6713 Do not generate tablejump insns which sometimes increase code size.
6716 @opindex mtiny-stack
6717 Change only the low 8 bits of the stack pointer.
6721 Assume int to be 8 bit integer. This affects the sizes of all types: A
6722 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
6723 and long long will be 4 bytes. Please note that this option does not
6724 comply to the C standards, but it will provide you with smaller code
6729 @subsection CRIS Options
6730 @cindex CRIS Options
6732 These options are defined specifically for the CRIS ports.
6735 @item -march=@var{architecture-type}
6736 @itemx -mcpu=@var{architecture-type}
6739 Generate code for the specified architecture. The choices for
6740 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
6741 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX.
6742 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
6745 @item -mtune=@var{architecture-type}
6747 Tune to @var{architecture-type} everything applicable about the generated
6748 code, except for the ABI and the set of available instructions. The
6749 choices for @var{architecture-type} are the same as for
6750 @option{-march=@var{architecture-type}}.
6752 @item -mmax-stack-frame=@var{n}
6753 @opindex mmax-stack-frame
6754 Warn when the stack frame of a function exceeds @var{n} bytes.
6756 @item -melinux-stacksize=@var{n}
6757 @opindex melinux-stacksize
6758 Only available with the @samp{cris-axis-aout} target. Arranges for
6759 indications in the program to the kernel loader that the stack of the
6760 program should be set to @var{n} bytes.
6766 The options @option{-metrax4} and @option{-metrax100} are synonyms for
6767 @option{-march=v3} and @option{-march=v8} respectively.
6769 @item -mmul-bug-workaround
6770 @itemx -mno-mul-bug-workaround
6771 @opindex mmul-bug-workaround
6772 @opindex mno-mul-bug-workaround
6773 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
6774 models where it applies. This option is active by default.
6778 Enable CRIS-specific verbose debug-related information in the assembly
6779 code. This option also has the effect to turn off the @samp{#NO_APP}
6780 formatted-code indicator to the assembler at the beginning of the
6785 Do not use condition-code results from previous instruction; always emit
6786 compare and test instructions before use of condition codes.
6788 @item -mno-side-effects
6789 @opindex mno-side-effects
6790 Do not emit instructions with side-effects in addressing modes other than
6794 @itemx -mno-stack-align
6796 @itemx -mno-data-align
6797 @itemx -mconst-align
6798 @itemx -mno-const-align
6799 @opindex mstack-align
6800 @opindex mno-stack-align
6801 @opindex mdata-align
6802 @opindex mno-data-align
6803 @opindex mconst-align
6804 @opindex mno-const-align
6805 These options (no-options) arranges (eliminate arrangements) for the
6806 stack-frame, individual data and constants to be aligned for the maximum
6807 single data access size for the chosen CPU model. The default is to
6808 arrange for 32-bit alignment. ABI details such as structure layout are
6809 not affected by these options.
6817 Similar to the stack- data- and const-align options above, these options
6818 arrange for stack-frame, writable data and constants to all be 32-bit,
6819 16-bit or 8-bit aligned. The default is 32-bit alignment.
6821 @item -mno-prologue-epilogue
6822 @itemx -mprologue-epilogue
6823 @opindex mno-prologue-epilogue
6824 @opindex mprologue-epilogue
6825 With @option{-mno-prologue-epilogue}, the normal function prologue and
6826 epilogue that sets up the stack-frame are omitted and no return
6827 instructions or return sequences are generated in the code. Use this
6828 option only together with visual inspection of the compiled code: no
6829 warnings or errors are generated when call-saved registers must be saved,
6830 or storage for local variable needs to be allocated.
6836 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
6837 instruction sequences that load addresses for functions from the PLT part
6838 of the GOT rather than (traditional on other architectures) calls to the
6839 PLT. The default is @option{-mgotplt}.
6843 Legacy no-op option only recognized with the cris-axis-aout target.
6847 Legacy no-op option only recognized with the cris-axis-elf and
6848 cris-axis-linux-gnu targets.
6852 Only recognized with the cris-axis-aout target, where it selects a
6853 GNU/linux-like multilib, include files and instruction set for
6858 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
6862 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
6863 to link with input-output functions from a simulator library. Code,
6864 initialized data and zero-initialized data are allocated consecutively.
6868 Like @option{-sim}, but pass linker options to locate initialized data at
6869 0x40000000 and zero-initialized data at 0x80000000.
6872 @node Darwin Options
6873 @subsection Darwin Options
6874 @cindex Darwin options
6876 These options are defined for all architectures running the Darwin operating
6877 system. They are useful for compatibility with other Mac OS compilers.
6882 Add the framework directory @var{dir} to the head of the list of
6883 directories to be searched for header files. These directories are
6884 interleaved with those specified by @option{-I} options and are
6885 scanned in a left-to-right order.
6887 A framework directory is a directory with frameworks in it. A
6888 framework is a directory with a @samp{"Headers"} and/or
6889 @samp{"PrivateHeaders"} directory contained directly in it that ends
6890 in @samp{".framework"}. The name of a framework is the name of this
6891 directory excluding the @samp{".framework"}. Headers associated with
6892 the framework are found in one of those two directories, with
6893 @samp{"Headers"} being searched first. A subframework is a framework
6894 directory that is in a framework's @samp{"Frameworks"} directory.
6895 Includes of subframework headers can only appear in a header of a
6896 framework that contains the subframework, or in a sibling subframework
6897 header. Two subframeworks are siblings if they occur in the same
6898 framework. A subframework should not have the same name as a
6899 framework, a warning will be issued if this is violated. Currently a
6900 subframework cannot have subframeworks, in the future, the mechanism
6901 may be extended to support this. The standard frameworks can be found
6902 in @samp{"/System/Library/Frameworks"}, @samp{"/Library/Frameworks"}
6903 and @samp{"/Local/Library/Frameworks"}. An example include looks like
6904 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
6905 the name of the framework and header.h is found in the
6906 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
6910 Emit debugging information for symbols that are used. For STABS
6911 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
6912 This is by default ON.
6916 Emit debugging information for all symbols and types.
6920 Loads all members of static archive libraries.
6921 See man ld(1) for more information.
6923 @item -arch_errors_fatal
6924 @opindex arch_errors_fatal
6925 Cause the errors having to do with files that have the wrong architecture
6929 @opindex bind_at_load
6930 Causes the output file to be marked such that the dynamic linker will
6931 bind all undefined references when the file is loaded or launched.
6935 Produce a Mach-o bundle format file.
6936 See man ld(1) for more information.
6938 @item -bundle_loader @var{executable}
6939 @opindex bundle_loader
6940 This specifies the @var{executable} that will be loading the build
6941 output file being linked. See man ld(1) for more information.
6943 @item -allowable_client @var{client_name}
6947 @itemx -compatibility_version
6948 @itemx -current_version
6949 @itemx -dependency-file
6951 @itemx -dylinker_install_name
6954 @itemx -exported_symbols_list
6956 @itemx -flat_namespace
6957 @itemx -force_cpusubtype_ALL
6958 @itemx -force_flat_namespace
6959 @itemx -headerpad_max_install_names
6962 @itemx -install_name
6963 @itemx -keep_private_externs
6964 @itemx -multi_module
6965 @itemx -multiply_defined
6966 @itemx -multiply_defined_unused
6968 @itemx -nofixprebinding
6971 @itemx -noseglinkedit
6972 @itemx -pagezero_size
6974 @itemx -prebind_all_twolevel_modules
6975 @itemx -private_bundle
6976 @itemx -read_only_relocs
6978 @itemx -sectobjectsymbols
6982 @itemx -sectobjectsymbols
6984 @itemx -seg_addr_table
6985 @itemx -seg_addr_table_filename
6988 @itemx -segs_read_only_addr
6989 @itemx -segs_read_write_addr
6990 @itemx -single_module
6993 @itemx -sub_umbrella
6994 @itemx -twolevel_namespace
6997 @itemx -unexported_symbols_list
6998 @itemx -weak_reference_mismatches
7001 @opindex allowable_client
7003 @opindex client_name
7004 @opindex compatibility_version
7005 @opindex current_version
7006 @opindex dependency-file
7008 @opindex dylinker_install_name
7011 @opindex exported_symbols_list
7013 @opindex flat_namespace
7014 @opindex force_cpusubtype_ALL
7015 @opindex force_flat_namespace
7016 @opindex headerpad_max_install_names
7019 @opindex install_name
7020 @opindex keep_private_externs
7021 @opindex multi_module
7022 @opindex multiply_defined
7023 @opindex multiply_defined_unused
7025 @opindex nofixprebinding
7026 @opindex nomultidefs
7028 @opindex noseglinkedit
7029 @opindex pagezero_size
7031 @opindex prebind_all_twolevel_modules
7032 @opindex private_bundle
7033 @opindex read_only_relocs
7035 @opindex sectobjectsymbols
7039 @opindex sectobjectsymbols
7041 @opindex seg_addr_table
7042 @opindex seg_addr_table_filename
7043 @opindex seglinkedit
7045 @opindex segs_read_only_addr
7046 @opindex segs_read_write_addr
7047 @opindex single_module
7049 @opindex sub_library
7050 @opindex sub_umbrella
7051 @opindex twolevel_namespace
7054 @opindex unexported_symbols_list
7055 @opindex weak_reference_mismatches
7056 @opindex whatsloaded
7058 These options are available for Darwin linker. Darwin linker man page
7059 describes them in detail.
7062 @node DEC Alpha Options
7063 @subsection DEC Alpha Options
7065 These @samp{-m} options are defined for the DEC Alpha implementations:
7068 @item -mno-soft-float
7070 @opindex mno-soft-float
7071 @opindex msoft-float
7072 Use (do not use) the hardware floating-point instructions for
7073 floating-point operations. When @option{-msoft-float} is specified,
7074 functions in @file{libgcc.a} will be used to perform floating-point
7075 operations. Unless they are replaced by routines that emulate the
7076 floating-point operations, or compiled in such a way as to call such
7077 emulations routines, these routines will issue floating-point
7078 operations. If you are compiling for an Alpha without floating-point
7079 operations, you must ensure that the library is built so as not to call
7082 Note that Alpha implementations without floating-point operations are
7083 required to have floating-point registers.
7088 @opindex mno-fp-regs
7089 Generate code that uses (does not use) the floating-point register set.
7090 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7091 register set is not used, floating point operands are passed in integer
7092 registers as if they were integers and floating-point results are passed
7093 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7094 so any function with a floating-point argument or return value called by code
7095 compiled with @option{-mno-fp-regs} must also be compiled with that
7098 A typical use of this option is building a kernel that does not use,
7099 and hence need not save and restore, any floating-point registers.
7103 The Alpha architecture implements floating-point hardware optimized for
7104 maximum performance. It is mostly compliant with the IEEE floating
7105 point standard. However, for full compliance, software assistance is
7106 required. This option generates code fully IEEE compliant code
7107 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7108 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7109 defined during compilation. The resulting code is less efficient but is
7110 able to correctly support denormalized numbers and exceptional IEEE
7111 values such as not-a-number and plus/minus infinity. Other Alpha
7112 compilers call this option @option{-ieee_with_no_inexact}.
7114 @item -mieee-with-inexact
7115 @opindex mieee-with-inexact
7116 This is like @option{-mieee} except the generated code also maintains
7117 the IEEE @var{inexact-flag}. Turning on this option causes the
7118 generated code to implement fully-compliant IEEE math. In addition to
7119 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7120 macro. On some Alpha implementations the resulting code may execute
7121 significantly slower than the code generated by default. Since there is
7122 very little code that depends on the @var{inexact-flag}, you should
7123 normally not specify this option. Other Alpha compilers call this
7124 option @option{-ieee_with_inexact}.
7126 @item -mfp-trap-mode=@var{trap-mode}
7127 @opindex mfp-trap-mode
7128 This option controls what floating-point related traps are enabled.
7129 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7130 The trap mode can be set to one of four values:
7134 This is the default (normal) setting. The only traps that are enabled
7135 are the ones that cannot be disabled in software (e.g., division by zero
7139 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7143 Like @samp{su}, but the instructions are marked to be safe for software
7144 completion (see Alpha architecture manual for details).
7147 Like @samp{su}, but inexact traps are enabled as well.
7150 @item -mfp-rounding-mode=@var{rounding-mode}
7151 @opindex mfp-rounding-mode
7152 Selects the IEEE rounding mode. Other Alpha compilers call this option
7153 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7158 Normal IEEE rounding mode. Floating point numbers are rounded towards
7159 the nearest machine number or towards the even machine number in case
7163 Round towards minus infinity.
7166 Chopped rounding mode. Floating point numbers are rounded towards zero.
7169 Dynamic rounding mode. A field in the floating point control register
7170 (@var{fpcr}, see Alpha architecture reference manual) controls the
7171 rounding mode in effect. The C library initializes this register for
7172 rounding towards plus infinity. Thus, unless your program modifies the
7173 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7176 @item -mtrap-precision=@var{trap-precision}
7177 @opindex mtrap-precision
7178 In the Alpha architecture, floating point traps are imprecise. This
7179 means without software assistance it is impossible to recover from a
7180 floating trap and program execution normally needs to be terminated.
7181 GCC can generate code that can assist operating system trap handlers
7182 in determining the exact location that caused a floating point trap.
7183 Depending on the requirements of an application, different levels of
7184 precisions can be selected:
7188 Program precision. This option is the default and means a trap handler
7189 can only identify which program caused a floating point exception.
7192 Function precision. The trap handler can determine the function that
7193 caused a floating point exception.
7196 Instruction precision. The trap handler can determine the exact
7197 instruction that caused a floating point exception.
7200 Other Alpha compilers provide the equivalent options called
7201 @option{-scope_safe} and @option{-resumption_safe}.
7203 @item -mieee-conformant
7204 @opindex mieee-conformant
7205 This option marks the generated code as IEEE conformant. You must not
7206 use this option unless you also specify @option{-mtrap-precision=i} and either
7207 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7208 is to emit the line @samp{.eflag 48} in the function prologue of the
7209 generated assembly file. Under DEC Unix, this has the effect that
7210 IEEE-conformant math library routines will be linked in.
7212 @item -mbuild-constants
7213 @opindex mbuild-constants
7214 Normally GCC examines a 32- or 64-bit integer constant to
7215 see if it can construct it from smaller constants in two or three
7216 instructions. If it cannot, it will output the constant as a literal and
7217 generate code to load it from the data segment at runtime.
7219 Use this option to require GCC to construct @emph{all} integer constants
7220 using code, even if it takes more instructions (the maximum is six).
7222 You would typically use this option to build a shared library dynamic
7223 loader. Itself a shared library, it must relocate itself in memory
7224 before it can find the variables and constants in its own data segment.
7230 Select whether to generate code to be assembled by the vendor-supplied
7231 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7249 Indicate whether GCC should generate code to use the optional BWX,
7250 CIX, FIX and MAX instruction sets. The default is to use the instruction
7251 sets supported by the CPU type specified via @option{-mcpu=} option or that
7252 of the CPU on which GCC was built if none was specified.
7257 @opindex mfloat-ieee
7258 Generate code that uses (does not use) VAX F and G floating point
7259 arithmetic instead of IEEE single and double precision.
7261 @item -mexplicit-relocs
7262 @itemx -mno-explicit-relocs
7263 @opindex mexplicit-relocs
7264 @opindex mno-explicit-relocs
7265 Older Alpha assemblers provided no way to generate symbol relocations
7266 except via assembler macros. Use of these macros does not allow
7267 optimal instruction scheduling. GNU binutils as of version 2.12
7268 supports a new syntax that allows the compiler to explicitly mark
7269 which relocations should apply to which instructions. This option
7270 is mostly useful for debugging, as GCC detects the capabilities of
7271 the assembler when it is built and sets the default accordingly.
7275 @opindex msmall-data
7276 @opindex mlarge-data
7277 When @option{-mexplicit-relocs} is in effect, static data is
7278 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7279 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7280 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7281 16-bit relocations off of the @code{$gp} register. This limits the
7282 size of the small data area to 64KB, but allows the variables to be
7283 directly accessed via a single instruction.
7285 The default is @option{-mlarge-data}. With this option the data area
7286 is limited to just below 2GB. Programs that require more than 2GB of
7287 data must use @code{malloc} or @code{mmap} to allocate the data in the
7288 heap instead of in the program's data segment.
7290 When generating code for shared libraries, @option{-fpic} implies
7291 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7295 @opindex msmall-text
7296 @opindex mlarge-text
7297 When @option{-msmall-text} is used, the compiler assumes that the
7298 code of the entire program (or shared library) fits in 4MB, and is
7299 thus reachable with a branch instruction. When @option{-msmall-data}
7300 is used, the compiler can assume that all local symbols share the
7301 same @code{$gp} value, and thus reduce the number of instructions
7302 required for a function call from 4 to 1.
7304 The default is @option{-mlarge-text}.
7306 @item -mcpu=@var{cpu_type}
7308 Set the instruction set and instruction scheduling parameters for
7309 machine type @var{cpu_type}. You can specify either the @samp{EV}
7310 style name or the corresponding chip number. GCC supports scheduling
7311 parameters for the EV4, EV5 and EV6 family of processors and will
7312 choose the default values for the instruction set from the processor
7313 you specify. If you do not specify a processor type, GCC will default
7314 to the processor on which the compiler was built.
7316 Supported values for @var{cpu_type} are
7322 Schedules as an EV4 and has no instruction set extensions.
7326 Schedules as an EV5 and has no instruction set extensions.
7330 Schedules as an EV5 and supports the BWX extension.
7335 Schedules as an EV5 and supports the BWX and MAX extensions.
7339 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7343 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7346 @item -mtune=@var{cpu_type}
7348 Set only the instruction scheduling parameters for machine type
7349 @var{cpu_type}. The instruction set is not changed.
7351 @item -mmemory-latency=@var{time}
7352 @opindex mmemory-latency
7353 Sets the latency the scheduler should assume for typical memory
7354 references as seen by the application. This number is highly
7355 dependent on the memory access patterns used by the application
7356 and the size of the external cache on the machine.
7358 Valid options for @var{time} are
7362 A decimal number representing clock cycles.
7368 The compiler contains estimates of the number of clock cycles for
7369 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7370 (also called Dcache, Scache, and Bcache), as well as to main memory.
7371 Note that L3 is only valid for EV5.
7376 @node DEC Alpha/VMS Options
7377 @subsection DEC Alpha/VMS Options
7379 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7382 @item -mvms-return-codes
7383 @opindex mvms-return-codes
7384 Return VMS condition codes from main. The default is to return POSIX
7385 style condition (e.g.@ error) codes.
7389 @subsection FRV Options
7396 Only use the first 32 general purpose registers.
7401 Use all 64 general purpose registers.
7406 Use only the first 32 floating point registers.
7411 Use all 64 floating point registers
7414 @opindex mhard-float
7416 Use hardware instructions for floating point operations.
7419 @opindex msoft-float
7421 Use library routines for floating point operations.
7426 Dynamically allocate condition code registers.
7431 Do not try to dynamically allocate condition code registers, only
7432 use @code{icc0} and @code{fcc0}.
7437 Change ABI to use double word insns.
7442 Do not use double word instructions.
7447 Use floating point double instructions.
7452 Do not use floating point double instructions.
7457 Use media instructions.
7462 Do not use media instructions.
7467 Use multiply and add/subtract instructions.
7472 Do not use multiply and add/subtract instructions.
7475 @opindex mlibrary-pic
7477 Generate position-independent EABI code.
7482 Use only the first four media accumulator registers.
7487 Use all eight media accumulator registers.
7492 Pack VLIW instructions.
7497 Do not pack VLIW instructions.
7502 Do not mark ABI switches in e_flags.
7507 Enable the use of conditional-move instructions (default).
7509 This switch is mainly for debugging the compiler and will likely be removed
7510 in a future version.
