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 and Objective-C++ Dialect Options:: Variations on Objective-C
126 * Language Independent Options:: Controlling how diagnostics should be
128 * Warning Options:: How picky should the compiler be?
129 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
130 * Optimize Options:: How much optimization?
131 * Preprocessor Options:: Controlling header files and macro definitions.
132 Also, getting dependency information for Make.
133 * Assembler Options:: Passing options to the assembler.
134 * Link Options:: Specifying libraries and so on.
135 * Directory Options:: Where to find header files and libraries.
136 Where to find the compiler executable files.
137 * Spec Files:: How to pass switches to sub-processes.
138 * Target Options:: Running a cross-compiler, or an old version of GCC.
139 * Submodel Options:: Specifying minor hardware or convention variations,
140 such as 68010 vs 68020.
141 * Code Gen Options:: Specifying conventions for function calls, data layout
143 * Environment Variables:: Env vars that affect GCC.
144 * Precompiled Headers:: Compiling a header once, and using it many times.
145 * Running Protoize:: Automatically adding or removing function prototypes.
151 @section Option Summary
153 Here is a summary of all the options, grouped by type. Explanations are
154 in the following sections.
157 @item Overall Options
158 @xref{Overall Options,,Options Controlling the Kind of Output}.
159 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
160 -x @var{language} -v -### --help --target-help --version}
162 @item C Language Options
163 @xref{C Dialect Options,,Options Controlling C Dialect}.
164 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
165 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
166 -fhosted -ffreestanding -fms-extensions @gol
167 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
168 -fallow-single-precision -fcond-mismatch @gol
169 -fsigned-bitfields -fsigned-char @gol
170 -funsigned-bitfields -funsigned-char}
172 @item C++ Language Options
173 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
174 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
175 -fconserve-space -fno-const-strings @gol
176 -fno-elide-constructors @gol
177 -fno-enforce-eh-specs @gol
178 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
179 -fno-implicit-templates @gol
180 -fno-implicit-inline-templates @gol
181 -fno-implement-inlines -fms-extensions @gol
182 -fno-nonansi-builtins -fno-operator-names @gol
183 -fno-optional-diags -fpermissive @gol
184 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
185 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
186 -fno-default-inline -fvisibility-inlines-hidden @gol
187 -Wabi -Wctor-dtor-privacy @gol
188 -Wnon-virtual-dtor -Wreorder @gol
189 -Weffc++ -Wno-deprecated @gol
190 -Wno-non-template-friend -Wold-style-cast @gol
191 -Woverloaded-virtual -Wno-pmf-conversions @gol
192 -Wsign-promo -Wsynth}
194 @item Objective-C and Objective-C++ Language Options
195 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
196 Objective-C and Objective-C++ Dialects}.
198 -fconstant-string-class=@var{class-name} @gol
199 -fgnu-runtime -fnext-runtime @gol
200 -fno-nil-receivers @gol
201 -fobjc-exceptions @gol
202 -freplace-objc-classes @gol
205 -Wno-protocol -Wselector -Wundeclared-selector}
207 @item Language Independent Options
208 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
209 @gccoptlist{-fmessage-length=@var{n} @gol
210 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
212 @item Warning Options
213 @xref{Warning Options,,Options to Request or Suppress Warnings}.
214 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
215 -w -Wextra -Wall -Waggregate-return @gol
216 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
217 -Wconversion -Wno-deprecated-declarations @gol
218 -Wdisabled-optimization -Wno-div-by-zero -Wendif-labels @gol
219 -Werror -Werror-implicit-function-declaration @gol
220 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
221 -Wno-format-extra-args -Wformat-nonliteral @gol
222 -Wformat-security -Wformat-y2k @gol
223 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
224 -Wimport -Wno-import -Winit-self -Winline @gol
225 -Wno-invalid-offsetof -Winvalid-pch @gol
226 -Wlarger-than-@var{len} -Wlong-long @gol
227 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
228 -Wmissing-format-attribute -Wmissing-include-dirs @gol
229 -Wmissing-noreturn @gol
230 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
231 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
232 -Wreturn-type -Wsequence-point -Wshadow @gol
233 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
234 -Wswitch -Wswitch-default -Wswitch-enum @gol
235 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
236 -Wunknown-pragmas -Wunreachable-code @gol
237 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
238 -Wunused-value -Wunused-variable -Wwrite-strings @gol
241 @item C-only Warning Options
242 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
243 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
244 -Wstrict-prototypes -Wtraditional @gol
245 -Wdeclaration-after-statement}
247 @item Debugging Options
248 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
249 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
250 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
251 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
252 -fdump-ipa-all -fdump-ipa-cgraph @gol
254 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
255 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
256 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
257 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
259 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
267 -fdump-tree-nrv -fdump-tree-vect @gol
268 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
270 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
271 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
272 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
273 -ftest-coverage -ftime-report -fvar-tracking @gol
274 -g -g@var{level} -gcoff -gdwarf-2 @gol
275 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
276 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
277 -print-multi-directory -print-multi-lib @gol
278 -print-prog-name=@var{program} -print-search-dirs -Q @gol
281 @item Optimization Options
282 @xref{Optimize Options,,Options that Control Optimization}.
283 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
284 -falign-labels=@var{n} -falign-loops=@var{n} @gol
285 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
286 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
287 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
288 -fcaller-saves -fcprop-registers @gol
289 -fcse-follow-jumps -fcse-skip-blocks -fdata-sections @gol
290 -fdelayed-branch -fdelete-null-pointer-checks @gol
291 -fexpensive-optimizations -ffast-math -ffloat-store @gol
292 -fforce-addr -fforce-mem -ffunction-sections @gol
293 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
294 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
295 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
296 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
297 -fmodulo-sched -fnew-ra -fno-branch-count-reg @gol
298 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
299 -fno-function-cse -fno-guess-branch-probability @gol
300 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
301 -funsafe-math-optimizations -ffinite-math-only @gol
302 -fno-trapping-math -fno-zero-initialized-in-bss @gol
303 -fomit-frame-pointer -foptimize-register-move @gol
304 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
305 -fprofile-generate -fprofile-use @gol
306 -fregmove -frename-registers @gol
307 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
308 -frerun-cse-after-loop -frerun-loop-opt @gol
309 -frounding-math -fschedule-insns -fschedule-insns2 @gol
310 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
311 -fsched-spec-load-dangerous @gol
312 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
313 -fsched2-use-superblocks @gol
314 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
315 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
316 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
317 -funroll-all-loops -funroll-loops -fpeel-loops @gol
318 -fsplit-ivs-in-unroller -funswitch-loops @gol
319 -fvariable-expansion-in-unroller @gol
320 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
321 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
322 -ftree-dominator-opts -ftree-dse -ftree-copyrename @gol
323 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
324 --param @var{name}=@var{value}
325 -O -O0 -O1 -O2 -O3 -Os}
327 @item Preprocessor Options
328 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
329 @gccoptlist{-A@var{question}=@var{answer} @gol
330 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
331 -C -dD -dI -dM -dN @gol
332 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
333 -idirafter @var{dir} @gol
334 -include @var{file} -imacros @var{file} @gol
335 -iprefix @var{file} -iwithprefix @var{dir} @gol
336 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
337 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
338 -P -fworking-directory -remap @gol
339 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
340 -Xpreprocessor @var{option}}
342 @item Assembler Option
343 @xref{Assembler Options,,Passing Options to the Assembler}.
344 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
347 @xref{Link Options,,Options for Linking}.
348 @gccoptlist{@var{object-file-name} -l@var{library} @gol
349 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
350 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
351 -Wl,@var{option} -Xlinker @var{option} @gol
354 @item Directory Options
355 @xref{Directory Options,,Options for Directory Search}.
356 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
359 @c I wrote this xref this way to avoid overfull hbox. -- rms
360 @xref{Target Options}.
361 @gccoptlist{-V @var{version} -b @var{machine}}
363 @item Machine Dependent Options
364 @xref{Submodel Options,,Hardware Models and Configurations}.
365 @c This list is ordered alphanumerically by subsection name.
366 @c Try and put the significant identifier (CPU or system) first,
367 @c so users have a clue at guessing where the ones they want will be.