7512 @item -mno-cond-move
7513 @opindex mno-cond-move
7515 Disable the use of conditional-move instructions.
7517 This switch is mainly for debugging the compiler and will likely be removed
7518 in a future version.
7523 Enable the use of conditional set instructions (default).
7525 This switch is mainly for debugging the compiler and will likely be removed
7526 in a future version.
7531 Disable the use of conditional set instructions.
7533 This switch is mainly for debugging the compiler and will likely be removed
7534 in a future version.
7539 Enable the use of conditional execution (default).
7541 This switch is mainly for debugging the compiler and will likely be removed
7542 in a future version.
7544 @item -mno-cond-exec
7545 @opindex mno-cond-exec
7547 Disable the use of conditional execution.
7549 This switch is mainly for debugging the compiler and will likely be removed
7550 in a future version.
7553 @opindex mvliw-branch
7555 Run a pass to pack branches into VLIW instructions (default).
7557 This switch is mainly for debugging the compiler and will likely be removed
7558 in a future version.
7560 @item -mno-vliw-branch
7561 @opindex mno-vliw-branch
7563 Do not run a pass to pack branches into VLIW instructions.
7565 This switch is mainly for debugging the compiler and will likely be removed
7566 in a future version.
7568 @item -mmulti-cond-exec
7569 @opindex mmulti-cond-exec
7571 Enable optimization of @code{&&} and @code{||} in conditional execution
7574 This switch is mainly for debugging the compiler and will likely be removed
7575 in a future version.
7577 @item -mno-multi-cond-exec
7578 @opindex mno-multi-cond-exec
7580 Disable optimization of @code{&&} and @code{||} in conditional execution.
7582 This switch is mainly for debugging the compiler and will likely be removed
7583 in a future version.
7585 @item -mnested-cond-exec
7586 @opindex mnested-cond-exec
7588 Enable nested conditional execution optimizations (default).
7590 This switch is mainly for debugging the compiler and will likely be removed
7591 in a future version.
7593 @item -mno-nested-cond-exec
7594 @opindex mno-nested-cond-exec
7596 Disable nested conditional execution optimizations.
7598 This switch is mainly for debugging the compiler and will likely be removed
7599 in a future version.
7601 @item -mtomcat-stats
7602 @opindex mtomcat-stats
7604 Cause gas to print out tomcat statistics.
7606 @item -mcpu=@var{cpu}
7609 Select the processor type for which to generate code. Possible values are
7610 @samp{simple}, @samp{tomcat}, @samp{fr500}, @samp{fr400}, @samp{fr300},
7615 @node H8/300 Options
7616 @subsection H8/300 Options
7618 These @samp{-m} options are defined for the H8/300 implementations:
7623 Shorten some address references at link time, when possible; uses the
7624 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
7625 ld, Using ld}, for a fuller description.
7629 Generate code for the H8/300H@.
7633 Generate code for the H8S@.
7637 Generate code for the H8S and H8/300H in the normal mode. This switch
7638 must be used either with -mh or -ms.
7642 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
7646 Make @code{int} data 32 bits by default.
7650 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
7651 The default for the H8/300H and H8S is to align longs and floats on 4
7653 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
7654 This option has no effect on the H8/300.
7658 @subsection HPPA Options
7659 @cindex HPPA Options
7661 These @samp{-m} options are defined for the HPPA family of computers:
7664 @item -march=@var{architecture-type}
7666 Generate code for the specified architecture. The choices for
7667 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
7668 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
7669 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
7670 architecture option for your machine. Code compiled for lower numbered
7671 architectures will run on higher numbered architectures, but not the
7674 PA 2.0 support currently requires gas snapshot 19990413 or later. The
7675 next release of binutils (current is 2.9.1) will probably contain PA 2.0
7679 @itemx -mpa-risc-1-1
7680 @itemx -mpa-risc-2-0
7681 @opindex mpa-risc-1-0
7682 @opindex mpa-risc-1-1
7683 @opindex mpa-risc-2-0
7684 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
7687 @opindex mbig-switch
7688 Generate code suitable for big switch tables. Use this option only if
7689 the assembler/linker complain about out of range branches within a switch
7692 @item -mjump-in-delay
7693 @opindex mjump-in-delay
7694 Fill delay slots of function calls with unconditional jump instructions
7695 by modifying the return pointer for the function call to be the target
7696 of the conditional jump.
7698 @item -mdisable-fpregs
7699 @opindex mdisable-fpregs
7700 Prevent floating point registers from being used in any manner. This is
7701 necessary for compiling kernels which perform lazy context switching of
7702 floating point registers. If you use this option and attempt to perform
7703 floating point operations, the compiler will abort.
7705 @item -mdisable-indexing
7706 @opindex mdisable-indexing
7707 Prevent the compiler from using indexing address modes. This avoids some
7708 rather obscure problems when compiling MIG generated code under MACH@.
7710 @item -mno-space-regs
7711 @opindex mno-space-regs
7712 Generate code that assumes the target has no space registers. This allows
7713 GCC to generate faster indirect calls and use unscaled index address modes.
7715 Such code is suitable for level 0 PA systems and kernels.
7717 @item -mfast-indirect-calls
7718 @opindex mfast-indirect-calls
7719 Generate code that assumes calls never cross space boundaries. This
7720 allows GCC to emit code which performs faster indirect calls.
7722 This option will not work in the presence of shared libraries or nested
7725 @item -mfixed-range=@var{register-range}
7726 @opindex mfixed-range
7727 Generate code treating the given register range as fixed registers.
7728 A fixed register is one that the register allocator can not use. This is
7729 useful when compiling kernel code. A register range is specified as
7730 two registers separated by a dash. Multiple register ranges can be
7731 specified separated by a comma.
7733 @item -mlong-load-store
7734 @opindex mlong-load-store
7735 Generate 3-instruction load and store sequences as sometimes required by
7736 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
7739 @item -mportable-runtime
7740 @opindex mportable-runtime
7741 Use the portable calling conventions proposed by HP for ELF systems.
7745 Enable the use of assembler directives only GAS understands.
7747 @item -mschedule=@var{cpu-type}
7749 Schedule code according to the constraints for the machine type
7750 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
7751 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
7752 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
7753 proper scheduling option for your machine. The default scheduling is
7757 @opindex mlinker-opt
7758 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
7759 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
7760 linkers in which they give bogus error messages when linking some programs.
7763 @opindex msoft-float
7764 Generate output containing library calls for floating point.
7765 @strong{Warning:} the requisite libraries are not available for all HPPA
7766 targets. Normally the facilities of the machine's usual C compiler are
7767 used, but this cannot be done directly in cross-compilation. You must make
7768 your own arrangements to provide suitable library functions for
7769 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
7770 does provide software floating point support.
7772 @option{-msoft-float} changes the calling convention in the output file;
7773 therefore, it is only useful if you compile @emph{all} of a program with
7774 this option. In particular, you need to compile @file{libgcc.a}, the
7775 library that comes with GCC, with @option{-msoft-float} in order for
7780 Generate the predefine, @code{_SIO}, for server IO. The default is
7781 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
7782 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO. These
7783 options are available under HP-UX and HI-UX.
7787 Use GNU ld specific options. This passes @option{-shared} to ld when
7788 building a shared library. It is the default when GCC is configured,
7789 explicitly or implicitly, with the GNU linker. This option does not
7790 have any affect on which ld is called, it only changes what parameters
7791 are passed to that ld. The ld that is called is determined by the
7792 @option{--with-ld} configure option, GCC's program search path, and
7793 finally by the user's @env{PATH}. The linker used by GCC can be printed
7794 using @samp{which `gcc -print-prog-name=ld`}.
7798 Use HP ld specific options. This passes @option{-b} to ld when building
7799 a shared library and passes @option{+Accept TypeMismatch} to ld on all
7800 links. It is the default when GCC is configured, explicitly or
7801 implicitly, with the HP linker. This option does not have any affect on
7802 which ld is called, it only changes what parameters are passed to that
7803 ld. The ld that is called is determined by the @option{--with-ld}
7804 configure option, GCC's program search path, and finally by the user's
7805 @env{PATH}. The linker used by GCC can be printed using @samp{which
7806 `gcc -print-prog-name=ld`}.
7811 Select the FDPIC ABI, that uses function descriptors to represent
7812 pointers to functions. Without any PIC/PIE-related options, it
7813 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
7814 assumes GOT entries and small data are within a 12-bit range from the
7815 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
7816 are computed with 32 bits.
7819 @opindex minline-plt
7821 Enable inlining of PLT entries in function calls to functions that are
7822 not known to bind locally. It has no effect without @option{-mfdpic}.
7823 It's enabled by default if optimizing for speed and compiling for
7824 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
7825 optimization option such as @option{-O3} or above is present in the
7831 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
7832 that is known to be in read-only sections. It's enabled by default,
7833 except for @option{-fpic} or @option{-fpie}: even though it may help
7834 make the global offset table smaller, it trades 1 instruction for 4.
7835 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
7836 one of which may be shared by multiple symbols, and it avoids the need
7837 for a GOT entry for the referenced symbol, so it's more likely to be a
7838 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
7840 @item -multilib-library-pic
7841 @opindex multilib-library-pic
7843 Link with the (library, not FD) pic libraries. It's implied by
7844 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
7845 @option{-fpic} without @option{-mfdpic}. You should never have to use
7851 Follow the EABI requirement of always creating a frame pointer whenever
7852 a stack frame is allocated. This option is enabled by default and can
7853 be disabled with @option{-mno-linked-fp}.
7856 @opindex mno-long-calls
7857 Generate code that uses long call sequences. This ensures that a call
7858 is always able to reach linker generated stubs. The default is to generate
7859 long calls only when the distance from the call site to the beginning
7860 of the function or translation unit, as the case may be, exceeds a
7861 predefined limit set by the branch type being used. The limits for
7862 normal calls are 7,600,000 and 240,000 bytes, respectively for the
7863 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
7866 Distances are measured from the beginning of functions when using the
7867 @option{-ffunction-sections} option, or when using the @option{-mgas}
7868 and @option{-mno-portable-runtime} options together under HP-UX with
7871 It is normally not desirable to use this option as it will degrade
7872 performance. However, it may be useful in large applications,
7873 particularly when partial linking is used to build the application.
7875 The types of long calls used depends on the capabilities of the
7876 assembler and linker, and the type of code being generated. The
7877 impact on systems that support long absolute calls, and long pic
7878 symbol-difference or pc-relative calls should be relatively small.
7879 However, an indirect call is used on 32-bit ELF systems in pic code
7880 and it is quite long.
7884 Suppress the generation of link options to search libdld.sl when the
7885 @option{-static} option is specified on HP-UX 10 and later.
7889 The HP-UX implementation of setlocale in libc has a dependency on
7890 libdld.sl. There isn't an archive version of libdld.sl. Thus,
7891 when the @option{-static} option is specified, special link options
7892 are needed to resolve this dependency.
7894 On HP-UX 10 and later, the GCC driver adds the necessary options to
7895 link with libdld.sl when the @option{-static} option is specified.
7896 This causes the resulting binary to be dynamic. On the 64-bit port,
7897 the linkers generate dynamic binaries by default in any case. The
7898 @option{-nolibdld} option can be used to prevent the GCC driver from
7899 adding these link options.
7903 Add support for multithreading with the @dfn{dce thread} library
7904 under HP-UX. This option sets flags for both the preprocessor and
7908 @node i386 and x86-64 Options
7909 @subsection Intel 386 and AMD x86-64 Options
7910 @cindex i386 Options
7911 @cindex x86-64 Options
7912 @cindex Intel 386 Options
7913 @cindex AMD x86-64 Options
7915 These @samp{-m} options are defined for the i386 and x86-64 family of
7919 @item -mtune=@var{cpu-type}
7921 Tune to @var{cpu-type} everything applicable about the generated code, except
7922 for the ABI and the set of available instructions. The choices for
7926 Original Intel's i386 CPU.
7928 Intel's i486 CPU. (No scheduling is implemented for this chip.)
7930 Intel Pentium CPU with no MMX support.
7932 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
7933 @item i686, pentiumpro
7934 Intel PentiumPro CPU.
7936 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
7937 @item pentium3, pentium3m
7938 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
7941 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
7942 support. Used by Centrino notebooks.
7943 @item pentium4, pentium4m
7944 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
7946 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
7949 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
7950 SSE2 and SSE3 instruction set support.
7952 AMD K6 CPU with MMX instruction set support.
7954 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
7955 @item athlon, athlon-tbird
7956 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
7958 @item athlon-4, athlon-xp, athlon-mp
7959 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
7960 instruction set support.
7961 @item k8, opteron, athlon64, athlon-fx
7962 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
7963 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
7965 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
7968 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
7969 instruction set support.
7971 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
7972 implemented for this chip.)
7974 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
7975 implemented for this chip.)
7978 While picking a specific @var{cpu-type} will schedule things appropriately
7979 for that particular chip, the compiler will not generate any code that
7980 does not run on the i386 without the @option{-march=@var{cpu-type}} option
7983 @item -march=@var{cpu-type}
7985 Generate instructions for the machine type @var{cpu-type}. The choices
7986 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
7987 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
7989 @item -mcpu=@var{cpu-type}
7991 A deprecated synonym for @option{-mtune}.
8000 @opindex mpentiumpro
8001 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8002 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8003 These synonyms are deprecated.
8005 @item -mfpmath=@var{unit}
8007 Generate floating point arithmetics for selected unit @var{unit}. The choices
8012 Use the standard 387 floating point coprocessor present majority of chips and
8013 emulated otherwise. Code compiled with this option will run almost everywhere.
8014 The temporary results are computed in 80bit precision instead of precision
8015 specified by the type resulting in slightly different results compared to most
8016 of other chips. See @option{-ffloat-store} for more detailed description.
8018 This is the default choice for i386 compiler.
8021 Use scalar floating point instructions present in the SSE instruction set.
8022 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8023 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8024 instruction set supports only single precision arithmetics, thus the double and
8025 extended precision arithmetics is still done using 387. Later version, present
8026 only in Pentium4 and the future AMD x86-64 chips supports double precision
8029 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8030 @option{-msse2} switches to enable SSE extensions and make this option
8031 effective. For x86-64 compiler, these extensions are enabled by default.
8033 The resulting code should be considerably faster in the majority of cases and avoid
8034 the numerical instability problems of 387 code, but may break some existing
8035 code that expects temporaries to be 80bit.
8037 This is the default choice for the x86-64 compiler.
8040 Attempt to utilize both instruction sets at once. This effectively double the
8041 amount of available registers and on chips with separate execution units for
8042 387 and SSE the execution resources too. Use this option with care, as it is
8043 still experimental, because the GCC register allocator does not model separate
8044 functional units well resulting in instable performance.
8047 @item -masm=@var{dialect}
8048 @opindex masm=@var{dialect}
8049 Output asm instructions using selected @var{dialect}. Supported choices are
8050 @samp{intel} or @samp{att} (the default one).
8055 @opindex mno-ieee-fp
8056 Control whether or not the compiler uses IEEE floating point
8057 comparisons. These handle correctly the case where the result of a
8058 comparison is unordered.
8061 @opindex msoft-float
8062 Generate output containing library calls for floating point.
8063 @strong{Warning:} the requisite libraries are not part of GCC@.
8064 Normally the facilities of the machine's usual C compiler are used, but
8065 this can't be done directly in cross-compilation. You must make your
8066 own arrangements to provide suitable library functions for
8069 On machines where a function returns floating point results in the 80387
8070 register stack, some floating point opcodes may be emitted even if
8071 @option{-msoft-float} is used.
8073 @item -mno-fp-ret-in-387
8074 @opindex mno-fp-ret-in-387
8075 Do not use the FPU registers for return values of functions.
8077 The usual calling convention has functions return values of types
8078 @code{float} and @code{double} in an FPU register, even if there
8079 is no FPU@. The idea is that the operating system should emulate
8082 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8083 in ordinary CPU registers instead.
8085 @item -mno-fancy-math-387
8086 @opindex mno-fancy-math-387
8087 Some 387 emulators do not support the @code{sin}, @code{cos} and
8088 @code{sqrt} instructions for the 387. Specify this option to avoid
8089 generating those instructions. This option is the default on FreeBSD,
8090 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8091 indicates that the target cpu will always have an FPU and so the
8092 instruction will not need emulation. As of revision 2.6.1, these
8093 instructions are not generated unless you also use the
8094 @option{-funsafe-math-optimizations} switch.
8096 @item -malign-double
8097 @itemx -mno-align-double
8098 @opindex malign-double
8099 @opindex mno-align-double
8100 Control whether GCC aligns @code{double}, @code{long double}, and
8101 @code{long long} variables on a two word boundary or a one word
8102 boundary. Aligning @code{double} variables on a two word boundary will
8103 produce code that runs somewhat faster on a @samp{Pentium} at the
8104 expense of more memory.
8106 @strong{Warning:} if you use the @option{-malign-double} switch,
8107 structures containing the above types will be aligned differently than
8108 the published application binary interface specifications for the 386
8109 and will not be binary compatible with structures in code compiled
8110 without that switch.
8112 @item -m96bit-long-double
8113 @itemx -m128bit-long-double
8114 @opindex m96bit-long-double
8115 @opindex m128bit-long-double
8116 These switches control the size of @code{long double} type. The i386
8117 application binary interface specifies the size to be 96 bits,
8118 so @option{-m96bit-long-double} is the default in 32 bit mode.
8120 Modern architectures (Pentium and newer) would prefer @code{long double}
8121 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8122 conforming to the ABI, this would not be possible. So specifying a
8123 @option{-m128bit-long-double} will align @code{long double}
8124 to a 16 byte boundary by padding the @code{long double} with an additional
8127 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8128 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8130 Notice that neither of these options enable any extra precision over the x87
8131 standard of 80 bits for a @code{long double}.
8133 @strong{Warning:} if you override the default value for your target ABI, the
8134 structures and arrays containing @code{long double} variables will change
8135 their size as well as function calling convention for function taking
8136 @code{long double} will be modified. Hence they will not be binary
8137 compatible with arrays or structures in code compiled without that switch.
8141 @itemx -mno-svr3-shlib
8142 @opindex msvr3-shlib
8143 @opindex mno-svr3-shlib
8144 Control whether GCC places uninitialized local variables into the
8145 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8146 into @code{bss}. These options are meaningful only on System V Release 3.
8150 Use a different function-calling convention, in which functions that
8151 take a fixed number of arguments return with the @code{ret} @var{num}
8152 instruction, which pops their arguments while returning. This saves one
8153 instruction in the caller since there is no need to pop the arguments
8156 You can specify that an individual function is called with this calling
8157 sequence with the function attribute @samp{stdcall}. You can also
8158 override the @option{-mrtd} option by using the function attribute
8159 @samp{cdecl}. @xref{Function Attributes}.
8161 @strong{Warning:} this calling convention is incompatible with the one
8162 normally used on Unix, so you cannot use it if you need to call
8163 libraries compiled with the Unix compiler.
8165 Also, you must provide function prototypes for all functions that
8166 take variable numbers of arguments (including @code{printf});
8167 otherwise incorrect code will be generated for calls to those
8170 In addition, seriously incorrect code will result if you call a
8171 function with too many arguments. (Normally, extra arguments are
8172 harmlessly ignored.)