370 @gccoptlist{-EB -EL @gol
371 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
372 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
375 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
376 -mabi=@var{name} @gol
377 -mapcs-stack-check -mno-apcs-stack-check @gol
378 -mapcs-float -mno-apcs-float @gol
379 -mapcs-reentrant -mno-apcs-reentrant @gol
380 -msched-prolog -mno-sched-prolog @gol
381 -mlittle-endian -mbig-endian -mwords-little-endian @gol
382 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
383 -mthumb-interwork -mno-thumb-interwork @gol
384 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
385 -mstructure-size-boundary=@var{n} @gol
386 -mabort-on-noreturn @gol
387 -mlong-calls -mno-long-calls @gol
388 -msingle-pic-base -mno-single-pic-base @gol
389 -mpic-register=@var{reg} @gol
390 -mnop-fun-dllimport @gol
391 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
392 -mpoke-function-name @gol
394 -mtpcs-frame -mtpcs-leaf-frame @gol
395 -mcaller-super-interworking -mcallee-super-interworking}
398 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
399 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
402 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
403 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
404 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
405 -mstack-align -mdata-align -mconst-align @gol
406 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
407 -melf -maout -melinux -mlinux -sim -sim2 @gol
408 -mmul-bug-workaround -mno-mul-bug-workaround}
410 @emph{Darwin Options}
411 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
412 -arch_only -bind_at_load -bundle -bundle_loader @gol
413 -client_name -compatibility_version -current_version @gol
415 -dependency-file -dylib_file -dylinker_install_name @gol
416 -dynamic -dynamiclib -exported_symbols_list @gol
417 -filelist -flat_namespace -force_cpusubtype_ALL @gol
418 -force_flat_namespace -headerpad_max_install_names @gol
419 -image_base -init -install_name -keep_private_externs @gol
420 -multi_module -multiply_defined -multiply_defined_unused @gol
421 -noall_load -no_dead_strip_inits_and_terms @gol
422 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
423 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
424 -private_bundle -read_only_relocs -sectalign @gol
425 -sectobjectsymbols -whyload -seg1addr @gol
426 -sectcreate -sectobjectsymbols -sectorder @gol
427 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
428 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
429 -segprot -segs_read_only_addr -segs_read_write_addr @gol
430 -single_module -static -sub_library -sub_umbrella @gol
431 -twolevel_namespace -umbrella -undefined @gol
432 -unexported_symbols_list -weak_reference_mismatches @gol
433 -whatsloaded -F -gused -gfull -mone-byte-bool}
435 @emph{DEC Alpha Options}
436 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
437 -mieee -mieee-with-inexact -mieee-conformant @gol
438 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
439 -mtrap-precision=@var{mode} -mbuild-constants @gol
440 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
441 -mbwx -mmax -mfix -mcix @gol
442 -mfloat-vax -mfloat-ieee @gol
443 -mexplicit-relocs -msmall-data -mlarge-data @gol
444 -msmall-text -mlarge-text @gol
445 -mmemory-latency=@var{time}}
447 @emph{DEC Alpha/VMS Options}
448 @gccoptlist{-mvms-return-codes}
451 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
452 -mhard-float -msoft-float @gol
453 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
454 -mdouble -mno-double @gol
455 -mmedia -mno-media -mmuladd -mno-muladd @gol
456 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
457 -mlinked-fp -mlong-calls -malign-labels @gol
458 -mlibrary-pic -macc-4 -macc-8 @gol
459 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
460 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
461 -mvliw-branch -mno-vliw-branch @gol
462 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
463 -mno-nested-cond-exec -mtomcat-stats @gol
466 @emph{H8/300 Options}
467 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
470 @gccoptlist{-march=@var{architecture-type} @gol
471 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
472 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
473 -mfixed-range=@var{register-range} @gol
474 -mjump-in-delay -mlinker-opt -mlong-calls @gol
475 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
476 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
477 -mno-jump-in-delay -mno-long-load-store @gol
478 -mno-portable-runtime -mno-soft-float @gol
479 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
480 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
481 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
482 -munix=@var{unix-std} -nolibdld -static -threads}
484 @emph{i386 and x86-64 Options}
485 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
486 -mfpmath=@var{unit} @gol
487 -masm=@var{dialect} -mno-fancy-math-387 @gol
488 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
489 -mno-wide-multiply -mrtd -malign-double @gol
490 -mpreferred-stack-boundary=@var{num} @gol
491 -mmmx -msse -msse2 -msse3 -m3dnow @gol
492 -mthreads -mno-align-stringops -minline-all-stringops @gol
493 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
494 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
495 -mno-red-zone -mno-tls-direct-seg-refs @gol
496 -mcmodel=@var{code-model} @gol
500 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
501 -mvolatile-asm-stop -mb-step -mregister-names -mno-sdata @gol
502 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
503 -minline-float-divide-max-throughput @gol
504 -minline-int-divide-min-latency @gol
505 -minline-int-divide-max-throughput -mno-dwarf2-asm @gol
506 -mfixed-range=@var{register-range}}
508 @emph{M32R/D Options}
509 @gccoptlist{-m32r2 -m32rx -m32r @gol
511 -malign-loops -mno-align-loops @gol
512 -missue-rate=@var{number} @gol
513 -mbranch-cost=@var{number} @gol
514 -mmodel=@var{code-size-model-type} @gol
515 -msdata=@var{sdata-type} @gol
516 -mno-flush-func -mflush-func=@var{name} @gol
517 -mno-flush-trap -mflush-trap=@var{number} @gol
520 @emph{M680x0 Options}
521 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
522 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
523 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
524 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
525 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