8174 @item -mregparm=@var{num}
8176 Control how many registers are used to pass integer arguments. By
8177 default, no registers are used to pass arguments, and at most 3
8178 registers can be used. You can control this behavior for a specific
8179 function by using the function attribute @samp{regparm}.
8180 @xref{Function Attributes}.
8182 @strong{Warning:} if you use this switch, and
8183 @var{num} is nonzero, then you must build all modules with the same
8184 value, including any libraries. This includes the system libraries and
8187 @item -mpreferred-stack-boundary=@var{num}
8188 @opindex mpreferred-stack-boundary
8189 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8190 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8191 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8192 size (@option{-Os}), in which case the default is the minimum correct
8193 alignment (4 bytes for x86, and 8 bytes for x86-64).
8195 On Pentium and PentiumPro, @code{double} and @code{long double} values
8196 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8197 suffer significant run time performance penalties. On Pentium III, the
8198 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8199 penalties if it is not 16 byte aligned.
8201 To ensure proper alignment of this values on the stack, the stack boundary
8202 must be as aligned as that required by any value stored on the stack.
8203 Further, every function must be generated such that it keeps the stack
8204 aligned. Thus calling a function compiled with a higher preferred
8205 stack boundary from a function compiled with a lower preferred stack
8206 boundary will most likely misalign the stack. It is recommended that
8207 libraries that use callbacks always use the default setting.
8209 This extra alignment does consume extra stack space, and generally
8210 increases code size. Code that is sensitive to stack space usage, such
8211 as embedded systems and operating system kernels, may want to reduce the
8212 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8230 These switches enable or disable the use of built-in functions that allow
8231 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8234 @xref{X86 Built-in Functions}, for details of the functions enabled
8235 and disabled by these switches.
8237 To have SSE/SSE2 instructions generated automatically from floating-point
8238 code, see @option{-mfpmath=sse}.
8241 @itemx -mno-push-args
8243 @opindex mno-push-args
8244 Use PUSH operations to store outgoing parameters. This method is shorter
8245 and usually equally fast as method using SUB/MOV operations and is enabled
8246 by default. In some cases disabling it may improve performance because of
8247 improved scheduling and reduced dependencies.
8249 @item -maccumulate-outgoing-args
8250 @opindex maccumulate-outgoing-args
8251 If enabled, the maximum amount of space required for outgoing arguments will be
8252 computed in the function prologue. This is faster on most modern CPUs
8253 because of reduced dependencies, improved scheduling and reduced stack usage
8254 when preferred stack boundary is not equal to 2. The drawback is a notable
8255 increase in code size. This switch implies @option{-mno-push-args}.
8259 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8260 on thread-safe exception handling must compile and link all code with the
8261 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8262 @option{-D_MT}; when linking, it links in a special thread helper library
8263 @option{-lmingwthrd} which cleans up per thread exception handling data.
8265 @item -mno-align-stringops
8266 @opindex mno-align-stringops
8267 Do not align destination of inlined string operations. This switch reduces
8268 code size and improves performance in case the destination is already aligned,
8269 but GCC doesn't know about it.
8271 @item -minline-all-stringops
8272 @opindex minline-all-stringops
8273 By default GCC inlines string operations only when destination is known to be
8274 aligned at least to 4 byte boundary. This enables more inlining, increase code
8275 size, but may improve performance of code that depends on fast memcpy, strlen
8276 and memset for short lengths.
8278 @item -momit-leaf-frame-pointer
8279 @opindex momit-leaf-frame-pointer
8280 Don't keep the frame pointer in a register for leaf functions. This
8281 avoids the instructions to save, set up and restore frame pointers and
8282 makes an extra register available in leaf functions. The option
8283 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8284 which might make debugging harder.
8286 @item -mtls-direct-seg-refs
8287 @itemx -mno-tls-direct-seg-refs
8288 @opindex mtls-direct-seg-refs
8289 Controls whether TLS variables may be accessed with offsets from the
8290 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8291 or whether the thread base pointer must be added. Whether or not this
8292 is legal depends on the operating system, and whether it maps the
8293 segment to cover the entire TLS area.
8295 For systems that use GNU libc, the default is on.
8298 These @samp{-m} switches are supported in addition to the above
8299 on AMD x86-64 processors in 64-bit environments.
8306 Generate code for a 32-bit or 64-bit environment.
8307 The 32-bit environment sets int, long and pointer to 32 bits and
8308 generates code that runs on any i386 system.
8309 The 64-bit environment sets int to 32 bits and long and pointer
8310 to 64 bits and generates code for AMD's x86-64 architecture.
8313 @opindex no-red-zone
8314 Do not use a so called red zone for x86-64 code. The red zone is mandated
8315 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8316 stack pointer that will not be modified by signal or interrupt handlers
8317 and therefore can be used for temporary data without adjusting the stack
8318 pointer. The flag @option{-mno-red-zone} disables this red zone.
8320 @item -mcmodel=small
8321 @opindex mcmodel=small
8322 Generate code for the small code model: the program and its symbols must
8323 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8324 Programs can be statically or dynamically linked. This is the default
8327 @item -mcmodel=kernel
8328 @opindex mcmodel=kernel
8329 Generate code for the kernel code model. The kernel runs in the
8330 negative 2 GB of the address space.
8331 This model has to be used for Linux kernel code.
8333 @item -mcmodel=medium
8334 @opindex mcmodel=medium
8335 Generate code for the medium model: The program is linked in the lower 2
8336 GB of the address space but symbols can be located anywhere in the
8337 address space. Programs can be statically or dynamically linked, but
8338 building of shared libraries are not supported with the medium model.
8340 @item -mcmodel=large
8341 @opindex mcmodel=large
8342 Generate code for the large model: This model makes no assumptions
8343 about addresses and sizes of sections. Currently GCC does not implement
8348 @subsection IA-64 Options
8349 @cindex IA-64 Options
8351 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8355 @opindex mbig-endian
8356 Generate code for a big endian target. This is the default for HP-UX@.
8358 @item -mlittle-endian
8359 @opindex mlittle-endian
8360 Generate code for a little endian target. This is the default for AIX5
8367 Generate (or don't) code for the GNU assembler. This is the default.
8368 @c Also, this is the default if the configure option @option{--with-gnu-as}
8375 Generate (or don't) code for the GNU linker. This is the default.
8376 @c Also, this is the default if the configure option @option{--with-gnu-ld}
8381 Generate code that does not use a global pointer register. The result
8382 is not position independent code, and violates the IA-64 ABI@.
8384 @item -mvolatile-asm-stop
8385 @itemx -mno-volatile-asm-stop
8386 @opindex mvolatile-asm-stop
8387 @opindex mno-volatile-asm-stop
8388 Generate (or don't) a stop bit immediately before and after volatile asm
8393 Generate code that works around Itanium B step errata.
8395 @item -mregister-names
8396 @itemx -mno-register-names
8397 @opindex mregister-names
8398 @opindex mno-register-names
8399 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
8400 the stacked registers. This may make assembler output more readable.
8406 Disable (or enable) optimizations that use the small data section. This may
8407 be useful for working around optimizer bugs.
8410 @opindex mconstant-gp
8411 Generate code that uses a single constant global pointer value. This is
8412 useful when compiling kernel code.
8416 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
8417 This is useful when compiling firmware code.
8419 @item -minline-float-divide-min-latency
8420 @opindex minline-float-divide-min-latency
8421 Generate code for inline divides of floating point values
8422 using the minimum latency algorithm.
8424 @item -minline-float-divide-max-throughput
8425 @opindex minline-float-divide-max-throughput
8426 Generate code for inline divides of floating point values
8427 using the maximum throughput algorithm.
8429 @item -minline-int-divide-min-latency
8430 @opindex minline-int-divide-min-latency
8431 Generate code for inline divides of integer values
8432 using the minimum latency algorithm.
8434 @item -minline-int-divide-max-throughput
8435 @opindex minline-int-divide-max-throughput
8436 Generate code for inline divides of integer values
8437 using the maximum throughput algorithm.
8439 @item -mno-dwarf2-asm
8441 @opindex mno-dwarf2-asm
8442 @opindex mdwarf2-asm
8443 Don't (or do) generate assembler code for the DWARF2 line number debugging
8444 info. This may be useful when not using the GNU assembler.
8446 @item -mfixed-range=@var{register-range}
8447 @opindex mfixed-range
8448 Generate code treating the given register range as fixed registers.
8449 A fixed register is one that the register allocator can not use. This is
8450 useful when compiling kernel code. A register range is specified as
8451 two registers separated by a dash. Multiple register ranges can be
8452 specified separated by a comma.
8454 @item -mearly-stop-bits
8455 @itemx -mno-early-stop-bits
8456 @opindex mearly-stop-bits
8457 @opindex mno-early-stop-bits
8458 Allow stop bits to be placed earlier than immediately preceding the
8459 instruction that triggered the stop bit. This can improve instruction
8460 scheduling, but does not always do so.
8463 @node M32R/D Options
8464 @subsection M32R/D Options
8465 @cindex M32R/D options
8467 These @option{-m} options are defined for Renesas M32R/D architectures:
8472 Generate code for the M32R/2@.
8476 Generate code for the M32R/X@.
8480 Generate code for the M32R@. This is the default.
8483 @opindex mmodel=small
8484 Assume all objects live in the lower 16MB of memory (so that their addresses
8485 can be loaded with the @code{ld24} instruction), and assume all subroutines
8486 are reachable with the @code{bl} instruction.
8487 This is the default.
8489 The addressability of a particular object can be set with the
8490 @code{model} attribute.
8492 @item -mmodel=medium
8493 @opindex mmodel=medium
8494 Assume objects may be anywhere in the 32-bit address space (the compiler
8495 will generate @code{seth/add3} instructions to load their addresses), and
8496 assume all subroutines are reachable with the @code{bl} instruction.
8499 @opindex mmodel=large
8500 Assume objects may be anywhere in the 32-bit address space (the compiler
8501 will generate @code{seth/add3} instructions to load their addresses), and
8502 assume subroutines may not be reachable with the @code{bl} instruction
8503 (the compiler will generate the much slower @code{seth/add3/jl}
8504 instruction sequence).
8507 @opindex msdata=none
8508 Disable use of the small data area. Variables will be put into
8509 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
8510 @code{section} attribute has been specified).
8511 This is the default.
8513 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
8514 Objects may be explicitly put in the small data area with the
8515 @code{section} attribute using one of these sections.
8518 @opindex msdata=sdata
8519 Put small global and static data in the small data area, but do not
8520 generate special code to reference them.
8524 Put small global and static data in the small data area, and generate
8525 special instructions to reference them.
8529 @cindex smaller data references
8530 Put global and static objects less than or equal to @var{num} bytes
8531 into the small data or bss sections instead of the normal data or bss
8532 sections. The default value of @var{num} is 8.
8533 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
8534 for this option to have any effect.
8536 All modules should be compiled with the same @option{-G @var{num}} value.
8537 Compiling with different values of @var{num} may or may not work; if it
8538 doesn't the linker will give an error message---incorrect code will not be
8543 Makes the M32R specific code in the compiler display some statistics
8544 that might help in debugging programs.
8547 @opindex malign-loops
8548 Align all loops to a 32-byte boundary.
8550 @item -mno-align-loops
8551 @opindex mno-align-loops
8552 Do not enforce a 32-byte alignment for loops. This is the default.
8554 @item -missue-rate=@var{number}
8555 @opindex missue-rate=@var{number}
8556 Issue @var{number} instructions per cycle. @var{number} can only be 1
8559 @item -mbranch-cost=@var{number}
8560 @opindex mbranch-cost=@var{number}
8561 @var{number} can only be 1 or 2. If it is 1 then branches will be
8562 preferred over conditional code, if it is 2, then the opposite will
8565 @item -mflush-trap=@var{number}
8566 @opindex mflush-trap=@var{number}
8567 Specifies the trap number to use to flush the cache. The default is
8568 12. Valid numbers are between 0 and 15 inclusive.
8570 @item -mno-flush-trap
8571 @opindex mno-flush-trap
8572 Specifies that the cache cannot be flushed by using a trap.
8574 @item -mflush-func=@var{name}
8575 @opindex mflush-func=@var{name}
8576 Specifies the name of the operating system function to call to flush
8577 the cache. The default is @emph{_flush_cache}, but a function call
8578 will only be used if a trap is not available.
8580 @item -mno-flush-func
8581 @opindex mno-flush-func
8582 Indicates that there is no OS function for flushing the cache.
8586 @node M680x0 Options
8587 @subsection M680x0 Options
8588 @cindex M680x0 options
8590 These are the @samp{-m} options defined for the 68000 series. The default
8591 values for these options depends on which style of 68000 was selected when
8592 the compiler was configured; the defaults for the most common choices are
8600 Generate output for a 68000. This is the default
8601 when the compiler is configured for 68000-based systems.
8603 Use this option for microcontrollers with a 68000 or EC000 core,
8604 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
8610 Generate output for a 68020. This is the default
8611 when the compiler is configured for 68020-based systems.
8615 Generate output containing 68881 instructions for floating point.
8616 This is the default for most 68020 systems unless @option{--nfp} was
8617 specified when the compiler was configured.
8621 Generate output for a 68030. This is the default when the compiler is
8622 configured for 68030-based systems.
8626 Generate output for a 68040. This is the default when the compiler is
8627 configured for 68040-based systems.
8629 This option inhibits the use of 68881/68882 instructions that have to be
8630 emulated by software on the 68040. Use this option if your 68040 does not
8631 have code to emulate those instructions.
8635 Generate output for a 68060. This is the default when the compiler is
8636 configured for 68060-based systems.
8638 This option inhibits the use of 68020 and 68881/68882 instructions that
8639 have to be emulated by software on the 68060. Use this option if your 68060
8640 does not have code to emulate those instructions.
8644 Generate output for a CPU32. This is the default
8645 when the compiler is configured for CPU32-based systems.
8647 Use this option for microcontrollers with a
8648 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
8649 68336, 68340, 68341, 68349 and 68360.
8653 Generate output for a 520X ``coldfire'' family cpu. This is the default
8654 when the compiler is configured for 520X-based systems.
8656 Use this option for microcontroller with a 5200 core, including
8657 the MCF5202, MCF5203, MCF5204 and MCF5202.
8662 Generate output for a 68040, without using any of the new instructions.
8663 This results in code which can run relatively efficiently on either a
8664 68020/68881 or a 68030 or a 68040. The generated code does use the
8665 68881 instructions that are emulated on the 68040.
8669 Generate output for a 68060, without using any of the new instructions.
8670 This results in code which can run relatively efficiently on either a
8671 68020/68881 or a 68030 or a 68040. The generated code does use the
8672 68881 instructions that are emulated on the 68060.
8675 @opindex msoft-float
8676 Generate output containing library calls for floating point.
8677 @strong{Warning:} the requisite libraries are not available for all m68k
8678 targets. Normally the facilities of the machine's usual C compiler are
8679 used, but this can't be done directly in cross-compilation. You must
8680 make your own arrangements to provide suitable library functions for
8681 cross-compilation. The embedded targets @samp{m68k-*-aout} and
8682 @samp{m68k-*-coff} do provide software floating point support.
8686 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8687 Additionally, parameters passed on the stack are also aligned to a
8688 16-bit boundary even on targets whose API mandates promotion to 32-bit.
8691 @opindex mnobitfield
8692 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
8693 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
8697 Do use the bit-field instructions. The @option{-m68020} option implies
8698 @option{-mbitfield}. This is the default if you use a configuration
8699 designed for a 68020.
8703 Use a different function-calling convention, in which functions
8704 that take a fixed number of arguments return with the @code{rtd}
8705 instruction, which pops their arguments while returning. This
8706 saves one instruction in the caller since there is no need to pop
8707 the arguments there.
8709 This calling convention is incompatible with the one normally
8710 used on Unix, so you cannot use it if you need to call libraries
8711 compiled with the Unix compiler.
8713 Also, you must provide function prototypes for all functions that
8714 take variable numbers of arguments (including @code{printf});
8715 otherwise incorrect code will be generated for calls to those
8718 In addition, seriously incorrect code will result if you call a
8719 function with too many arguments. (Normally, extra arguments are
8720 harmlessly ignored.)
8722 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
8723 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
8726 @itemx -mno-align-int
8728 @opindex mno-align-int
8729 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
8730 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
8731 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
8732 Aligning variables on 32-bit boundaries produces code that runs somewhat
8733 faster on processors with 32-bit busses at the expense of more memory.
8735 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
8736 align structures containing the above types differently than
8737 most published application binary interface specifications for the m68k.
8741 Use the pc-relative addressing mode of the 68000 directly, instead of
8742 using a global offset table. At present, this option implies @option{-fpic},
8743 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
8744 not presently supported with @option{-mpcrel}, though this could be supported for
8745 68020 and higher processors.
8747 @item -mno-strict-align
8748 @itemx -mstrict-align
8749 @opindex mno-strict-align
8750 @opindex mstrict-align
8751 Do not (do) assume that unaligned memory references will be handled by
8755 Generate code that allows the data segment to be located in a different
8756 area of memory from the text segment. This allows for execute in place in
8757 an environment without virtual memory management. This option implies -fPIC.
8760 Generate code that assumes that the data segment follows the text segment.
8761 This is the default.
8763 @item -mid-shared-library
8764 Generate code that supports shared libraries via the library ID method.
8765 This allows for execute in place and shared libraries in an environment
8766 without virtual memory management. This option implies -fPIC.
8768 @item -mno-id-shared-library
8769 Generate code that doesn't assume ID based shared libraries are being used.
8770 This is the default.
8772 @item -mshared-library-id=n
8773 Specified the identification number of the ID based shared library being
8774 compiled. Specifying a value of 0 will generate more compact code, specifying
8775 other values will force the allocation of that number to the current
8776 library but is no more space or time efficient than omitting this option.
8780 @node M68hc1x Options
8781 @subsection M68hc1x Options
8782 @cindex M68hc1x options
8784 These are the @samp{-m} options defined for the 68hc11 and 68hc12
8785 microcontrollers. The default values for these options depends on
8786 which style of microcontroller was selected when the compiler was configured;
8787 the defaults for the most common choices are given below.
8794 Generate output for a 68HC11. This is the default
8795 when the compiler is configured for 68HC11-based systems.
8801 Generate output for a 68HC12. This is the default
8802 when the compiler is configured for 68HC12-based systems.
8808 Generate output for a 68HCS12.
8811 @opindex mauto-incdec
8812 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
8819 Enable the use of 68HC12 min and max instructions.
8822 @itemx -mno-long-calls
8823 @opindex mlong-calls
8824 @opindex mno-long-calls
8825 Treat all calls as being far away (near). If calls are assumed to be
8826 far away, the compiler will use the @code{call} instruction to
8827 call a function and the @code{rtc} instruction for returning.
8831 Consider type @code{int} to be 16 bits wide, like @code{short int}.
8833 @item -msoft-reg-count=@var{count}
8834 @opindex msoft-reg-count
8835 Specify the number of pseudo-soft registers which are used for the
8836 code generation. The maximum number is 32. Using more pseudo-soft
8837 register may or may not result in better code depending on the program.
8838 The default is 4 for 68HC11 and 2 for 68HC12.
8843 @subsection MCore Options
8844 @cindex MCore options
8846 These are the @samp{-m} options defined for the Motorola M*Core
8854 @opindex mno-hardlit
8855 Inline constants into the code stream if it can be done in two
8856 instructions or less.
8862 Use the divide instruction. (Enabled by default).
8864 @item -mrelax-immediate
8865 @itemx -mno-relax-immediate
8866 @opindex mrelax-immediate
8867 @opindex mno-relax-immediate
8868 Allow arbitrary sized immediates in bit operations.
8870 @item -mwide-bitfields
8871 @itemx -mno-wide-bitfields
8872 @opindex mwide-bitfields
8873 @opindex mno-wide-bitfields
8874 Always treat bit-fields as int-sized.
8876 @item -m4byte-functions
8877 @itemx -mno-4byte-functions
8878 @opindex m4byte-functions
8879 @opindex mno-4byte-functions
8880 Force all functions to be aligned to a four byte boundary.
8882 @item -mcallgraph-data
8883 @itemx -mno-callgraph-data
8884 @opindex mcallgraph-data
8885 @opindex mno-callgraph-data
8886 Emit callgraph information.
8889 @itemx -mno-slow-bytes
8890 @opindex mslow-bytes
8891 @opindex mno-slow-bytes
8892 Prefer word access when reading byte quantities.
8894 @item -mlittle-endian
8896 @opindex mlittle-endian
8897 @opindex mbig-endian
8898 Generate code for a little endian target.
8904 Generate code for the 210 processor.
8908 @subsection MIPS Options
8909 @cindex MIPS options
8915 Generate big-endian code.
8919 Generate little-endian code. This is the default for @samp{mips*el-*-*}
8922 @item -march=@var{arch}
8924 Generate code that will run on @var{arch}, which can be the name of a
8925 generic MIPS ISA, or the name of a particular processor.
8927 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
8928 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
8929 The processor names are:
8930 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
8932 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
8933 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
8937 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
8938 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
8939 The special value @samp{from-abi} selects the
8940 most compatible architecture for the selected ABI (that is,
8941 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
8943 In processor names, a final @samp{000} can be abbreviated as @samp{k}
8944 (for example, @samp{-march=r2k}). Prefixes are optional, and
8945 @samp{vr} may be written @samp{r}.
8947 GCC defines two macros based on the value of this option. The first
8948 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
8949 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
8950 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
8951 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
8952 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
8954 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
8955 above. In other words, it will have the full prefix and will not
8956 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
8957 the macro names the resolved architecture (either @samp{"mips1"} or
8958 @samp{"mips3"}). It names the default architecture when no
8959 @option{-march} option is given.
8961 @item -mtune=@var{arch}
8963 Optimize for @var{arch}. Among other things, this option controls
8964 the way instructions are scheduled, and the perceived cost of arithmetic
8965 operations. The list of @var{arch} values is the same as for
8968 When this option is not used, GCC will optimize for the processor
8969 specified by @option{-march}. By using @option{-march} and
8970 @option{-mtune} together, it is possible to generate code that will
8971 run on a family of processors, but optimize the code for one
8972 particular member of that family.
8974 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
8975 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
8976 @samp{-march} ones described above.
8980 Equivalent to @samp{-march=mips1}.
8984 Equivalent to @samp{-march=mips2}.
8988 Equivalent to @samp{-march=mips3}.
8992 Equivalent to @samp{-march=mips4}.
8996 Equivalent to @samp{-march=mips32}.
9000 Equivalent to @samp{-march=mips32r2}.
9004 Equivalent to @samp{-march=mips64}.
9010 Use (do not use) the MIPS16 ISA.
9022 Generate code for the given ABI@.
9024 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9025 generates 64-bit code when you select a 64-bit architecture, but you
9026 can use @option{-mgp32} to get 32-bit code instead.
9028 For information about the O64 ABI, see
9029 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9032 @itemx -mno-abicalls
9034 @opindex mno-abicalls
9035 Generate (do not generate) SVR4-style position-independent code.
9036 @option{-mabicalls} is the default for SVR4-based systems.
9042 Lift (do not lift) the usual restrictions on the size of the global
9045 GCC normally uses a single instruction to load values from the GOT.
9046 While this is relatively efficient, it will only work if the GOT
9047 is smaller than about 64k. Anything larger will cause the linker
9048 to report an error such as:
9050 @cindex relocation truncated to fit (MIPS)
9052 relocation truncated to fit: R_MIPS_GOT16 foobar
9055 If this happens, you should recompile your code with @option{-mxgot}.
9056 It should then work with very large GOTs, although it will also be
9057 less efficient, since it will take three instructions to fetch the
9058 value of a global symbol.
9060 Note that some linkers can create multiple GOTs. If you have such a
9061 linker, you should only need to use @option{-mxgot} when a single object
9062 file accesses more than 64k's worth of GOT entries. Very few do.
9064 These options have no effect unless GCC is generating position
9069 Assume that general-purpose registers are 32 bits wide.
9073 Assume that general-purpose registers are 64 bits wide.
9077 Assume that floating-point registers are 32 bits wide.
9081 Assume that floating-point registers are 64 bits wide.
9084 @opindex mhard-float
9085 Use floating-point coprocessor instructions.
9088 @opindex msoft-float
9089 Do not use floating-point coprocessor instructions. Implement
9090 floating-point calculations using library calls instead.
9092 @item -msingle-float
9093 @opindex msingle-float
9094 Assume that the floating-point coprocessor only supports single-precision
9097 @itemx -mdouble-float
9098 @opindex mdouble-float
9099 Assume that the floating-point coprocessor supports double-precision
9100 operations. This is the default.
9104 Force @code{int} and @code{long} types to be 64 bits wide. See
9105 @option{-mlong32} for an explanation of the default and the way
9106 that the pointer size is determined.
9110 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9111 an explanation of the default and the way that the pointer size is
9116 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9118 The default size of @code{int}s, @code{long}s and pointers depends on
9119 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9120 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9121 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9122 or the same size as integer registers, whichever is smaller.
9126 @cindex smaller data references (MIPS)
9127 @cindex gp-relative references (MIPS)
9128 Put global and static items less than or equal to @var{num} bytes into
9129 the small data or bss section instead of the normal data or bss section.
9130 This allows the data to be accessed using a single instruction.
9132 All modules should be compiled with the same @option{-G @var{num}}
9135 @item -membedded-data
9136 @itemx -mno-embedded-data
9137 @opindex membedded-data
9138 @opindex mno-embedded-data
9139 Allocate variables to the read-only data section first if possible, then
9140 next in the small data section if possible, otherwise in data. This gives
9141 slightly slower code than the default, but reduces the amount of RAM required
9142 when executing, and thus may be preferred for some embedded systems.
9144 @item -muninit-const-in-rodata
9145 @itemx -mno-uninit-const-in-rodata
9146 @opindex muninit-const-in-rodata
9147 @opindex mno-uninit-const-in-rodata
9148 Put uninitialized @code{const} variables in the read-only data section.
9149 This option is only meaningful in conjunction with @option{-membedded-data}.
9151 @item -msplit-addresses
9152 @itemx -mno-split-addresses
9153 @opindex msplit-addresses
9154 @opindex mno-split-addresses
9155 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9156 relocation operators. This option has been superceded by
9157 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9159 @item -mexplicit-relocs
9160 @itemx -mno-explicit-relocs
9161 @opindex mexplicit-relocs
9162 @opindex mno-explicit-relocs
9163 Use (do not use) assembler relocation operators when dealing with symbolic
9164 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9165 is to use assembler macros instead.
9167 @option{-mexplicit-relocs} is usually the default if GCC was configured
9168 to use an assembler that supports relocation operators. However, the
9169 combination of @option{-mabicalls} and @option{-fno-unit-at-a-time}
9170 implies @option{-mno-explicit-relocs} unless explicitly overridden.
9171 This is because, when generating abicalls, the choice of relocation
9172 depends on whether a symbol is local or global. In some rare cases,
9173 GCC will not be able to decide this until the whole compilation unit
9176 @item -mcheck-zero-division
9177 @itemx -mno-check-zero-division
9178 @opindex mcheck-zero-division
9179 @opindex mno-check-zero-division
9180 Trap (do not trap) on integer division by zero. The default is
9181 @option{-mcheck-zero-division}.
9187 Force (do not force) the use of @code{memcpy()} for non-trivial block
9188 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9189 most constant-sized copies.
9192 @itemx -mno-long-calls
9193 @opindex mlong-calls
9194 @opindex mno-long-calls
9195 Disable (do not disable) use of the @code{jal} instruction. Calling
9196 functions using @code{jal} is more efficient but requires the caller
9197 and callee to be in the same 256 megabyte segment.
9199 This option has no effect on abicalls code. The default is
9200 @option{-mno-long-calls}.
9206 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9207 instructions, as provided by the R4650 ISA.
9210 @itemx -mno-fused-madd
9211 @opindex mfused-madd
9212 @opindex mno-fused-madd
9213 Enable (disable) use of the floating point multiply-accumulate
9214 instructions, when they are available. The default is
9215 @option{-mfused-madd}.
9217 When multiply-accumulate instructions are used, the intermediate
9218 product is calculated to infinite precision and is not subject to
9219 the FCSR Flush to Zero bit. This may be undesirable in some
9224 Tell the MIPS assembler to not run its preprocessor over user
9225 assembler files (with a @samp{.s} suffix) when assembling them.
9228 @itemx -mno-fix-r4000
9230 @opindex mno-fix-r4000
9231 Work around certain R4000 CPU errata:
9234 A double-word or a variable shift may give an incorrect result if executed
9235 immediately after starting an integer division.
9237 A double-word or a variable shift may give an incorrect result if executed
9238 while an integer multiplication is in progress.
9240 An integer division may give an incorrect result if started in a delay slot
9241 of a taken branch or a jump.
9245 @itemx -mno-fix-r4400
9247 @opindex mno-fix-r4400
9248 Work around certain R4400 CPU errata:
9251 A double-word or a variable shift may give an incorrect result if executed
9252 immediately after starting an integer division.
9256 @itemx -mno-fix-vr4120
9257 @opindex mfix-vr4120
9258 Work around certain VR4120 errata:
9261 @code{dmultu} does not always produce the correct result.
9263 @code{div} and @code{ddiv} do not always produce the correct result if one
9264 of the operands is negative.
9266 The workarounds for the division errata rely on special functions in
9267 @file{libgcc.a}. At present, these functions are only provided by
9268 the @code{mips64vr*-elf} configurations.
9270 Other VR4120 errata require a nop to be inserted between certain pairs of
9271 instructions. These errata are handled by the assembler, not by GCC itself.
9276 Work around certain SB-1 CPU core errata.
9277 (This flag currently works around the SB-1 revision 2
9278 ``F1'' and ``F2'' floating point errata.)
9280 @item -mflush-func=@var{func}
9281 @itemx -mno-flush-func
9282 @opindex mflush-func
9283 Specifies the function to call to flush the I and D caches, or to not
9284 call any such function. If called, the function must take the same
9285 arguments as the common @code{_flush_func()}, that is, the address of the
9286 memory range for which the cache is being flushed, the size of the
9287 memory range, and the number 3 (to flush both caches). The default
9288 depends on the target GCC was configured for, but commonly is either
9289 @samp{_flush_func} or @samp{__cpu_flush}.
9291 @item -mbranch-likely
9292 @itemx -mno-branch-likely
9293 @opindex mbranch-likely
9294 @opindex mno-branch-likely
9295 Enable or disable use of Branch Likely instructions, regardless of the
9296 default for the selected architecture. By default, Branch Likely
9297 instructions may be generated if they are supported by the selected
9298 architecture. An exception is for the MIPS32 and MIPS64 architectures
9299 and processors which implement those architectures; for those, Branch
9300 Likely instructions will not be generated by default because the MIPS32
9301 and MIPS64 architectures specifically deprecate their use.
9303 @item -mfp-exceptions
9304 @itemx -mno-fp-exceptions
9305 @opindex mfp-exceptions
9306 Specifies whether FP exceptions are enabled. This affects how we schedule
9307 FP instructions for some processors. The default is that FP exceptions are
9310 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
9311 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
9314 @item -mvr4130-align
9315 @itemx -mno-vr4130-align
9316 @opindex mvr4130-align
9317 The VR4130 pipeline is two-way superscalar, but can only issue two
9318 instructions together if the first one is 8-byte aligned. When this
9319 option is enabled, GCC will align pairs of instructions that it
9320 thinks should execute in parallel.
9322 This option only has an effect when optimizing for the VR4130.
9323 It normally makes code faster, but at the expense of making it bigger.
9324 It is enabled by default at optimization level @option{-O3}.
9328 @subsection MMIX Options
9329 @cindex MMIX Options
9331 These options are defined for the MMIX:
9335 @itemx -mno-libfuncs
9337 @opindex mno-libfuncs
9338 Specify that intrinsic library functions are being compiled, passing all
9339 values in registers, no matter the size.
9344 @opindex mno-epsilon
9345 Generate floating-point comparison instructions that compare with respect
9346 to the @code{rE} epsilon register.
9348 @item -mabi=mmixware
9350 @opindex mabi-mmixware
9352 Generate code that passes function parameters and return values that (in
9353 the called function) are seen as registers @code{$0} and up, as opposed to
9354 the GNU ABI which uses global registers @code{$231} and up.
9357 @itemx -mno-zero-extend
9358 @opindex mzero-extend
9359 @opindex mno-zero-extend
9360 When reading data from memory in sizes shorter than 64 bits, use (do not
9361 use) zero-extending load instructions by default, rather than
9362 sign-extending ones.
9365 @itemx -mno-knuthdiv
9367 @opindex mno-knuthdiv
9368 Make the result of a division yielding a remainder have the same sign as
9369 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
9370 remainder follows the sign of the dividend. Both methods are
9371 arithmetically valid, the latter being almost exclusively used.
9373 @item -mtoplevel-symbols
9374 @itemx -mno-toplevel-symbols
9375 @opindex mtoplevel-symbols
9376 @opindex mno-toplevel-symbols
9377 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
9378 code can be used with the @code{PREFIX} assembly directive.
9382 Generate an executable in the ELF format, rather than the default
9383 @samp{mmo} format used by the @command{mmix} simulator.
9385 @item -mbranch-predict
9386 @itemx -mno-branch-predict
9387 @opindex mbranch-predict
9388 @opindex mno-branch-predict
9389 Use (do not use) the probable-branch instructions, when static branch
9390 prediction indicates a probable branch.
9392 @item -mbase-addresses
9393 @itemx -mno-base-addresses
9394 @opindex mbase-addresses
9395 @opindex mno-base-addresses
9396 Generate (do not generate) code that uses @emph{base addresses}. Using a
9397 base address automatically generates a request (handled by the assembler
9398 and the linker) for a constant to be set up in a global register. The
9399 register is used for one or more base address requests within the range 0
9400 to 255 from the value held in the register. The generally leads to short
9401 and fast code, but the number of different data items that can be
9402 addressed is limited. This means that a program that uses lots of static
9403 data may require @option{-mno-base-addresses}.
9406 @itemx -mno-single-exit
9407 @opindex msingle-exit
9408 @opindex mno-single-exit
9409 Force (do not force) generated code to have a single exit point in each
9413 @node MN10300 Options
9414 @subsection MN10300 Options
9415 @cindex MN10300 options
9417 These @option{-m} options are defined for Matsushita MN10300 architectures:
9422 Generate code to avoid bugs in the multiply instructions for the MN10300
9423 processors. This is the default.
9426 @opindex mno-mult-bug
9427 Do not generate code to avoid bugs in the multiply instructions for the
9432 Generate code which uses features specific to the AM33 processor.
9436 Do not generate code which uses features specific to the AM33 processor. This
9441 Do not link in the C run-time initialization object file.
9445 Indicate to the linker that it should perform a relaxation optimization pass
9446 to shorten branches, calls and absolute memory addresses. This option only
9447 has an effect when used on the command line for the final link step.
9449 This option makes symbolic debugging impossible.
9453 @subsection NS32K Options
9454 @cindex NS32K options
9456 These are the @samp{-m} options defined for the 32000 series. The default
9457 values for these options depends on which style of 32000 was selected when
9458 the compiler was configured; the defaults for the most common choices are
9466 Generate output for a 32032. This is the default
9467 when the compiler is configured for 32032 and 32016 based systems.
9473 Generate output for a 32332. This is the default
9474 when the compiler is configured for 32332-based systems.
9480 Generate output for a 32532. This is the default
9481 when the compiler is configured for 32532-based systems.
9485 Generate output containing 32081 instructions for floating point.
9486 This is the default for all systems.
9490 Generate output containing 32381 instructions for floating point. This
9491 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9492 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9496 Try and generate multiply-add floating point instructions @code{polyF}
9497 and @code{dotF}. This option is only available if the @option{-m32381}
9498 option is in effect. Using these instructions requires changes to
9499 register allocation which generally has a negative impact on
9500 performance. This option should only be enabled when compiling code
9501 particularly likely to make heavy use of multiply-add instructions.
9504 @opindex mnomulti-add
9505 Do not try and generate multiply-add floating point instructions
9506 @code{polyF} and @code{dotF}. This is the default on all platforms.
9509 @opindex msoft-float
9510 Generate output containing library calls for floating point.
9511 @strong{Warning:} the requisite libraries may not be available.
9513 @item -mieee-compare
9514 @itemx -mno-ieee-compare
9515 @opindex mieee-compare
9516 @opindex mno-ieee-compare
9517 Control whether or not the compiler uses IEEE floating point
9518 comparisons. These handle correctly the case where the result of a
9519 comparison is unordered.
9520 @strong{Warning:} the requisite kernel support may not be available.
9523 @opindex mnobitfield
9524 Do not use the bit-field instructions. On some machines it is faster to
9525 use shifting and masking operations. This is the default for the pc532.
9529 Do use the bit-field instructions. This is the default for all platforms
9534 Use a different function-calling convention, in which functions
9535 that take a fixed number of arguments return pop their
9536 arguments on return with the @code{ret} instruction.
9538 This calling convention is incompatible with the one normally
9539 used on Unix, so you cannot use it if you need to call libraries
9540 compiled with the Unix compiler.
9542 Also, you must provide function prototypes for all functions that
9543 take variable numbers of arguments (including @code{printf});
9544 otherwise incorrect code will be generated for calls to those
9547 In addition, seriously incorrect code will result if you call a
9548 function with too many arguments. (Normally, extra arguments are
9549 harmlessly ignored.)
9551 This option takes its name from the 680x0 @code{rtd} instruction.
9556 Use a different function-calling convention where the first two arguments
9557 are passed in registers.
9559 This calling convention is incompatible with the one normally
9560 used on Unix, so you cannot use it if you need to call libraries
9561 compiled with the Unix compiler.
9564 @opindex mnoregparam
9565 Do not pass any arguments in registers. This is the default for all
9570 It is OK to use the sb as an index register which is always loaded with
9571 zero. This is the default for the pc532-netbsd target.
9575 The sb register is not available for use or has not been initialized to
9576 zero by the run time system. This is the default for all targets except
9577 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
9578 @option{-fpic} is set.
9582 Many ns32000 series addressing modes use displacements of up to 512MB@.
9583 If an address is above 512MB then displacements from zero can not be used.
9584 This option causes code to be generated which can be loaded above 512MB@.
9585 This may be useful for operating systems or ROM code.
9589 Assume code will be loaded in the first 512MB of virtual address space.
9590 This is the default for all platforms.
9594 @node PDP-11 Options
9595 @subsection PDP-11 Options
9596 @cindex PDP-11 Options
9598 These options are defined for the PDP-11:
9603 Use hardware FPP floating point. This is the default. (FIS floating
9604 point on the PDP-11/40 is not supported.)