527 @emph{M68hc1x Options}
528 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
529 -mauto-incdec -minmax -mlong-calls -mshort @gol
530 -msoft-reg-count=@var{count}}
533 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
534 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
535 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
536 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
537 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
540 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
541 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
542 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
543 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
544 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
545 -mpaired-single -mips3d @gol
546 -mint64 -mlong64 -mlong32 @gol
547 -G@var{num} -membedded-data -mno-embedded-data @gol
548 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
549 -msplit-addresses -mno-split-addresses @gol
550 -mexplicit-relocs -mno-explicit-relocs @gol
551 -mcheck-zero-division -mno-check-zero-division @gol
552 -mdivide-traps -mdivide-breaks @gol
553 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
554 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
555 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
556 -mfix-vr4120 -mno-fix-vr4120 -mfix-sb1 -mno-fix-sb1 @gol
557 -mflush-func=@var{func} -mno-flush-func @gol
558 -mbranch-likely -mno-branch-likely @gol
559 -mfp-exceptions -mno-fp-exceptions @gol
560 -mvr4130-align -mno-vr4130-align}
563 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
564 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
565 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
566 -mno-base-addresses -msingle-exit -mno-single-exit}
568 @emph{MN10300 Options}
569 @gccoptlist{-mmult-bug -mno-mult-bug @gol
570 -mam33 -mno-am33 @gol
571 -mam33-2 -mno-am33-2 @gol
575 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
576 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
577 -mregparam -mnoregparam -msb -mnosb @gol
578 -mbitfield -mnobitfield -mhimem -mnohimem}
580 @emph{PDP-11 Options}
581 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
582 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
583 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
584 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
585 -mbranch-expensive -mbranch-cheap @gol
586 -msplit -mno-split -munix-asm -mdec-asm}
588 @emph{PowerPC Options}
589 See RS/6000 and PowerPC Options.
591 @emph{RS/6000 and PowerPC Options}
592 @gccoptlist{-mcpu=@var{cpu-type} @gol
593 -mtune=@var{cpu-type} @gol
594 -mpower -mno-power -mpower2 -mno-power2 @gol
595 -mpowerpc -mpowerpc64 -mno-powerpc @gol
596 -maltivec -mno-altivec @gol
597 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
598 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
599 -mnew-mnemonics -mold-mnemonics @gol
600 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
601 -m64 -m32 -mxl-call -mno-xl-call -mpe @gol
602 -malign-power -malign-natural @gol
603 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
604 -mstring -mno-string -mupdate -mno-update @gol
605 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
606 -mstrict-align -mno-strict-align -mrelocatable @gol
607 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
608 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
609 -mdynamic-no-pic @gol
610 -mprioritize-restricted-insns=@var{priority} @gol
611 -msched-costly-dep=@var{dependence_type} @gol
612 -minsert-sched-nops=@var{scheme} @gol
613 -mcall-sysv -mcall-netbsd @gol
614 -maix-struct-return -msvr4-struct-return @gol
615 -mabi=altivec -mabi=no-altivec @gol
616 -mabi=spe -mabi=no-spe @gol
617 -misel=yes -misel=no @gol
618 -mspe=yes -mspe=no @gol
619 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
620 -mprototype -mno-prototype @gol
621 -msim -mmvme -mads -myellowknife -memb -msdata @gol
622 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
624 @emph{S/390 and zSeries Options}
625 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
626 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
627 -mpacked-stack -mno-packed-stack @gol
628 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
629 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
630 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
631 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
634 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
635 -m4-nofpu -m4-single-only -m4-single -m4 @gol
636 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
637 -m5-64media -m5-64media-nofpu @gol
638 -m5-32media -m5-32media-nofpu @gol
639 -m5-compact -m5-compact-nofpu @gol
640 -mb -ml -mdalign -mrelax @gol
641 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
642 -mieee -misize -mpadstruct -mspace @gol
643 -mprefergot -musermode}
646 @gccoptlist{-mcpu=@var{cpu-type} @gol
647 -mtune=@var{cpu-type} @gol
648 -mcmodel=@var{code-model} @gol
649 -m32 -m64 -mapp-regs -mno-app-regs @gol
650 -mfaster-structs -mno-faster-structs @gol
651 -mfpu -mno-fpu -mhard-float -msoft-float @gol
652 -mhard-quad-float -msoft-quad-float @gol
653 -mimpure-text -mno-impure-text -mlittle-endian @gol
654 -mstack-bias -mno-stack-bias @gol
655 -munaligned-doubles -mno-unaligned-doubles @gol
656 -mv8plus -mno-v8plus -mvis -mno-vis
659 @emph{System V Options}
660 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
662 @emph{TMS320C3x/C4x Options}
663 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
664 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
665 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
666 -mparallel-insns -mparallel-mpy -mpreserve-float}
669 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
670 -mprolog-function -mno-prolog-function -mspace @gol
671 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
672 -mapp-regs -mno-app-regs @gol
673 -mdisable-callt -mno-disable-callt @gol
679 @gccoptlist{-mg -mgnu -munix}
681 @emph{x86-64 Options}
682 See i386 and x86-64 Options.
684 @emph{Xstormy16 Options}
687 @emph{Xtensa Options}
688 @gccoptlist{-mconst16 -mno-const16 @gol
689 -mfused-madd -mno-fused-madd @gol
690 -mtext-section-literals -mno-text-section-literals @gol
691 -mtarget-align -mno-target-align @gol
692 -mlongcalls -mno-longcalls}
694 @emph{zSeries Options}
695 See S/390 and zSeries Options.
697 @item Code Generation Options
698 @xref{Code Gen Options,,Options for Code Generation Conventions}.