9607 @opindex msoft-float
9608 Do not use hardware floating point.
9612 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
9616 Return floating-point results in memory. This is the default.
9620 Generate code for a PDP-11/40.
9624 Generate code for a PDP-11/45. This is the default.
9628 Generate code for a PDP-11/10.
9630 @item -mbcopy-builtin
9631 @opindex bcopy-builtin
9632 Use inline @code{movmemhi} patterns for copying memory. This is the
9637 Do not use inline @code{movmemhi} patterns for copying memory.
9643 Use 16-bit @code{int}. This is the default.
9649 Use 32-bit @code{int}.
9654 @opindex mno-float32
9655 Use 64-bit @code{float}. This is the default.
9660 @opindex mno-float64
9661 Use 32-bit @code{float}.
9665 Use @code{abshi2} pattern. This is the default.
9669 Do not use @code{abshi2} pattern.
9671 @item -mbranch-expensive
9672 @opindex mbranch-expensive
9673 Pretend that branches are expensive. This is for experimenting with
9674 code generation only.
9676 @item -mbranch-cheap
9677 @opindex mbranch-cheap
9678 Do not pretend that branches are expensive. This is the default.
9682 Generate code for a system with split I&D.
9686 Generate code for a system without split I&D. This is the default.
9690 Use Unix assembler syntax. This is the default when configured for
9695 Use DEC assembler syntax. This is the default when configured for any
9696 PDP-11 target other than @samp{pdp11-*-bsd}.
9699 @node PowerPC Options
9700 @subsection PowerPC Options
9701 @cindex PowerPC options
9703 These are listed under @xref{RS/6000 and PowerPC Options}.
9705 @node RS/6000 and PowerPC Options
9706 @subsection IBM RS/6000 and PowerPC Options
9707 @cindex RS/6000 and PowerPC Options
9708 @cindex IBM RS/6000 and PowerPC Options
9710 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
9718 @itemx -mpowerpc-gpopt
9719 @itemx -mno-powerpc-gpopt
9720 @itemx -mpowerpc-gfxopt
9721 @itemx -mno-powerpc-gfxopt
9723 @itemx -mno-powerpc64
9729 @opindex mno-powerpc
9730 @opindex mpowerpc-gpopt
9731 @opindex mno-powerpc-gpopt
9732 @opindex mpowerpc-gfxopt
9733 @opindex mno-powerpc-gfxopt
9735 @opindex mno-powerpc64
9736 GCC supports two related instruction set architectures for the
9737 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
9738 instructions supported by the @samp{rios} chip set used in the original
9739 RS/6000 systems and the @dfn{PowerPC} instruction set is the
9740 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
9741 the IBM 4xx microprocessors.
9743 Neither architecture is a subset of the other. However there is a
9744 large common subset of instructions supported by both. An MQ
9745 register is included in processors supporting the POWER architecture.
9747 You use these options to specify which instructions are available on the
9748 processor you are using. The default value of these options is
9749 determined when configuring GCC@. Specifying the
9750 @option{-mcpu=@var{cpu_type}} overrides the specification of these
9751 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
9752 rather than the options listed above.
9754 The @option{-mpower} option allows GCC to generate instructions that
9755 are found only in the POWER architecture and to use the MQ register.
9756 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
9757 to generate instructions that are present in the POWER2 architecture but
9758 not the original POWER architecture.
9760 The @option{-mpowerpc} option allows GCC to generate instructions that
9761 are found only in the 32-bit subset of the PowerPC architecture.
9762 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
9763 GCC to use the optional PowerPC architecture instructions in the
9764 General Purpose group, including floating-point square root. Specifying
9765 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
9766 use the optional PowerPC architecture instructions in the Graphics
9767 group, including floating-point select.
9769 The @option{-mpowerpc64} option allows GCC to generate the additional
9770 64-bit instructions that are found in the full PowerPC64 architecture
9771 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
9772 @option{-mno-powerpc64}.
9774 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
9775 will use only the instructions in the common subset of both
9776 architectures plus some special AIX common-mode calls, and will not use
9777 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
9778 permits GCC to use any instruction from either architecture and to
9779 allow use of the MQ register; specify this for the Motorola MPC601.
9781 @item -mnew-mnemonics
9782 @itemx -mold-mnemonics
9783 @opindex mnew-mnemonics
9784 @opindex mold-mnemonics
9785 Select which mnemonics to use in the generated assembler code. With
9786 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
9787 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
9788 assembler mnemonics defined for the POWER architecture. Instructions
9789 defined in only one architecture have only one mnemonic; GCC uses that
9790 mnemonic irrespective of which of these options is specified.
9792 GCC defaults to the mnemonics appropriate for the architecture in
9793 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
9794 value of these option. Unless you are building a cross-compiler, you
9795 should normally not specify either @option{-mnew-mnemonics} or
9796 @option{-mold-mnemonics}, but should instead accept the default.
9798 @item -mcpu=@var{cpu_type}
9800 Set architecture type, register usage, choice of mnemonics, and
9801 instruction scheduling parameters for machine type @var{cpu_type}.
9802 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
9803 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
9804 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
9805 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
9806 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
9807 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
9808 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
9809 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
9810 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
9812 @option{-mcpu=common} selects a completely generic processor. Code
9813 generated under this option will run on any POWER or PowerPC processor.
9814 GCC will use only the instructions in the common subset of both
9815 architectures, and will not use the MQ register. GCC assumes a generic
9816 processor model for scheduling purposes.
9818 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
9819 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
9820 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
9821 types, with an appropriate, generic processor model assumed for
9822 scheduling purposes.
9824 The other options specify a specific processor. Code generated under
9825 those options will run best on that processor, and may not run at all on
9828 The @option{-mcpu} options automatically enable or disable the
9829 following options: @option{-maltivec}, @option{-mhard-float},
9830 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
9831 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
9832 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
9833 @option{-mstring}. The particular options set for any particular CPU
9834 will vary between compiler versions, depending on what setting seems
9835 to produce optimal code for that CPU; it doesn't necessarily reflect
9836 the actual hardware's capabilities. If you wish to set an individual
9837 option to a particular value, you may specify it after the
9838 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
9840 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
9841 not enabled or disabled by the @option{-mcpu} option at present, since
9842 AIX does not have full support for these options. You may still
9843 enable or disable them individually if you're sure it'll work in your
9846 @item -mtune=@var{cpu_type}
9848 Set the instruction scheduling parameters for machine type
9849 @var{cpu_type}, but do not set the architecture type, register usage, or
9850 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
9851 values for @var{cpu_type} are used for @option{-mtune} as for
9852 @option{-mcpu}. If both are specified, the code generated will use the
9853 architecture, registers, and mnemonics set by @option{-mcpu}, but the
9854 scheduling parameters set by @option{-mtune}.
9859 @opindex mno-altivec
9860 These switches enable or disable the use of built-in functions that
9861 allow access to the AltiVec instruction set. You may also need to set
9862 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
9867 Extend the current ABI with SPE ABI extensions. This does not change
9868 the default ABI, instead it adds the SPE ABI extensions to the current
9872 @opindex mabi=no-spe
9873 Disable Booke SPE ABI extensions for the current ABI.
9875 @item -misel=@var{yes/no}
9878 This switch enables or disables the generation of ISEL instructions.
9880 @item -mspe=@var{yes/no}
9883 This switch enables or disables the generation of SPE simd
9886 @item -mfloat-gprs=@var{yes/no}
9888 @opindex mfloat-gprs
9889 This switch enables or disables the generation of floating point
9890 operations on the general purpose registers for architectures that
9891 support it. This option is currently only available on the MPC8540.
9894 @itemx -mno-fp-in-toc
9895 @itemx -mno-sum-in-toc
9896 @itemx -mminimal-toc
9898 @opindex mno-fp-in-toc
9899 @opindex mno-sum-in-toc
9900 @opindex mminimal-toc
9901 Modify generation of the TOC (Table Of Contents), which is created for
9902 every executable file. The @option{-mfull-toc} option is selected by
9903 default. In that case, GCC will allocate at least one TOC entry for
9904 each unique non-automatic variable reference in your program. GCC
9905 will also place floating-point constants in the TOC@. However, only
9906 16,384 entries are available in the TOC@.
9908 If you receive a linker error message that saying you have overflowed
9909 the available TOC space, you can reduce the amount of TOC space used
9910 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
9911 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
9912 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
9913 generate code to calculate the sum of an address and a constant at
9914 run-time instead of putting that sum into the TOC@. You may specify one
9915 or both of these options. Each causes GCC to produce very slightly
9916 slower and larger code at the expense of conserving TOC space.
9918 If you still run out of space in the TOC even when you specify both of
9919 these options, specify @option{-mminimal-toc} instead. This option causes
9920 GCC to make only one TOC entry for every file. When you specify this
9921 option, GCC will produce code that is slower and larger but which
9922 uses extremely little TOC space. You may wish to use this option
9923 only on files that contain less frequently executed code.
9929 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
9930 @code{long} type, and the infrastructure needed to support them.
9931 Specifying @option{-maix64} implies @option{-mpowerpc64} and
9932 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
9933 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
9938 @opindex mno-xl-call
9939 On AIX, pass floating-point arguments to prototyped functions beyond the
9940 register save area (RSA) on the stack in addition to argument FPRs. The
9941 AIX calling convention was extended but not initially documented to
9942 handle an obscure K&R C case of calling a function that takes the
9943 address of its arguments with fewer arguments than declared. AIX XL
9944 compilers access floating point arguments which do not fit in the
9945 RSA from the stack when a subroutine is compiled without
9946 optimization. Because always storing floating-point arguments on the
9947 stack is inefficient and rarely needed, this option is not enabled by
9948 default and only is necessary when calling subroutines compiled by AIX
9949 XL compilers without optimization.
9953 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
9954 application written to use message passing with special startup code to
9955 enable the application to run. The system must have PE installed in the
9956 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
9957 must be overridden with the @option{-specs=} option to specify the
9958 appropriate directory location. The Parallel Environment does not
9959 support threads, so the @option{-mpe} option and the @option{-pthread}
9960 option are incompatible.
9962 @item -malign-natural
9963 @itemx -malign-power
9964 @opindex malign-natural
9965 @opindex malign-power
9966 On AIX, Darwin, and 64-bit PowerPC GNU/Linux, the option
9967 @option{-malign-natural} overrides the ABI-defined alignment of larger
9968 types, such as floating-point doubles, on their natural size-based boundary.
9969 The option @option{-malign-power} instructs GCC to follow the ABI-specified
9970 alignment rules. GCC defaults to the standard alignment defined in the ABI.
9974 @opindex msoft-float
9975 @opindex mhard-float
9976 Generate code that does not use (uses) the floating-point register set.
9977 Software floating point emulation is provided if you use the
9978 @option{-msoft-float} option, and pass the option to GCC when linking.
9981 @itemx -mno-multiple
9983 @opindex mno-multiple
9984 Generate code that uses (does not use) the load multiple word
9985 instructions and the store multiple word instructions. These
9986 instructions are generated by default on POWER systems, and not
9987 generated on PowerPC systems. Do not use @option{-mmultiple} on little
9988 endian PowerPC systems, since those instructions do not work when the
9989 processor is in little endian mode. The exceptions are PPC740 and
9990 PPC750 which permit the instructions usage in little endian mode.
9996 Generate code that uses (does not use) the load string instructions
9997 and the store string word instructions to save multiple registers and
9998 do small block moves. These instructions are generated by default on
9999 POWER systems, and not generated on PowerPC systems. Do not use
10000 @option{-mstring} on little endian PowerPC systems, since those
10001 instructions do not work when the processor is in little endian mode.
10002 The exceptions are PPC740 and PPC750 which permit the instructions
10003 usage in little endian mode.
10008 @opindex mno-update
10009 Generate code that uses (does not use) the load or store instructions
10010 that update the base register to the address of the calculated memory
10011 location. These instructions are generated by default. If you use
10012 @option{-mno-update}, there is a small window between the time that the
10013 stack pointer is updated and the address of the previous frame is
10014 stored, which means code that walks the stack frame across interrupts or
10015 signals may get corrupted data.
10018 @itemx -mno-fused-madd
10019 @opindex mfused-madd
10020 @opindex mno-fused-madd
10021 Generate code that uses (does not use) the floating point multiply and
10022 accumulate instructions. These instructions are generated by default if
10023 hardware floating is used.
10025 @item -mno-bit-align
10027 @opindex mno-bit-align
10028 @opindex mbit-align
10029 On System V.4 and embedded PowerPC systems do not (do) force structures
10030 and unions that contain bit-fields to be aligned to the base type of the
10033 For example, by default a structure containing nothing but 8
10034 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10035 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10036 the structure would be aligned to a 1 byte boundary and be one byte in
10039 @item -mno-strict-align
10040 @itemx -mstrict-align
10041 @opindex mno-strict-align
10042 @opindex mstrict-align
10043 On System V.4 and embedded PowerPC systems do not (do) assume that
10044 unaligned memory references will be handled by the system.
10046 @item -mrelocatable
10047 @itemx -mno-relocatable
10048 @opindex mrelocatable
10049 @opindex mno-relocatable
10050 On embedded PowerPC systems generate code that allows (does not allow)
10051 the program to be relocated to a different address at runtime. If you
10052 use @option{-mrelocatable} on any module, all objects linked together must
10053 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10055 @item -mrelocatable-lib
10056 @itemx -mno-relocatable-lib
10057 @opindex mrelocatable-lib
10058 @opindex mno-relocatable-lib
10059 On embedded PowerPC systems generate code that allows (does not allow)
10060 the program to be relocated to a different address at runtime. Modules
10061 compiled with @option{-mrelocatable-lib} can be linked with either modules
10062 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10063 with modules compiled with the @option{-mrelocatable} options.
10069 On System V.4 and embedded PowerPC systems do not (do) assume that
10070 register 2 contains a pointer to a global area pointing to the addresses
10071 used in the program.
10074 @itemx -mlittle-endian
10076 @opindex mlittle-endian
10077 On System V.4 and embedded PowerPC systems compile code for the
10078 processor in little endian mode. The @option{-mlittle-endian} option is
10079 the same as @option{-mlittle}.
10082 @itemx -mbig-endian
10084 @opindex mbig-endian
10085 On System V.4 and embedded PowerPC systems compile code for the
10086 processor in big endian mode. The @option{-mbig-endian} option is
10087 the same as @option{-mbig}.
10089 @item -mdynamic-no-pic
10090 @opindex mdynamic-no-pic
10091 On Darwin and Mac OS X systems, compile code so that it is not
10092 relocatable, but that its external references are relocatable. The
10093 resulting code is suitable for applications, but not shared
10096 @item -mprioritize-restricted-insns=@var{priority}
10097 @opindex mprioritize-restricted-insns
10098 This option controls the priority that is assigned to
10099 dispatch-slot restricted instructions during the second scheduling
10100 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10101 @var{no/highest/second-highest} priority to dispatch slot restricted
10104 @item -msched-costly-dep=@var{dependence_type}
10105 @opindex msched-costly-dep
10106 This option controls which dependences are considered costly
10107 by the target during instruction scheduling. The argument
10108 @var{dependence_type} takes one of the following values:
10109 @var{no}: no dependence is costly,
10110 @var{all}: all dependences are costly,
10111 @var{true_store_to_load}: a true dependence from store to load is costly,
10112 @var{store_to_load}: any dependence from store to load is costly,
10113 @var{number}: any dependence which latency >= @var{number} is costly.
10115 @item -minsert-sched-nops=@var{scheme}
10116 @opindex minsert-sched-nops
10117 This option controls which nop insertion scheme will be used during
10118 the second scheduling pass. The argument @var{scheme} takes one of the
10120 @var{no}: Don't insert nops.
10121 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10122 according to the scheduler's grouping.
10123 @var{regroup_exact}: Insert nops to force costly dependent insns into
10124 separate groups. Insert exactly as many nops as needed to force an insn
10125 to a new group, according to the estimated processor grouping.
10126 @var{number}: Insert nops to force costly dependent insns into
10127 separate groups. Insert @var{number} nops to force an insn to a new group.
10130 @opindex mcall-sysv
10131 On System V.4 and embedded PowerPC systems compile code using calling
10132 conventions that adheres to the March 1995 draft of the System V
10133 Application Binary Interface, PowerPC processor supplement. This is the
10134 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10136 @item -mcall-sysv-eabi
10137 @opindex mcall-sysv-eabi
10138 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10140 @item -mcall-sysv-noeabi
10141 @opindex mcall-sysv-noeabi
10142 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10144 @item -mcall-solaris
10145 @opindex mcall-solaris
10146 On System V.4 and embedded PowerPC systems compile code for the Solaris
10150 @opindex mcall-linux
10151 On System V.4 and embedded PowerPC systems compile code for the
10152 Linux-based GNU system.
10156 On System V.4 and embedded PowerPC systems compile code for the
10157 Hurd-based GNU system.
10159 @item -mcall-netbsd
10160 @opindex mcall-netbsd
10161 On System V.4 and embedded PowerPC systems compile code for the
10162 NetBSD operating system.
10164 @item -maix-struct-return
10165 @opindex maix-struct-return
10166 Return all structures in memory (as specified by the AIX ABI)@.
10168 @item -msvr4-struct-return
10169 @opindex msvr4-struct-return
10170 Return structures smaller than 8 bytes in registers (as specified by the
10173 @item -mabi=altivec
10174 @opindex mabi=altivec
10175 Extend the current ABI with AltiVec ABI extensions. This does not
10176 change the default ABI, instead it adds the AltiVec ABI extensions to
10179 @item -mabi=no-altivec
10180 @opindex mabi=no-altivec
10181 Disable AltiVec ABI extensions for the current ABI.
10184 @itemx -mno-prototype
10185 @opindex mprototype
10186 @opindex mno-prototype
10187 On System V.4 and embedded PowerPC systems assume that all calls to
10188 variable argument functions are properly prototyped. Otherwise, the
10189 compiler must insert an instruction before every non prototyped call to
10190 set or clear bit 6 of the condition code register (@var{CR}) to
10191 indicate whether floating point values were passed in the floating point
10192 registers in case the function takes a variable arguments. With
10193 @option{-mprototype}, only calls to prototyped variable argument functions
10194 will set or clear the bit.
10198 On embedded PowerPC systems, assume that the startup module is called
10199 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10200 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10205 On embedded PowerPC systems, assume that the startup module is called
10206 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10211 On embedded PowerPC systems, assume that the startup module is called
10212 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10215 @item -myellowknife
10216 @opindex myellowknife
10217 On embedded PowerPC systems, assume that the startup module is called
10218 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10223 On System V.4 and embedded PowerPC systems, specify that you are
10224 compiling for a VxWorks system.
10228 Specify that you are compiling for the WindISS simulation environment.
10232 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10233 header to indicate that @samp{eabi} extended relocations are used.
10239 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10240 Embedded Applications Binary Interface (eabi) which is a set of
10241 modifications to the System V.4 specifications. Selecting @option{-meabi}
10242 means that the stack is aligned to an 8 byte boundary, a function
10243 @code{__eabi} is called to from @code{main} to set up the eabi
10244 environment, and the @option{-msdata} option can use both @code{r2} and
10245 @code{r13} to point to two separate small data areas. Selecting
10246 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10247 do not call an initialization function from @code{main}, and the
10248 @option{-msdata} option will only use @code{r13} to point to a single
10249 small data area. The @option{-meabi} option is on by default if you
10250 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10253 @opindex msdata=eabi
10254 On System V.4 and embedded PowerPC systems, put small initialized
10255 @code{const} global and static data in the @samp{.sdata2} section, which
10256 is pointed to by register @code{r2}. Put small initialized
10257 non-@code{const} global and static data in the @samp{.sdata} section,
10258 which is pointed to by register @code{r13}. Put small uninitialized
10259 global and static data in the @samp{.sbss} section, which is adjacent to
10260 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10261 incompatible with the @option{-mrelocatable} option. The
10262 @option{-msdata=eabi} option also sets the @option{-memb} option.
10265 @opindex msdata=sysv
10266 On System V.4 and embedded PowerPC systems, put small global and static
10267 data in the @samp{.sdata} section, which is pointed to by register
10268 @code{r13}. Put small uninitialized global and static data in the
10269 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
10270 The @option{-msdata=sysv} option is incompatible with the
10271 @option{-mrelocatable} option.
10273 @item -msdata=default
10275 @opindex msdata=default
10277 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
10278 compile code the same as @option{-msdata=eabi}, otherwise compile code the
10279 same as @option{-msdata=sysv}.
10282 @opindex msdata-data
10283 On System V.4 and embedded PowerPC systems, put small global and static
10284 data in the @samp{.sdata} section. Put small uninitialized global and
10285 static data in the @samp{.sbss} section. Do not use register @code{r13}
10286 to address small data however. This is the default behavior unless
10287 other @option{-msdata} options are used.
10291 @opindex msdata=none
10293 On embedded PowerPC systems, put all initialized global and static data
10294 in the @samp{.data} section, and all uninitialized data in the
10295 @samp{.bss} section.
10299 @cindex smaller data references (PowerPC)
10300 @cindex .sdata/.sdata2 references (PowerPC)
10301 On embedded PowerPC systems, put global and static items less than or
10302 equal to @var{num} bytes into the small data or bss sections instead of
10303 the normal data or bss section. By default, @var{num} is 8. The
10304 @option{-G @var{num}} switch is also passed to the linker.
10305 All modules should be compiled with the same @option{-G @var{num}} value.
10308 @itemx -mno-regnames
10310 @opindex mno-regnames
10311 On System V.4 and embedded PowerPC systems do (do not) emit register
10312 names in the assembly language output using symbolic forms.
10315 @itemx -mno-longcall
10317 @opindex mno-longcall
10318 Default to making all function calls indirectly, using a register, so
10319 that functions which reside further than 32 megabytes (33,554,432
10320 bytes) from the current location can be called. This setting can be
10321 overridden by the @code{shortcall} function attribute, or by
10322 @code{#pragma longcall(0)}.
10324 Some linkers are capable of detecting out-of-range calls and generating
10325 glue code on the fly. On these systems, long calls are unnecessary and
10326 generate slower code. As of this writing, the AIX linker can do this,
10327 as can the GNU linker for PowerPC/64. It is planned to add this feature
10328 to the GNU linker for 32-bit PowerPC systems as well.
10330 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
10331 callee, L42'', plus a ``branch island'' (glue code). The two target
10332 addresses represent the callee and the ``branch island.'' The
10333 Darwin/PPC linker will prefer the first address and generate a ``bl
10334 callee'' if the PPC ``bl'' instruction will reach the callee directly;
10335 otherwise, the linker will generate ``bl L42'' to call the ``branch
10336 island.'' The ``branch island'' is appended to the body of the
10337 calling function; it computes the full 32-bit address of the callee
10340 On Mach-O (Darwin) systems, this option directs the compiler emit to
10341 the glue for every direct call, and the Darwin linker decides whether
10342 to use or discard it.
10344 In the future, we may cause GCC to ignore all longcall specifications
10345 when the linker is known to generate glue.
10349 Adds support for multithreading with the @dfn{pthreads} library.
10350 This option sets flags for both the preprocessor and linker.
10354 @node S/390 and zSeries Options
10355 @subsection S/390 and zSeries Options
10356 @cindex S/390 and zSeries Options
10358 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10362 @itemx -msoft-float
10363 @opindex mhard-float
10364 @opindex msoft-float
10365 Use (do not use) the hardware floating-point instructions and registers
10366 for floating-point operations. When @option{-msoft-float} is specified,
10367 functions in @file{libgcc.a} will be used to perform floating-point
10368 operations. When @option{-mhard-float} is specified, the compiler
10369 generates IEEE floating-point instructions. This is the default.
10372 @itemx -mno-backchain
10373 @opindex mbackchain
10374 @opindex mno-backchain
10375 Generate (or do not generate) code which maintains an explicit
10376 backchain within the stack frame that points to the caller's frame.
10377 This may be needed to allow debugging using tools that do not understand
10378 DWARF-2 call frame information. The default is not to generate the
10382 @itemx -mno-small-exec
10383 @opindex msmall-exec
10384 @opindex mno-small-exec
10385 Generate (or do not generate) code using the @code{bras} instruction
10386 to do subroutine calls.
10387 This only works reliably if the total executable size does not
10388 exceed 64k. The default is to use the @code{basr} instruction instead,
10389 which does not have this limitation.
10395 When @option{-m31} is specified, generate code compliant to the
10396 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
10397 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
10398 particular to generate 64-bit instructions. For the @samp{s390}
10399 targets, the default is @option{-m31}, while the @samp{s390x}
10400 targets default to @option{-m64}.
10406 When @option{-mzarch} is specified, generate code using the
10407 instructions available on z/Architecture.
10408 When @option{-mesa} is specified, generate code using the
10409 instructions available on ESA/390. Note that @option{-mesa} is
10410 not possible with @option{-m64}.
10411 When generating code compliant to the GNU/Linux for S/390 ABI,
10412 the default is @option{-mesa}. When generating code compliant
10413 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
10419 Generate (or do not generate) code using the @code{mvcle} instruction
10420 to perform block moves. When @option{-mno-mvcle} is specified,
10421 use a @code{mvc} loop instead. This is the default.
10427 Print (or do not print) additional debug information when compiling.
10428 The default is to not print debug information.
10430 @item -march=@var{cpu-type}
10432 Generate code that will run on @var{cpu-type}, which is the name of a system
10433 representing a certain processor type. Possible values for
10434 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10435 When generating code using the instructions available on z/Architecture,
10436 the default is @option{-march=z900}. Otherwise, the default is
10437 @option{-march=g5}.
10439 @item -mtune=@var{cpu-type}
10441 Tune to @var{cpu-type} everything applicable about the generated code,
10442 except for the ABI and the set of available instructions.
10443 The list of @var{cpu-type} values is the same as for @option{-march}.
10444 The default is the value used for @option{-march}.
10447 @itemx -mno-tpf-trace
10448 @opindex mtpf-trace
10449 @opindex mno-tpf-trace
10450 Generate code that adds (does not add) in TPF OS specific branches to trace
10451 routines in the operating system. This option is off by default, even
10452 when compiling for the TPF OS.
10455 @itemx -mno-fused-madd
10456 @opindex mfused-madd
10457 @opindex mno-fused-madd
10458 Generate code that uses (does not use) the floating point multiply and
10459 accumulate instructions. These instructions are generated by default if
10460 hardware floating point is used.
10464 @subsection SH Options
10466 These @samp{-m} options are defined for the SH implementations:
10471 Generate code for the SH1.
10475 Generate code for the SH2.
10478 Generate code for the SH2e.
10482 Generate code for the SH3.
10486 Generate code for the SH3e.
10490 Generate code for the SH4 without a floating-point unit.
10492 @item -m4-single-only
10493 @opindex m4-single-only
10494 Generate code for the SH4 with a floating-point unit that only
10495 supports single-precision arithmetic.
10499 Generate code for the SH4 assuming the floating-point unit is in
10500 single-precision mode by default.
10504 Generate code for the SH4.
10508 Compile code for the processor in big endian mode.
10512 Compile code for the processor in little endian mode.
10516 Align doubles at 64-bit boundaries. Note that this changes the calling
10517 conventions, and thus some functions from the standard C library will
10518 not work unless you recompile it first with @option{-mdalign}.
10522 Shorten some address references at link time, when possible; uses the
10523 linker option @option{-relax}.
10527 Use 32-bit offsets in @code{switch} tables. The default is to use
10532 Enable the use of the instruction @code{fmovd}.
10536 Comply with the calling conventions defined by Renesas.
10539 @opindex mnomacsave
10540 Mark the @code{MAC} register as call-clobbered, even if
10541 @option{-mhitachi} is given.
10545 Increase IEEE-compliance of floating-point code.
10549 Dump instruction size and location in the assembly code.
10552 @opindex mpadstruct
10553 This option is deprecated. It pads structures to multiple of 4 bytes,
10554 which is incompatible with the SH ABI@.
10558 Optimize for space instead of speed. Implied by @option{-Os}.
10561 @opindex mprefergot
10562 When generating position-independent code, emit function calls using
10563 the Global Offset Table instead of the Procedure Linkage Table.
10567 Generate a library function call to invalidate instruction cache
10568 entries, after fixing up a trampoline. This library function call
10569 doesn't assume it can write to the whole memory address space. This
10570 is the default when the target is @code{sh-*-linux*}.
10573 @node SPARC Options
10574 @subsection SPARC Options
10575 @cindex SPARC options
10577 These @samp{-m} options are supported on the SPARC:
10580 @item -mno-app-regs
10582 @opindex mno-app-regs
10584 Specify @option{-mapp-regs} to generate output using the global registers
10585 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
10588 To be fully SVR4 ABI compliant at the cost of some performance loss,
10589 specify @option{-mno-app-regs}. You should compile libraries and system
10590 software with this option.
10593 @itemx -mhard-float
10595 @opindex mhard-float
10596 Generate output containing floating point instructions. This is the
10600 @itemx -msoft-float
10602 @opindex msoft-float
10603 Generate output containing library calls for floating point.
10604 @strong{Warning:} the requisite libraries are not available for all SPARC
10605 targets. Normally the facilities of the machine's usual C compiler are
10606 used, but this cannot be done directly in cross-compilation. You must make
10607 your own arrangements to provide suitable library functions for
10608 cross-compilation. The embedded targets @samp{sparc-*-aout} and
10609 @samp{sparclite-*-*} do provide software floating point support.
10611 @option{-msoft-float} changes the calling convention in the output file;
10612 therefore, it is only useful if you compile @emph{all} of a program with
10613 this option. In particular, you need to compile @file{libgcc.a}, the
10614 library that comes with GCC, with @option{-msoft-float} in order for
10617 @item -mhard-quad-float
10618 @opindex mhard-quad-float
10619 Generate output containing quad-word (long double) floating point
10622 @item -msoft-quad-float
10623 @opindex msoft-quad-float
10624 Generate output containing library calls for quad-word (long double)
10625 floating point instructions. The functions called are those specified
10626 in the SPARC ABI@. This is the default.
10628 As of this writing, there are no SPARC implementations that have hardware
10629 support for the quad-word floating point instructions. They all invoke
10630 a trap handler for one of these instructions, and then the trap handler
10631 emulates the effect of the instruction. Because of the trap handler overhead,
10632 this is much slower than calling the ABI library routines. Thus the
10633 @option{-msoft-quad-float} option is the default.
10635 @item -mno-unaligned-doubles
10636 @itemx -munaligned-doubles
10637 @opindex mno-unaligned-doubles
10638 @opindex munaligned-doubles
10639 Assume that doubles have 8 byte alignment. This is the default.
10641 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
10642 alignment only if they are contained in another type, or if they have an
10643 absolute address. Otherwise, it assumes they have 4 byte alignment.
10644 Specifying this option avoids some rare compatibility problems with code
10645 generated by other compilers. It is not the default because it results
10646 in a performance loss, especially for floating point code.
10648 @item -mno-faster-structs
10649 @itemx -mfaster-structs
10650 @opindex mno-faster-structs
10651 @opindex mfaster-structs
10652 With @option{-mfaster-structs}, the compiler assumes that structures
10653 should have 8 byte alignment. This enables the use of pairs of
10654 @code{ldd} and @code{std} instructions for copies in structure
10655 assignment, in place of twice as many @code{ld} and @code{st} pairs.
10656 However, the use of this changed alignment directly violates the SPARC
10657 ABI@. Thus, it's intended only for use on targets where the developer
10658 acknowledges that their resulting code will not be directly in line with
10659 the rules of the ABI@.
10661 @item -mimpure-text
10662 @opindex mimpure-text
10663 @option{-mimpure-text}, used in addition to @option{-shared}, tells
10664 the compiler to not pass @option{-z text} to the linker when linking a
10665 shared object. Using this option, you can link position-dependent
10666 code into a shared object.
10668 @option{-mimpure-text} suppresses the ``relocations remain against
10669 allocatable but non-writable sections'' linker error message.
10670 However, the necessary relocations will trigger copy-on-write, and the
10671 shared object is not actually shared across processes. Instead of
10672 using @option{-mimpure-text}, you should compile all source code with
10673 @option{-fpic} or @option{-fPIC}.
10675 This option is only available on SunOS and Solaris.
10677 @item -mcpu=@var{cpu_type}
10679 Set the instruction set, register set, and instruction scheduling parameters
10680 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
10681 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
10682 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
10683 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
10684 @samp{ultrasparc3}.
10686 Default instruction scheduling parameters are used for values that select
10687 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
10688 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
10690 Here is a list of each supported architecture and their supported
10695 v8: supersparc, hypersparc
10696 sparclite: f930, f934, sparclite86x
10698 v9: ultrasparc, ultrasparc3
10701 By default (unless configured otherwise), GCC generates code for the V7
10702 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
10703 additionally optimizes it for the Cypress CY7C602 chip, as used in the
10704 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
10705 SPARCStation 1, 2, IPX etc.
10707 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
10708 architecture. The only difference from V7 code is that the compiler emits
10709 the integer multiply and integer divide instructions which exist in SPARC-V8
10710 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
10711 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
10714 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
10715 the SPARC architecture. This adds the integer multiply, integer divide step
10716 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
10717 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
10718 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU. With
10719 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
10720 MB86934 chip, which is the more recent SPARClite with FPU.
10722 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
10723 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
10724 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
10725 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
10726 optimizes it for the TEMIC SPARClet chip.
10728 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
10729 architecture. This adds 64-bit integer and floating-point move instructions,
10730 3 additional floating-point condition code registers and conditional move
10731 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
10732 optimizes it for the Sun UltraSPARC I/II chips. With
10733 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
10734 Sun UltraSPARC III chip.
10736 @item -mtune=@var{cpu_type}
10738 Set the instruction scheduling parameters for machine type
10739 @var{cpu_type}, but do not set the instruction set or register set that the
10740 option @option{-mcpu=@var{cpu_type}} would.
10742 The same values for @option{-mcpu=@var{cpu_type}} can be used for
10743 @option{-mtune=@var{cpu_type}}, but the only useful values are those
10744 that select a particular cpu implementation. Those are @samp{cypress},
10745 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
10746 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
10747 @samp{ultrasparc3}.
10752 @opindex mno-v8plus
10753 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI. The
10754 difference from the V8 ABI is that the global and out registers are
10755 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
10756 mode for all SPARC-V9 processors.
10762 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
10763 Visual Instruction Set extensions. The default is @option{-mno-vis}.
10766 These @samp{-m} options are supported in addition to the above
10767 on SPARC-V9 processors in 64-bit environments:
10770 @item -mlittle-endian
10771 @opindex mlittle-endian
10772 Generate code for a processor running in little-endian mode. It is only
10773 available for a few configurations and most notably not on Solaris.
10779 Generate code for a 32-bit or 64-bit environment.
10780 The 32-bit environment sets int, long and pointer to 32 bits.
10781 The 64-bit environment sets int to 32 bits and long and pointer
10784 @item -mcmodel=medlow
10785 @opindex mcmodel=medlow
10786 Generate code for the Medium/Low code model: 64-bit addresses, programs
10787 must be linked in the low 32 bits of memory. Programs can be statically
10788 or dynamically linked.
10790 @item -mcmodel=medmid
10791 @opindex mcmodel=medmid
10792 Generate code for the Medium/Middle code model: 64-bit addresses, programs
10793 must be linked in the low 44 bits of memory, the text and data segments must
10794 be less than 2GB in size and the data segment must be located within 2GB of
10797 @item -mcmodel=medany
10798 @opindex mcmodel=medany
10799 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
10800 may be linked anywhere in memory, the text and data segments must be less
10801 than 2GB in size and the data segment must be located within 2GB of the
10804 @item -mcmodel=embmedany
10805 @opindex mcmodel=embmedany
10806 Generate code for the Medium/Anywhere code model for embedded systems:
10807 64-bit addresses, the text and data segments must be less than 2GB in
10808 size, both starting anywhere in memory (determined at link time). The
10809 global register %g4 points to the base of the data segment. Programs
10810 are statically linked and PIC is not supported.
10813 @itemx -mno-stack-bias
10814 @opindex mstack-bias
10815 @opindex mno-stack-bias
10816 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
10817 frame pointer if present, are offset by @minus{}2047 which must be added back
10818 when making stack frame references. This is the default in 64-bit mode.
10819 Otherwise, assume no such offset is present.
10822 These switches are supported in addition to the above on Solaris:
10827 Add support for multithreading using the Solaris threads library. This
10828 option sets flags for both the preprocessor and linker. This option does
10829 not affect the thread safety of object code produced by the compiler or
10830 that of libraries supplied with it.
10834 Add support for multithreading using the POSIX threads library. This
10835 option sets flags for both the preprocessor and linker. This option does
10836 not affect the thread safety of object code produced by the compiler or
10837 that of libraries supplied with it.
10840 @node System V Options
10841 @subsection Options for System V
10843 These additional options are available on System V Release 4 for
10844 compatibility with other compilers on those systems:
10849 Create a shared object.
10850 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
10854 Identify the versions of each tool used by the compiler, in a
10855 @code{.ident} assembler directive in the output.
10859 Refrain from adding @code{.ident} directives to the output file (this is
10862 @item -YP,@var{dirs}
10864 Search the directories @var{dirs}, and no others, for libraries
10865 specified with @option{-l}.
10867 @item -Ym,@var{dir}
10869 Look in the directory @var{dir} to find the M4 preprocessor.
10870 The assembler uses this option.
10871 @c This is supposed to go with a -Yd for predefined M4 macro files, but
10872 @c the generic assembler that comes with Solaris takes just -Ym.
10875 @node TMS320C3x/C4x Options
10876 @subsection TMS320C3x/C4x Options
10877 @cindex TMS320C3x/C4x Options
10879 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
10883 @item -mcpu=@var{cpu_type}
10885 Set the instruction set, register set, and instruction scheduling
10886 parameters for machine type @var{cpu_type}. Supported values for
10887 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
10888 @samp{c44}. The default is @samp{c40} to generate code for the
10893 @itemx -msmall-memory
10895 @opindex mbig-memory
10897 @opindex msmall-memory
10899 Generates code for the big or small memory model. The small memory
10900 model assumed that all data fits into one 64K word page. At run-time
10901 the data page (DP) register must be set to point to the 64K page
10902 containing the .bss and .data program sections. The big memory model is
10903 the default and requires reloading of the DP register for every direct
10910 Allow (disallow) allocation of general integer operands into the block
10911 count register BK@.
10917 Enable (disable) generation of code using decrement and branch,
10918 DBcond(D), instructions. This is enabled by default for the C4x. To be
10919 on the safe side, this is disabled for the C3x, since the maximum
10920 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
10921 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
10922 that it can utilize the decrement and branch instruction, but will give
10923 up if there is more than one memory reference in the loop. Thus a loop
10924 where the loop counter is decremented can generate slightly more
10925 efficient code, in cases where the RPTB instruction cannot be utilized.