699 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
700 -ffixed-@var{reg} -fexceptions @gol
701 -fnon-call-exceptions -funwind-tables @gol
702 -fasynchronous-unwind-tables @gol
703 -finhibit-size-directive -finstrument-functions @gol
704 -fno-common -fno-ident @gol
705 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
706 -freg-struct-return -fshared-data -fshort-enums @gol
707 -fshort-double -fshort-wchar @gol
708 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
709 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
710 -fargument-alias -fargument-noalias @gol
711 -fargument-noalias-global -fleading-underscore @gol
712 -ftls-model=@var{model} @gol
713 -ftrapv -fwrapv -fbounds-check @gol
718 * Overall Options:: Controlling the kind of output:
719 an executable, object files, assembler files,
720 or preprocessed source.
721 * C Dialect Options:: Controlling the variant of C language compiled.
722 * C++ Dialect Options:: Variations on C++.
723 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
725 * Language Independent Options:: Controlling how diagnostics should be
727 * Warning Options:: How picky should the compiler be?
728 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
729 * Optimize Options:: How much optimization?
730 * Preprocessor Options:: Controlling header files and macro definitions.
731 Also, getting dependency information for Make.
732 * Assembler Options:: Passing options to the assembler.
733 * Link Options:: Specifying libraries and so on.
734 * Directory Options:: Where to find header files and libraries.
735 Where to find the compiler executable files.
736 * Spec Files:: How to pass switches to sub-processes.
737 * Target Options:: Running a cross-compiler, or an old version of GCC.
740 @node Overall Options
741 @section Options Controlling the Kind of Output
743 Compilation can involve up to four stages: preprocessing, compilation
744 proper, assembly and linking, always in that order. GCC is capable of
745 preprocessing and compiling several files either into several
746 assembler input files, or into one assembler input file; then each
747 assembler input file produces an object file, and linking combines all
748 the object files (those newly compiled, and those specified as input)
749 into an executable file.
751 @cindex file name suffix
752 For any given input file, the file name suffix determines what kind of
757 C source code which must be preprocessed.
760 C source code which should not be preprocessed.
763 C++ source code which should not be preprocessed.
766 Objective-C source code. Note that you must link with the @file{libobjc}
767 library to make an Objective-C program work.
770 Objective-C source code which should not be preprocessed.
774 Objective-C++ source code. Note that you must link with the @file{libobjc}
775 library to make an Objective-C++ program work. Note that @samp{.M} refers
776 to a literal capital M@.
779 Objective-C++ source code which should not be preprocessed.
782 C, C++, Objective-C or Objective-C++ header file to be turned into a
787 @itemx @var{file}.cxx
788 @itemx @var{file}.cpp
789 @itemx @var{file}.CPP
790 @itemx @var{file}.c++
792 C++ source code which must be preprocessed. Note that in @samp{.cxx},
793 the last two letters must both be literally @samp{x}. Likewise,
794 @samp{.C} refers to a literal capital C@.
798 C++ header file to be turned into a precompiled header.
801 @itemx @var{file}.for
802 @itemx @var{file}.FOR
803 Fortran source code which should not be preprocessed.
806 @itemx @var{file}.fpp
807 @itemx @var{file}.FPP
808 Fortran source code which must be preprocessed (with the traditional
812 Fortran source code which must be preprocessed with a RATFOR
813 preprocessor (not included with GCC)@.
816 @itemx @var{file}.f95
817 Fortran 90/95 source code which should not be preprocessed.
819 @c FIXME: Descriptions of Java file types.
826 Ada source code file which contains a library unit declaration (a
827 declaration of a package, subprogram, or generic, or a generic
828 instantiation), or a library unit renaming declaration (a package,
829 generic, or subprogram renaming declaration). Such files are also
832 @itemx @var{file}.adb
833 Ada source code file containing a library unit body (a subprogram or
834 package body). Such files are also called @dfn{bodies}.
836 @c GCC also knows about some suffixes for languages not yet included:
845 Assembler code which must be preprocessed.
848 An object file to be fed straight into linking.
849 Any file name with no recognized suffix is treated this way.
853 You can specify the input language explicitly with the @option{-x} option:
856 @item -x @var{language}
857 Specify explicitly the @var{language} for the following input files
858 (rather than letting the compiler choose a default based on the file
859 name suffix). This option applies to all following input files until
860 the next @option{-x} option. Possible values for @var{language} are:
862 c c-header c-cpp-output
863 c++ c++-header c++-cpp-output
864 objective-c objective-c-header objective-c-cpp-output
865 objective-c++ objective-c++-header objective-c++-cpp-output
866 assembler assembler-with-cpp
868 f77 f77-cpp-input ratfor
875 Turn off any specification of a language, so that subsequent files are
876 handled according to their file name suffixes (as they are if @option{-x}
877 has not been used at all).
879 @item -pass-exit-codes
880 @opindex pass-exit-codes
881 Normally the @command{gcc} program will exit with the code of 1 if any
882 phase of the compiler returns a non-success return code. If you specify
883 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
884 numerically highest error produced by any phase that returned an error
888 If you only want some of the stages of compilation, you can use
889 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
890 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
891 @command{gcc} is to stop. Note that some combinations (for example,
892 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
897 Compile or assemble the source files, but do not link. The linking
898 stage simply is not done. The ultimate output is in the form of an
899 object file for each source file.