10927 @item -mdp-isr-reload
10929 @opindex mdp-isr-reload
10931 Force the DP register to be saved on entry to an interrupt service
10932 routine (ISR), reloaded to point to the data section, and restored on
10933 exit from the ISR@. This should not be required unless someone has
10934 violated the small memory model by modifying the DP register, say within
10941 For the C3x use the 24-bit MPYI instruction for integer multiplies
10942 instead of a library call to guarantee 32-bit results. Note that if one
10943 of the operands is a constant, then the multiplication will be performed
10944 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
10945 then squaring operations are performed inline instead of a library call.
10948 @itemx -mno-fast-fix
10950 @opindex mno-fast-fix
10951 The C3x/C4x FIX instruction to convert a floating point value to an
10952 integer value chooses the nearest integer less than or equal to the
10953 floating point value rather than to the nearest integer. Thus if the
10954 floating point number is negative, the result will be incorrectly
10955 truncated an additional code is necessary to detect and correct this
10956 case. This option can be used to disable generation of the additional
10957 code required to correct the result.
10963 Enable (disable) generation of repeat block sequences using the RPTB
10964 instruction for zero overhead looping. The RPTB construct is only used
10965 for innermost loops that do not call functions or jump across the loop
10966 boundaries. There is no advantage having nested RPTB loops due to the
10967 overhead required to save and restore the RC, RS, and RE registers.
10968 This is enabled by default with @option{-O2}.
10970 @item -mrpts=@var{count}
10974 Enable (disable) the use of the single instruction repeat instruction
10975 RPTS@. If a repeat block contains a single instruction, and the loop
10976 count can be guaranteed to be less than the value @var{count}, GCC will
10977 emit a RPTS instruction instead of a RPTB@. If no value is specified,
10978 then a RPTS will be emitted even if the loop count cannot be determined
10979 at compile time. Note that the repeated instruction following RPTS does
10980 not have to be reloaded from memory each iteration, thus freeing up the
10981 CPU buses for operands. However, since interrupts are blocked by this
10982 instruction, it is disabled by default.
10984 @item -mloop-unsigned
10985 @itemx -mno-loop-unsigned
10986 @opindex mloop-unsigned
10987 @opindex mno-loop-unsigned
10988 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
10989 is @math{2^{31} + 1} since these instructions test if the iteration count is
10990 negative to terminate the loop. If the iteration count is unsigned
10991 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
10992 exceeded. This switch allows an unsigned iteration count.
10996 Try to emit an assembler syntax that the TI assembler (asm30) is happy
10997 with. This also enforces compatibility with the API employed by the TI
10998 C3x C compiler. For example, long doubles are passed as structures
10999 rather than in floating point registers.
11005 Generate code that uses registers (stack) for passing arguments to functions.
11006 By default, arguments are passed in registers where possible rather
11007 than by pushing arguments on to the stack.
11009 @item -mparallel-insns
11010 @itemx -mno-parallel-insns
11011 @opindex mparallel-insns
11012 @opindex mno-parallel-insns
11013 Allow the generation of parallel instructions. This is enabled by
11014 default with @option{-O2}.
11016 @item -mparallel-mpy
11017 @itemx -mno-parallel-mpy
11018 @opindex mparallel-mpy
11019 @opindex mno-parallel-mpy
11020 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11021 provided @option{-mparallel-insns} is also specified. These instructions have
11022 tight register constraints which can pessimize the code generation
11023 of large functions.
11028 @subsection V850 Options
11029 @cindex V850 Options
11031 These @samp{-m} options are defined for V850 implementations:
11035 @itemx -mno-long-calls
11036 @opindex mlong-calls
11037 @opindex mno-long-calls
11038 Treat all calls as being far away (near). If calls are assumed to be
11039 far away, the compiler will always load the functions address up into a
11040 register, and call indirect through the pointer.
11046 Do not optimize (do optimize) basic blocks that use the same index
11047 pointer 4 or more times to copy pointer into the @code{ep} register, and
11048 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11049 option is on by default if you optimize.
11051 @item -mno-prolog-function
11052 @itemx -mprolog-function
11053 @opindex mno-prolog-function
11054 @opindex mprolog-function
11055 Do not use (do use) external functions to save and restore registers
11056 at the prologue and epilogue of a function. The external functions
11057 are slower, but use less code space if more than one function saves
11058 the same number of registers. The @option{-mprolog-function} option
11059 is on by default if you optimize.
11063 Try to make the code as small as possible. At present, this just turns
11064 on the @option{-mep} and @option{-mprolog-function} options.
11066 @item -mtda=@var{n}
11068 Put static or global variables whose size is @var{n} bytes or less into
11069 the tiny data area that register @code{ep} points to. The tiny data
11070 area can hold up to 256 bytes in total (128 bytes for byte references).
11072 @item -msda=@var{n}
11074 Put static or global variables whose size is @var{n} bytes or less into
11075 the small data area that register @code{gp} points to. The small data
11076 area can hold up to 64 kilobytes.
11078 @item -mzda=@var{n}
11080 Put static or global variables whose size is @var{n} bytes or less into
11081 the first 32 kilobytes of memory.
11085 Specify that the target processor is the V850.
11088 @opindex mbig-switch
11089 Generate code suitable for big switch tables. Use this option only if
11090 the assembler/linker complain about out of range branches within a switch
11095 This option will cause r2 and r5 to be used in the code generated by
11096 the compiler. This setting is the default.
11098 @item -mno-app-regs
11099 @opindex mno-app-regs
11100 This option will cause r2 and r5 to be treated as fixed registers.
11104 Specify that the target processor is the V850E1. The preprocessor
11105 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11106 this option is used.
11110 Specify that the target processor is the V850E. The preprocessor
11111 constant @samp{__v850e__} will be defined if this option is used.
11113 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11114 are defined then a default target processor will be chosen and the
11115 relevant @samp{__v850*__} preprocessor constant will be defined.
11117 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11118 defined, regardless of which processor variant is the target.
11120 @item -mdisable-callt
11121 @opindex mdisable-callt
11122 This option will suppress generation of the CALLT instruction for the
11123 v850e and v850e1 flavors of the v850 architecture. The default is
11124 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11129 @subsection VAX Options
11130 @cindex VAX options
11132 These @samp{-m} options are defined for the VAX:
11137 Do not output certain jump instructions (@code{aobleq} and so on)
11138 that the Unix assembler for the VAX cannot handle across long
11143 Do output those jump instructions, on the assumption that you
11144 will assemble with the GNU assembler.
11148 Output code for g-format floating point numbers instead of d-format.
11151 @node x86-64 Options
11152 @subsection x86-64 Options
11153 @cindex x86-64 options
11155 These are listed under @xref{i386 and x86-64 Options}.
11157 @node Xstormy16 Options
11158 @subsection Xstormy16 Options
11159 @cindex Xstormy16 Options
11161 These options are defined for Xstormy16:
11166 Choose startup files and linker script suitable for the simulator.
11169 @node Xtensa Options
11170 @subsection Xtensa Options
11171 @cindex Xtensa Options
11173 These options are supported for Xtensa targets:
11177 @itemx -mno-const16
11179 @opindex mno-const16
11180 Enable or disable use of @code{CONST16} instructions for loading
11181 constant values. The @code{CONST16} instruction is currently not a
11182 standard option from Tensilica. When enabled, @code{CONST16}
11183 instructions are always used in place of the standard @code{L32R}
11184 instructions. The use of @code{CONST16} is enabled by default only if
11185 the @code{L32R} instruction is not available.
11188 @itemx -mno-fused-madd
11189 @opindex mfused-madd
11190 @opindex mno-fused-madd
11191 Enable or disable use of fused multiply/add and multiply/subtract
11192 instructions in the floating-point option. This has no effect if the
11193 floating-point option is not also enabled. Disabling fused multiply/add
11194 and multiply/subtract instructions forces the compiler to use separate
11195 instructions for the multiply and add/subtract operations. This may be
11196 desirable in some cases where strict IEEE 754-compliant results are
11197 required: the fused multiply add/subtract instructions do not round the
11198 intermediate result, thereby producing results with @emph{more} bits of
11199 precision than specified by the IEEE standard. Disabling fused multiply
11200 add/subtract instructions also ensures that the program output is not
11201 sensitive to the compiler's ability to combine multiply and add/subtract
11204 @item -mtext-section-literals
11205 @itemx -mno-text-section-literals
11206 @opindex mtext-section-literals
11207 @opindex mno-text-section-literals
11208 Control the treatment of literal pools. The default is
11209 @option{-mno-text-section-literals}, which places literals in a separate
11210 section in the output file. This allows the literal pool to be placed
11211 in a data RAM/ROM, and it also allows the linker to combine literal
11212 pools from separate object files to remove redundant literals and
11213 improve code size. With @option{-mtext-section-literals}, the literals
11214 are interspersed in the text section in order to keep them as close as
11215 possible to their references. This may be necessary for large assembly
11218 @item -mtarget-align
11219 @itemx -mno-target-align
11220 @opindex mtarget-align
11221 @opindex mno-target-align
11222 When this option is enabled, GCC instructs the assembler to
11223 automatically align instructions to reduce branch penalties at the
11224 expense of some code density. The assembler attempts to widen density
11225 instructions to align branch targets and the instructions following call
11226 instructions. If there are not enough preceding safe density
11227 instructions to align a target, no widening will be performed. The
11228 default is @option{-mtarget-align}. These options do not affect the
11229 treatment of auto-aligned instructions like @code{LOOP}, which the
11230 assembler will always align, either by widening density instructions or
11231 by inserting no-op instructions.
11234 @itemx -mno-longcalls
11235 @opindex mlongcalls
11236 @opindex mno-longcalls
11237 When this option is enabled, GCC instructs the assembler to translate
11238 direct calls to indirect calls unless it can determine that the target
11239 of a direct call is in the range allowed by the call instruction. This
11240 translation typically occurs for calls to functions in other source
11241 files. Specifically, the assembler translates a direct @code{CALL}
11242 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11243 The default is @option{-mno-longcalls}. This option should be used in
11244 programs where the call target can potentially be out of range. This
11245 option is implemented in the assembler, not the compiler, so the
11246 assembly code generated by GCC will still show direct call
11247 instructions---look at the disassembled object code to see the actual
11248 instructions. Note that the assembler will use an indirect call for
11249 every cross-file call, not just those that really will be out of range.
11252 @node zSeries Options
11253 @subsection zSeries Options
11254 @cindex zSeries options
11256 These are listed under @xref{S/390 and zSeries Options}.
11258 @node Code Gen Options
11259 @section Options for Code Generation Conventions
11260 @cindex code generation conventions
11261 @cindex options, code generation
11262 @cindex run-time options
11264 These machine-independent options control the interface conventions
11265 used in code generation.
11267 Most of them have both positive and negative forms; the negative form
11268 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11269 one of the forms is listed---the one which is not the default. You
11270 can figure out the other form by either removing @samp{no-} or adding
11274 @item -fbounds-check
11275 @opindex fbounds-check
11276 For front-ends that support it, generate additional code to check that
11277 indices used to access arrays are within the declared range. This is
11278 currently only supported by the Java and Fortran 77 front-ends, where
11279 this option defaults to true and false respectively.
11283 This option generates traps for signed overflow on addition, subtraction,
11284 multiplication operations.
11288 This option instructs the compiler to assume that signed arithmetic
11289 overflow of addition, subtraction and multiplication wraps around
11290 using twos-complement representation. This flag enables some optimizations
11291 and disables other. This option is enabled by default for the Java
11292 front-end, as required by the Java language specification.
11295 @opindex fexceptions
11296 Enable exception handling. Generates extra code needed to propagate
11297 exceptions. For some targets, this implies GCC will generate frame
11298 unwind information for all functions, which can produce significant data
11299 size overhead, although it does not affect execution. If you do not
11300 specify this option, GCC will enable it by default for languages like
11301 C++ which normally require exception handling, and disable it for
11302 languages like C that do not normally require it. However, you may need
11303 to enable this option when compiling C code that needs to interoperate
11304 properly with exception handlers written in C++. You may also wish to
11305 disable this option if you are compiling older C++ programs that don't
11306 use exception handling.
11308 @item -fnon-call-exceptions
11309 @opindex fnon-call-exceptions
11310 Generate code that allows trapping instructions to throw exceptions.
11311 Note that this requires platform-specific runtime support that does
11312 not exist everywhere. Moreover, it only allows @emph{trapping}
11313 instructions to throw exceptions, i.e.@: memory references or floating
11314 point instructions. It does not allow exceptions to be thrown from
11315 arbitrary signal handlers such as @code{SIGALRM}.
11317 @item -funwind-tables
11318 @opindex funwind-tables
11319 Similar to @option{-fexceptions}, except that it will just generate any needed
11320 static data, but will not affect the generated code in any other way.
11321 You will normally not enable this option; instead, a language processor
11322 that needs this handling would enable it on your behalf.
11324 @item -fasynchronous-unwind-tables
11325 @opindex fasynchronous-unwind-tables
11326 Generate unwind table in dwarf2 format, if supported by target machine. The
11327 table is exact at each instruction boundary, so it can be used for stack
11328 unwinding from asynchronous events (such as debugger or garbage collector).
11330 @item -fpcc-struct-return
11331 @opindex fpcc-struct-return
11332 Return ``short'' @code{struct} and @code{union} values in memory like
11333 longer ones, rather than in registers. This convention is less
11334 efficient, but it has the advantage of allowing intercallability between
11335 GCC-compiled files and files compiled with other compilers, particularly
11336 the Portable C Compiler (pcc).
11338 The precise convention for returning structures in memory depends
11339 on the target configuration macros.
11341 Short structures and unions are those whose size and alignment match
11342 that of some integer type.
11344 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11345 switch is not binary compatible with code compiled with the
11346 @option{-freg-struct-return} switch.
11347 Use it to conform to a non-default application binary interface.
11349 @item -freg-struct-return
11350 @opindex freg-struct-return
11351 Return @code{struct} and @code{union} values in registers when possible.
11352 This is more efficient for small structures than
11353 @option{-fpcc-struct-return}.
11355 If you specify neither @option{-fpcc-struct-return} nor
11356 @option{-freg-struct-return}, GCC defaults to whichever convention is
11357 standard for the target. If there is no standard convention, GCC
11358 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11359 the principal compiler. In those cases, we can choose the standard, and
11360 we chose the more efficient register return alternative.
11362 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11363 switch is not binary compatible with code compiled with the
11364 @option{-fpcc-struct-return} switch.
11365 Use it to conform to a non-default application binary interface.
11367 @item -fshort-enums
11368 @opindex fshort-enums
11369 Allocate to an @code{enum} type only as many bytes as it needs for the
11370 declared range of possible values. Specifically, the @code{enum} type
11371 will be equivalent to the smallest integer type which has enough room.
11373 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11374 code that is not binary compatible with code generated without that switch.
11375 Use it to conform to a non-default application binary interface.
11377 @item -fshort-double
11378 @opindex fshort-double
11379 Use the same size for @code{double} as for @code{float}.
11381 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11382 code that is not binary compatible with code generated without that switch.
11383 Use it to conform to a non-default application binary interface.
11385 @item -fshort-wchar
11386 @opindex fshort-wchar
11387 Override the underlying type for @samp{wchar_t} to be @samp{short
11388 unsigned int} instead of the default for the target. This option is
11389 useful for building programs to run under WINE@.
11391 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11392 code that is not binary compatible with code generated without that switch.
11393 Use it to conform to a non-default application binary interface.
11395 @item -fshared-data
11396 @opindex fshared-data
11397 Requests that the data and non-@code{const} variables of this
11398 compilation be shared data rather than private data. The distinction
11399 makes sense only on certain operating systems, where shared data is
11400 shared between processes running the same program, while private data
11401 exists in one copy per process.
11404 @opindex fno-common
11405 In C, allocate even uninitialized global variables in the data section of the
11406 object file, rather than generating them as common blocks. This has the
11407 effect that if the same variable is declared (without @code{extern}) in
11408 two different compilations, you will get an error when you link them.
11409 The only reason this might be useful is if you wish to verify that the
11410 program will work on other systems which always work this way.
11414 Ignore the @samp{#ident} directive.
11416 @item -finhibit-size-directive
11417 @opindex finhibit-size-directive
11418 Don't output a @code{.size} assembler directive, or anything else that
11419 would cause trouble if the function is split in the middle, and the
11420 two halves are placed at locations far apart in memory. This option is
11421 used when compiling @file{crtstuff.c}; you should not need to use it
11424 @item -fverbose-asm
11425 @opindex fverbose-asm
11426 Put extra commentary information in the generated assembly code to
11427 make it more readable. This option is generally only of use to those
11428 who actually need to read the generated assembly code (perhaps while
11429 debugging the compiler itself).
11431 @option{-fno-verbose-asm}, the default, causes the
11432 extra information to be omitted and is useful when comparing two assembler
11437 @cindex global offset table
11439 Generate position-independent code (PIC) suitable for use in a shared
11440 library, if supported for the target machine. Such code accesses all
11441 constant addresses through a global offset table (GOT)@. The dynamic
11442 loader resolves the GOT entries when the program starts (the dynamic
11443 loader is not part of GCC; it is part of the operating system). If
11444 the GOT size for the linked executable exceeds a machine-specific
11445 maximum size, you get an error message from the linker indicating that
11446 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11447 instead. (These maximums are 8k on the SPARC and 32k
11448 on the m68k and RS/6000. The 386 has no such limit.)
11450 Position-independent code requires special support, and therefore works
11451 only on certain machines. For the 386, GCC supports PIC for System V
11452 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11453 position-independent.
11457 If supported for the target machine, emit position-independent code,
11458 suitable for dynamic linking and avoiding any limit on the size of the
11459 global offset table. This option makes a difference on the m68k
11462 Position-independent code requires special support, and therefore works
11463 only on certain machines.
11469 These options are similar to @option{-fpic} and @option{-fPIC}, but
11470 generated position independent code can be only linked into executables.
11471 Usually these options are used when @option{-pie} GCC option will be
11472 used during linking.
11474 @item -ffixed-@var{reg}
11476 Treat the register named @var{reg} as a fixed register; generated code
11477 should never refer to it (except perhaps as a stack pointer, frame
11478 pointer or in some other fixed role).
11480 @var{reg} must be the name of a register. The register names accepted
11481 are machine-specific and are defined in the @code{REGISTER_NAMES}
11482 macro in the machine description macro file.
11484 This flag does not have a negative form, because it specifies a
11487 @item -fcall-used-@var{reg}
11488 @opindex fcall-used
11489 Treat the register named @var{reg} as an allocable register that is
11490 clobbered by function calls. It may be allocated for temporaries or
11491 variables that do not live across a call. Functions compiled this way
11492 will not save and restore the register @var{reg}.
11494 It is an error to used this flag with the frame pointer or stack pointer.
11495 Use of this flag for other registers that have fixed pervasive roles in
11496 the machine's execution model will produce disastrous results.
11498 This flag does not have a negative form, because it specifies a
11501 @item -fcall-saved-@var{reg}
11502 @opindex fcall-saved
11503 Treat the register named @var{reg} as an allocable register saved by
11504 functions. It may be allocated even for temporaries or variables that
11505 live across a call. Functions compiled this way will save and restore
11506 the register @var{reg} if they use it.