901 By default, the object file name for a source file is made by replacing
902 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
904 Unrecognized input files, not requiring compilation or assembly, are
909 Stop after the stage of compilation proper; do not assemble. The output
910 is in the form of an assembler code file for each non-assembler input
913 By default, the assembler file name for a source file is made by
914 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
916 Input files that don't require compilation are ignored.
920 Stop after the preprocessing stage; do not run the compiler proper. The
921 output is in the form of preprocessed source code, which is sent to the
924 Input files which don't require preprocessing are ignored.
926 @cindex output file option
929 Place output in file @var{file}. This applies regardless to whatever
930 sort of output is being produced, whether it be an executable file,
931 an object file, an assembler file or preprocessed C code.
933 If @option{-o} is not specified, the default is to put an executable
934 file in @file{a.out}, the object file for
935 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
936 assembler file in @file{@var{source}.s}, a precompiled header file in
937 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
942 Print (on standard error output) the commands executed to run the stages
943 of compilation. Also print the version number of the compiler driver
944 program and of the preprocessor and the compiler proper.
948 Like @option{-v} except the commands are not executed and all command
949 arguments are quoted. This is useful for shell scripts to capture the
950 driver-generated command lines.
954 Use pipes rather than temporary files for communication between the
955 various stages of compilation. This fails to work on some systems where
956 the assembler is unable to read from a pipe; but the GNU assembler has
961 If you are compiling multiple source files, this option tells the driver
962 to pass all the source files to the compiler at once (for those
963 languages for which the compiler can handle this). This will allow
964 intermodule analysis (IMA) to be performed by the compiler. Currently the only
965 language for which this is supported is C@. If you pass source files for
966 multiple languages to the driver, using this option, the driver will invoke
967 the compiler(s) that support IMA once each, passing each compiler all the
968 source files appropriate for it. For those languages that do not support
969 IMA this option will be ignored, and the compiler will be invoked once for
970 each source file in that language. If you use this option in conjunction
971 with @option{-save-temps}, the compiler will generate multiple
973 (one for each source file), but only one (combined) @file{.o} or
978 Print (on the standard output) a description of the command line options
979 understood by @command{gcc}. If the @option{-v} option is also specified
980 then @option{--help} will also be passed on to the various processes
981 invoked by @command{gcc}, so that they can display the command line options
982 they accept. If the @option{-Wextra} option is also specified then command
983 line options which have no documentation associated with them will also
988 Print (on the standard output) a description of target specific command
989 line options for each tool.
993 Display the version number and copyrights of the invoked GCC@.
997 @section Compiling C++ Programs
999 @cindex suffixes for C++ source
1000 @cindex C++ source file suffixes
1001 C++ source files conventionally use one of the suffixes @samp{.C},
1002 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1003 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1004 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1005 files with these names and compiles them as C++ programs even if you
1006 call the compiler the same way as for compiling C programs (usually
1007 with the name @command{gcc}).
1011 However, C++ programs often require class libraries as well as a
1012 compiler that understands the C++ language---and under some
1013 circumstances, you might want to compile programs or header files from
1014 standard input, or otherwise without a suffix that flags them as C++
1015 programs. You might also like to precompile a C header file with a
1016 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1017 program that calls GCC with the default language set to C++, and
1018 automatically specifies linking against the C++ library. On many
1019 systems, @command{g++} is also installed with the name @command{c++}.
1021 @cindex invoking @command{g++}
1022 When you compile C++ programs, you may specify many of the same
1023 command-line options that you use for compiling programs in any
1024 language; or command-line options meaningful for C and related
1025 languages; or options that are meaningful only for C++ programs.
1026 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1027 explanations of options for languages related to C@.
1028 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1029 explanations of options that are meaningful only for C++ programs.
1031 @node C Dialect Options
1032 @section Options Controlling C Dialect
1033 @cindex dialect options
1034 @cindex language dialect options
1035 @cindex options, dialect
1037 The following options control the dialect of C (or languages derived
1038 from C, such as C++, Objective-C and Objective-C++) that the compiler
1042 @cindex ANSI support
1046 In C mode, support all ISO C90 programs. In C++ mode,
1047 remove GNU extensions that conflict with ISO C++.
1049 This turns off certain features of GCC that are incompatible with ISO
1050 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1051 such as the @code{asm} and @code{typeof} keywords, and
1052 predefined macros such as @code{unix} and @code{vax} that identify the
1053 type of system you are using. It also enables the undesirable and
1054 rarely used ISO trigraph feature. For the C compiler,
1055 it disables recognition of C++ style @samp{//} comments as well as
1056 the @code{inline} keyword.