11508 It is an error to used this flag with the frame pointer or stack pointer.
11509 Use of this flag for other registers that have fixed pervasive roles in
11510 the machine's execution model will produce disastrous results.
11512 A different sort of disaster will result from the use of this flag for
11513 a register in which function values may be returned.
11515 This flag does not have a negative form, because it specifies a
11518 @item -fpack-struct
11519 @opindex fpack-struct
11520 Pack all structure members together without holes.
11522 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11523 code that is not binary compatible with code generated without that switch.
11524 Additionally, it makes the code suboptimal.
11525 Use it to conform to a non-default application binary interface.
11527 @item -finstrument-functions
11528 @opindex finstrument-functions
11529 Generate instrumentation calls for entry and exit to functions. Just
11530 after function entry and just before function exit, the following
11531 profiling functions will be called with the address of the current
11532 function and its call site. (On some platforms,
11533 @code{__builtin_return_address} does not work beyond the current
11534 function, so the call site information may not be available to the
11535 profiling functions otherwise.)
11538 void __cyg_profile_func_enter (void *this_fn,
11540 void __cyg_profile_func_exit (void *this_fn,
11544 The first argument is the address of the start of the current function,
11545 which may be looked up exactly in the symbol table.
11547 This instrumentation is also done for functions expanded inline in other
11548 functions. The profiling calls will indicate where, conceptually, the
11549 inline function is entered and exited. This means that addressable
11550 versions of such functions must be available. If all your uses of a
11551 function are expanded inline, this may mean an additional expansion of
11552 code size. If you use @samp{extern inline} in your C code, an
11553 addressable version of such functions must be provided. (This is
11554 normally the case anyways, but if you get lucky and the optimizer always
11555 expands the functions inline, you might have gotten away without
11556 providing static copies.)
11558 A function may be given the attribute @code{no_instrument_function}, in
11559 which case this instrumentation will not be done. This can be used, for
11560 example, for the profiling functions listed above, high-priority
11561 interrupt routines, and any functions from which the profiling functions
11562 cannot safely be called (perhaps signal handlers, if the profiling
11563 routines generate output or allocate memory).
11565 @item -fstack-check
11566 @opindex fstack-check
11567 Generate code to verify that you do not go beyond the boundary of the
11568 stack. You should specify this flag if you are running in an
11569 environment with multiple threads, but only rarely need to specify it in
11570 a single-threaded environment since stack overflow is automatically
11571 detected on nearly all systems if there is only one stack.
11573 Note that this switch does not actually cause checking to be done; the
11574 operating system must do that. The switch causes generation of code
11575 to ensure that the operating system sees the stack being extended.
11577 @item -fstack-limit-register=@var{reg}
11578 @itemx -fstack-limit-symbol=@var{sym}
11579 @itemx -fno-stack-limit
11580 @opindex fstack-limit-register
11581 @opindex fstack-limit-symbol
11582 @opindex fno-stack-limit
11583 Generate code to ensure that the stack does not grow beyond a certain value,
11584 either the value of a register or the address of a symbol. If the stack
11585 would grow beyond the value, a signal is raised. For most targets,
11586 the signal is raised before the stack overruns the boundary, so
11587 it is possible to catch the signal without taking special precautions.
11589 For instance, if the stack starts at absolute address @samp{0x80000000}
11590 and grows downwards, you can use the flags
11591 @option{-fstack-limit-symbol=__stack_limit} and
11592 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
11593 of 128KB@. Note that this may only work with the GNU linker.
11595 @cindex aliasing of parameters
11596 @cindex parameters, aliased
11597 @item -fargument-alias
11598 @itemx -fargument-noalias
11599 @itemx -fargument-noalias-global
11600 @opindex fargument-alias
11601 @opindex fargument-noalias
11602 @opindex fargument-noalias-global
11603 Specify the possible relationships among parameters and between
11604 parameters and global data.
11606 @option{-fargument-alias} specifies that arguments (parameters) may
11607 alias each other and may alias global storage.@*
11608 @option{-fargument-noalias} specifies that arguments do not alias
11609 each other, but may alias global storage.@*
11610 @option{-fargument-noalias-global} specifies that arguments do not
11611 alias each other and do not alias global storage.
11613 Each language will automatically use whatever option is required by
11614 the language standard. You should not need to use these options yourself.
11616 @item -fleading-underscore
11617 @opindex fleading-underscore
11618 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11619 change the way C symbols are represented in the object file. One use
11620 is to help link with legacy assembly code.
11622 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11623 generate code that is not binary compatible with code generated without that
11624 switch. Use it to conform to a non-default application binary interface.
11625 Not all targets provide complete support for this switch.
11627 @item -ftls-model=@var{model}
11628 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11629 The @var{model} argument should be one of @code{global-dynamic},
11630 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
11632 The default without @option{-fpic} is @code{initial-exec}; with
11633 @option{-fpic} the default is @code{global-dynamic}.
11638 @node Environment Variables
11639 @section Environment Variables Affecting GCC
11640 @cindex environment variables
11642 @c man begin ENVIRONMENT
11643 This section describes several environment variables that affect how GCC
11644 operates. Some of them work by specifying directories or prefixes to use
11645 when searching for various kinds of files. Some are used to specify other
11646 aspects of the compilation environment.
11648 Note that you can also specify places to search using options such as
11649 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
11650 take precedence over places specified using environment variables, which
11651 in turn take precedence over those specified by the configuration of GCC@.
11652 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
11653 GNU Compiler Collection (GCC) Internals}.
11658 @c @itemx LC_COLLATE
11660 @c @itemx LC_MONETARY
11661 @c @itemx LC_NUMERIC
11666 @c @findex LC_COLLATE
11667 @findex LC_MESSAGES
11668 @c @findex LC_MONETARY
11669 @c @findex LC_NUMERIC
11673 These environment variables control the way that GCC uses
11674 localization information that allow GCC to work with different
11675 national conventions. GCC inspects the locale categories
11676 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
11677 so. These locale categories can be set to any value supported by your
11678 installation. A typical value is @samp{en_UK} for English in the United
11681 The @env{LC_CTYPE} environment variable specifies character
11682 classification. GCC uses it to determine the character boundaries in
11683 a string; this is needed for some multibyte encodings that contain quote
11684 and escape characters that would otherwise be interpreted as a string
11687 The @env{LC_MESSAGES} environment variable specifies the language to
11688 use in diagnostic messages.
11690 If the @env{LC_ALL} environment variable is set, it overrides the value
11691 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
11692 and @env{LC_MESSAGES} default to the value of the @env{LANG}
11693 environment variable. If none of these variables are set, GCC
11694 defaults to traditional C English behavior.
11698 If @env{TMPDIR} is set, it specifies the directory to use for temporary
11699 files. GCC uses temporary files to hold the output of one stage of
11700 compilation which is to be used as input to the next stage: for example,
11701 the output of the preprocessor, which is the input to the compiler
11704 @item GCC_EXEC_PREFIX
11705 @findex GCC_EXEC_PREFIX
11706 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
11707 names of the subprograms executed by the compiler. No slash is added
11708 when this prefix is combined with the name of a subprogram, but you can
11709 specify a prefix that ends with a slash if you wish.
11711 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
11712 an appropriate prefix to use based on the pathname it was invoked with.
11714 If GCC cannot find the subprogram using the specified prefix, it
11715 tries looking in the usual places for the subprogram.
11717 The default value of @env{GCC_EXEC_PREFIX} is
11718 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
11719 of @code{prefix} when you ran the @file{configure} script.
11721 Other prefixes specified with @option{-B} take precedence over this prefix.
11723 This prefix is also used for finding files such as @file{crt0.o} that are
11726 In addition, the prefix is used in an unusual way in finding the
11727 directories to search for header files. For each of the standard
11728 directories whose name normally begins with @samp{/usr/local/lib/gcc}
11729 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
11730 replacing that beginning with the specified prefix to produce an
11731 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
11732 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
11733 These alternate directories are searched first; the standard directories
11736 @item COMPILER_PATH
11737 @findex COMPILER_PATH
11738 The value of @env{COMPILER_PATH} is a colon-separated list of
11739 directories, much like @env{PATH}. GCC tries the directories thus
11740 specified when searching for subprograms, if it can't find the
11741 subprograms using @env{GCC_EXEC_PREFIX}.
11744 @findex LIBRARY_PATH
11745 The value of @env{LIBRARY_PATH} is a colon-separated list of
11746 directories, much like @env{PATH}. When configured as a native compiler,
11747 GCC tries the directories thus specified when searching for special
11748 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
11749 using GCC also uses these directories when searching for ordinary
11750 libraries for the @option{-l} option (but directories specified with
11751 @option{-L} come first).
11755 @cindex locale definition
11756 This variable is used to pass locale information to the compiler. One way in
11757 which this information is used is to determine the character set to be used
11758 when character literals, string literals and comments are parsed in C and C++.
11759 When the compiler is configured to allow multibyte characters,
11760 the following values for @env{LANG} are recognized:
11764 Recognize JIS characters.
11766 Recognize SJIS characters.
11768 Recognize EUCJP characters.
11771 If @env{LANG} is not defined, or if it has some other value, then the
11772 compiler will use mblen and mbtowc as defined by the default locale to
11773 recognize and translate multibyte characters.
11777 Some additional environments variables affect the behavior of the
11780 @include cppenv.texi
11784 @node Precompiled Headers
11785 @section Using Precompiled Headers
11786 @cindex precompiled headers
11787 @cindex speed of compilation
11789 Often large projects have many header files that are included in every
11790 source file. The time the compiler takes to process these header files
11791 over and over again can account for nearly all of the time required to
11792 build the project. To make builds faster, GCC allows users to
11793 `precompile' a header file; then, if builds can use the precompiled
11794 header file they will be much faster.
11796 @strong{Caution:} There are a few known situations where GCC will
11797 crash when trying to use a precompiled header. If you have trouble
11798 with a precompiled header, you should remove the precompiled header
11799 and compile without it. In addition, please use GCC's on-line
11800 defect-tracking system to report any problems you encounter with
11801 precompiled headers. @xref{Bugs}.
11803 To create a precompiled header file, simply compile it as you would any
11804 other file, if necessary using the @option{-x} option to make the driver
11805 treat it as a C or C++ header file. You will probably want to use a
11806 tool like @command{make} to keep the precompiled header up-to-date when
11807 the headers it contains change.
11809 A precompiled header file will be searched for when @code{#include} is
11810 seen in the compilation. As it searches for the included file
11811 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
11812 compiler looks for a precompiled header in each directory just before it
11813 looks for the include file in that directory. The name searched for is
11814 the name specified in the @code{#include} with @samp{.gch} appended. If
11815 the precompiled header file can't be used, it is ignored.
11817 For instance, if you have @code{#include "all.h"}, and you have
11818 @file{all.h.gch} in the same directory as @file{all.h}, then the
11819 precompiled header file will be used if possible, and the original
11820 header will be used otherwise.
11822 Alternatively, you might decide to put the precompiled header file in a
11823 directory and use @option{-I} to ensure that directory is searched
11824 before (or instead of) the directory containing the original header.
11825 Then, if you want to check that the precompiled header file is always
11826 used, you can put a file of the same name as the original header in this
11827 directory containing an @code{#error} command.
11829 This also works with @option{-include}. So yet another way to use
11830 precompiled headers, good for projects not designed with precompiled
11831 header files in mind, is to simply take most of the header files used by
11832 a project, include them from another header file, precompile that header
11833 file, and @option{-include} the precompiled header. If the header files
11834 have guards against multiple inclusion, they will be skipped because
11835 they've already been included (in the precompiled header).
11837 If you need to precompile the same header file for different
11838 languages, targets, or compiler options, you can instead make a
11839 @emph{directory} named like @file{all.h.gch}, and put each precompiled
11840 header in the directory, perhaps using @option{-o}. It doesn't matter
11841 what you call the files in the directory, every precompiled header in
11842 the directory will be considered. The first precompiled header
11843 encountered in the directory that is valid for this compilation will
11844 be used; they're searched in no particular order.
11846 There are many other possibilities, limited only by your imagination,
11847 good sense, and the constraints of your build system.
11849 A precompiled header file can be used only when these conditions apply:
11853 Only one precompiled header can be used in a particular compilation.
11856 A precompiled header can't be used once the first C token is seen. You
11857 can have preprocessor directives before a precompiled header; you can
11858 even include a precompiled header from inside another header, so long as
11859 there are no C tokens before the @code{#include}.
11862 The precompiled header file must be produced for the same language as
11863 the current compilation. You can't use a C precompiled header for a C++
11867 The precompiled header file must be produced by the same compiler
11868 version and configuration as the current compilation is using.
11869 The easiest way to guarantee this is to use the same compiler binary
11870 for creating and using precompiled headers.
11873 Any macros defined before the precompiled header is included must
11874 either be defined in the same way as when the precompiled header was
11875 generated, or must not affect the precompiled header, which usually
11876 means that the they don't appear in the precompiled header at all.
11878 The @option{-D} option is one way to define a macro before a
11879 precompiled header is included; using a @code{#define} can also do it.
11880 There are also some options that define macros implicitly, like
11881 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
11884 @item If debugging information is output when using the precompiled
11885 header, using @option{-g} or similar, the same kind of debugging information
11886 must have been output when building the precompiled header. However,
11887 a precompiled header built using @option{-g} can be used in a compilation
11888 when no debugging information is being output.
11890 @item The same @option{-m} options must generally be used when building
11891 and using the precompiled header. @xref{Submodel Options},
11892 for any cases where this rule is relaxed.
11894 @item Each of the following options must be the same when building and using
11895 the precompiled header:
11897 @gccoptlist{-fexceptions -funit-at-a-time}
11900 Some other command-line options starting with @option{-f},
11901 @option{-p}, or @option{-O} must be defined in the same way as when
11902 the precompiled header was generated. At present, it's not clear
11903 which options are safe to change and which are not; the safest choice
11904 is to use exactly the same options when generating and using the
11905 precompiled header. The following are known to be safe:
11907 @gccoptlist{-fpreprocessed -pedantic-errors}
11911 For all of these except the last, the compiler will automatically
11912 ignore the precompiled header if the conditions aren't met. If you
11913 find an option combination that doesn't work and doesn't cause the
11914 precompiled header to be ignored, please consider filing a bug report,
11917 If you do use differing options when generating and using the
11918 precompiled header, the actual behaviour will be a mixture of the
11919 behaviour for the options. For instance, if you use @option{-g} to
11920 generate the precompiled header but not when using it, you may or may
11921 not get debugging information for routines in the precompiled header.
11923 @node Running Protoize
11924 @section Running Protoize
11926 The program @code{protoize} is an optional part of GCC@. You can use
11927 it to add prototypes to a program, thus converting the program to ISO
11928 C in one respect. The companion program @code{unprotoize} does the
11929 reverse: it removes argument types from any prototypes that are found.
11931 When you run these programs, you must specify a set of source files as
11932 command line arguments. The conversion programs start out by compiling
11933 these files to see what functions they define. The information gathered
11934 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
11936 After scanning comes actual conversion. The specified files are all
11937 eligible to be converted; any files they include (whether sources or
11938 just headers) are eligible as well.
11940 But not all the eligible files are converted. By default,
11941 @code{protoize} and @code{unprotoize} convert only source and header
11942 files in the current directory. You can specify additional directories
11943 whose files should be converted with the @option{-d @var{directory}}
11944 option. You can also specify particular files to exclude with the
11945 @option{-x @var{file}} option. A file is converted if it is eligible, its
11946 directory name matches one of the specified directory names, and its
11947 name within the directory has not been excluded.
11949 Basic conversion with @code{protoize} consists of rewriting most
11950 function definitions and function declarations to specify the types of
11951 the arguments. The only ones not rewritten are those for varargs
11954 @code{protoize} optionally inserts prototype declarations at the
11955 beginning of the source file, to make them available for any calls that
11956 precede the function's definition. Or it can insert prototype
11957 declarations with block scope in the blocks where undeclared functions
11960 Basic conversion with @code{unprotoize} consists of rewriting most
11961 function declarations to remove any argument types, and rewriting
11962 function definitions to the old-style pre-ISO form.
11964 Both conversion programs print a warning for any function declaration or
11965 definition that they can't convert. You can suppress these warnings
11968 The output from @code{protoize} or @code{unprotoize} replaces the
11969 original source file. The original file is renamed to a name ending
11970 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
11971 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
11972 for DOS) file already exists, then the source file is simply discarded.
11974 @code{protoize} and @code{unprotoize} both depend on GCC itself to
11975 scan the program and collect information about the functions it uses.
11976 So neither of these programs will work until GCC is installed.
11978 Here is a table of the options you can use with @code{protoize} and
11979 @code{unprotoize}. Each option works with both programs unless
11983 @item -B @var{directory}
11984 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
11985 usual directory (normally @file{/usr/local/lib}). This file contains
11986 prototype information about standard system functions. This option
11987 applies only to @code{protoize}.
11989 @item -c @var{compilation-options}
11990 Use @var{compilation-options} as the options when running @command{gcc} to
11991 produce the @samp{.X} files. The special option @option{-aux-info} is
11992 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
11994 Note that the compilation options must be given as a single argument to
11995 @code{protoize} or @code{unprotoize}. If you want to specify several
11996 @command{gcc} options, you must quote the entire set of compilation options
11997 to make them a single word in the shell.
11999 There are certain @command{gcc} arguments that you cannot use, because they
12000 would produce the wrong kind of output. These include @option{-g},
12001 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12002 the @var{compilation-options}, they are ignored.
12005 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12006 systems) instead of @samp{.c}. This is convenient if you are converting
12007 a C program to C++. This option applies only to @code{protoize}.
12010 Add explicit global declarations. This means inserting explicit
12011 declarations at the beginning of each source file for each function
12012 that is called in the file and was not declared. These declarations
12013 precede the first function definition that contains a call to an
12014 undeclared function. This option applies only to @code{protoize}.
12016 @item -i @var{string}
12017 Indent old-style parameter declarations with the string @var{string}.
12018 This option applies only to @code{protoize}.
12020 @code{unprotoize} converts prototyped function definitions to old-style
12021 function definitions, where the arguments are declared between the
12022 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12023 uses five spaces as the indentation. If you want to indent with just
12024 one space instead, use @option{-i " "}.
12027 Keep the @samp{.X} files. Normally, they are deleted after conversion
12031 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12032 a prototype declaration for each function in each block which calls the
12033 function without any declaration. This option applies only to
12037 Make no real changes. This mode just prints information about the conversions
12038 that would have been done without @option{-n}.
12041 Make no @samp{.save} files. The original files are simply deleted.
12042 Use this option with caution.
12044 @item -p @var{program}
12045 Use the program @var{program} as the compiler. Normally, the name
12046 @file{gcc} is used.
12049 Work quietly. Most warnings are suppressed.
12052 Print the version number, just like @option{-v} for @command{gcc}.
12055 If you need special compiler options to compile one of your program's
12056 source files, then you should generate that file's @samp{.X} file
12057 specially, by running @command{gcc} on that source file with the
12058 appropriate options and the option @option{-aux-info}. Then run
12059 @code{protoize} on the entire set of files. @code{protoize} will use
12060 the existing @samp{.X} file because it is newer than the source file.
12064 gcc -Dfoo=bar file1.c -aux-info file1.X
12069 You need to include the special files along with the rest in the
12070 @code{protoize} command, even though their @samp{.X} files already
12071 exist, because otherwise they won't get converted.
12073 @xref{Protoize Caveats}, for more information on how to use
12074 @code{protoize} successfully.