1058 The alternate keywords @code{__asm__}, @code{__extension__},
1059 @code{__inline__} and @code{__typeof__} continue to work despite
1060 @option{-ansi}. You would not want to use them in an ISO C program, of
1061 course, but it is useful to put them in header files that might be included
1062 in compilations done with @option{-ansi}. Alternate predefined macros
1063 such as @code{__unix__} and @code{__vax__} are also available, with or
1064 without @option{-ansi}.
1066 The @option{-ansi} option does not cause non-ISO programs to be
1067 rejected gratuitously. For that, @option{-pedantic} is required in
1068 addition to @option{-ansi}. @xref{Warning Options}.
1070 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1071 option is used. Some header files may notice this macro and refrain
1072 from declaring certain functions or defining certain macros that the
1073 ISO standard doesn't call for; this is to avoid interfering with any
1074 programs that might use these names for other things.
1076 Functions which would normally be built in but do not have semantics
1077 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1078 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1079 built-in functions provided by GCC}, for details of the functions
1084 Determine the language standard. This option is currently only
1085 supported when compiling C or C++. A value for this option must be
1086 provided; possible values are
1091 ISO C90 (same as @option{-ansi}).
1093 @item iso9899:199409
1094 ISO C90 as modified in amendment 1.
1100 ISO C99. Note that this standard is not yet fully supported; see
1101 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1102 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1105 Default, ISO C90 plus GNU extensions (including some C99 features).
1109 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1110 this will become the default. The name @samp{gnu9x} is deprecated.
1113 The 1998 ISO C++ standard plus amendments.
1116 The same as @option{-std=c++98} plus GNU extensions. This is the
1117 default for C++ code.
1120 Even when this option is not specified, you can still use some of the
1121 features of newer standards in so far as they do not conflict with
1122 previous C standards. For example, you may use @code{__restrict__} even
1123 when @option{-std=c99} is not specified.
1125 The @option{-std} options specifying some version of ISO C have the same
1126 effects as @option{-ansi}, except that features that were not in ISO C90
1127 but are in the specified version (for example, @samp{//} comments and
1128 the @code{inline} keyword in ISO C99) are not disabled.
1130 @xref{Standards,,Language Standards Supported by GCC}, for details of
1131 these standard versions.
1133 @item -aux-info @var{filename}
1135 Output to the given filename prototyped declarations for all functions
1136 declared and/or defined in a translation unit, including those in header
1137 files. This option is silently ignored in any language other than C@.
1139 Besides declarations, the file indicates, in comments, the origin of
1140 each declaration (source file and line), whether the declaration was
1141 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1142 @samp{O} for old, respectively, in the first character after the line
1143 number and the colon), and whether it came from a declaration or a
1144 definition (@samp{C} or @samp{F}, respectively, in the following
1145 character). In the case of function definitions, a K&R-style list of
1146 arguments followed by their declarations is also provided, inside
1147 comments, after the declaration.
1151 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1152 keyword, so that code can use these words as identifiers. You can use
1153 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1154 instead. @option{-ansi} implies @option{-fno-asm}.
1156 In C++, this switch only affects the @code{typeof} keyword, since
1157 @code{asm} and @code{inline} are standard keywords. You may want to
1158 use the @option{-fno-gnu-keywords} flag instead, which has the same
1159 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1160 switch only affects the @code{asm} and @code{typeof} keywords, since
1161 @code{inline} is a standard keyword in ISO C99.
1164 @itemx -fno-builtin-@var{function}
1165 @opindex fno-builtin
1166 @cindex built-in functions
1167 Don't recognize built-in functions that do not begin with
1168 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1169 functions provided by GCC}, for details of the functions affected,
1170 including those which are not built-in functions when @option{-ansi} or
1171 @option{-std} options for strict ISO C conformance are used because they
1172 do not have an ISO standard meaning.
1174 GCC normally generates special code to handle certain built-in functions
1175 more efficiently; for instance, calls to @code{alloca} may become single
1176 instructions that adjust the stack directly, and calls to @code{memcpy}
1177 may become inline copy loops. The resulting code is often both smaller
1178 and faster, but since the function calls no longer appear as such, you
1179 cannot set a breakpoint on those calls, nor can you change the behavior
1180 of the functions by linking with a different library. In addition,
1181 when a function is recognized as a built-in function, GCC may use
1182 information about that function to warn about problems with calls to
1183 that function, or to generate more efficient code, even if the
1184 resulting code still contains calls to that function. For example,
1185 warnings are given with @option{-Wformat} for bad calls to
1186 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1187 known not to modify global memory.
1189 With the @option{-fno-builtin-@var{function}} option
1190 only the built-in function @var{function} is
1191 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1192 function is named this is not built-in in this version of GCC, this
1193 option is ignored. There is no corresponding
1194 @option{-fbuiltin-@var{function}} option; if you wish to enable
1195 built-in functions selectively when using @option{-fno-builtin} or
1196 @option{-ffreestanding}, you may define macros such as:
1199 #define abs(n) __builtin_abs ((n))
1200 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1205 @cindex hosted environment
1207 Assert that compilation takes place in a hosted environment. This implies
1208 @option{-fbuiltin}. A hosted environment is one in which the
1209 entire standard library is available, and in which @code{main} has a return
1210 type of @code{int}. Examples are nearly everything except a kernel.
1211 This is equivalent to @option{-fno-freestanding}.
1213 @item -ffreestanding
1214 @opindex ffreestanding
1215 @cindex hosted environment
1217 Assert that compilation takes place in a freestanding environment. This
1218 implies @option{-fno-builtin}. A freestanding environment
1219 is one in which the standard library may not exist, and program startup may
1220 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1221 This is equivalent to @option{-fno-hosted}.
1223 @xref{Standards,,Language Standards Supported by GCC}, for details of
1224 freestanding and hosted environments.
1226 @item -fms-extensions
1227 @opindex fms-extensions
1228 Accept some non-standard constructs used in Microsoft header files.
1230 Some cases of unnamed fields in structures and unions are only
1231 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1232 fields within structs/unions}, for details.
1236 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1237 options for strict ISO C conformance) implies @option{-trigraphs}.
1239 @item -no-integrated-cpp
1240 @opindex no-integrated-cpp
1241 Performs a compilation in two passes: preprocessing and compiling. This
1242 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1243 @option{-B} option. The user supplied compilation step can then add in
1244 an additional preprocessing step after normal preprocessing but before
1245 compiling. The default is to use the integrated cpp (internal cpp)
1247 The semantics of this option will change if "cc1", "cc1plus", and
1248 "cc1obj" are merged.
1250 @cindex traditional C language
1251 @cindex C language, traditional
1253 @itemx -traditional-cpp
1254 @opindex traditional-cpp
1255 @opindex traditional
1256 Formerly, these options caused GCC to attempt to emulate a pre-standard
1257 C compiler. They are now only supported with the @option{-E} switch.
1258 The preprocessor continues to support a pre-standard mode. See the GNU
1259 CPP manual for details.
1261 @item -fcond-mismatch
1262 @opindex fcond-mismatch
1263 Allow conditional expressions with mismatched types in the second and
1264 third arguments. The value of such an expression is void. This option
1265 is not supported for C++.
1267 @item -funsigned-char
1268 @opindex funsigned-char
1269 Let the type @code{char} be unsigned, like @code{unsigned char}.
1271 Each kind of machine has a default for what @code{char} should
1272 be. It is either like @code{unsigned char} by default or like
1273 @code{signed char} by default.
1275 Ideally, a portable program should always use @code{signed char} or
1276 @code{unsigned char} when it depends on the signedness of an object.
1277 But many programs have been written to use plain @code{char} and
1278 expect it to be signed, or expect it to be unsigned, depending on the
1279 machines they were written for. This option, and its inverse, let you
1280 make such a program work with the opposite default.
1282 The type @code{char} is always a distinct type from each of
1283 @code{signed char} or @code{unsigned char}, even though its behavior
1284 is always just like one of those two.
1287 @opindex fsigned-char
1288 Let the type @code{char} be signed, like @code{signed char}.
1290 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1291 the negative form of @option{-funsigned-char}. Likewise, the option
1292 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1294 @item -fsigned-bitfields
1295 @itemx -funsigned-bitfields
1296 @itemx -fno-signed-bitfields
1297 @itemx -fno-unsigned-bitfields
1298 @opindex fsigned-bitfields
1299 @opindex funsigned-bitfields
1300 @opindex fno-signed-bitfields
1301 @opindex fno-unsigned-bitfields
1302 These options control whether a bit-field is signed or unsigned, when the
1303 declaration does not use either @code{signed} or @code{unsigned}. By
1304 default, such a bit-field is signed, because this is consistent: the
1305 basic integer types such as @code{int} are signed types.
1308 @node C++ Dialect Options
1309 @section Options Controlling C++ Dialect
1311 @cindex compiler options, C++
1312 @cindex C++ options, command line
1313 @cindex options, C++
1314 This section describes the command-line options that are only meaningful
1315 for C++ programs; but you can also use most of the GNU compiler options
1316 regardless of what language your program is in. For example, you
1317 might compile a file @code{firstClass.C} like this:
1320 g++ -g -frepo -O -c firstClass.C
1324 In this example, only @option{-frepo} is an option meant
1325 only for C++ programs; you can use the other options with any
1326 language supported by GCC@.
1328 Here is a list of options that are @emph{only} for compiling C++ programs:
1332 @item -fabi-version=@var{n}
1333 @opindex fabi-version
1334 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1335 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1336 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1337 the version that conforms most closely to the C++ ABI specification.
1338 Therefore, the ABI obtained using version 0 will change as ABI bugs
1341 The default is version 2.
1343 @item -fno-access-control
1344 @opindex fno-access-control
1345 Turn off all access checking. This switch is mainly useful for working
1346 around bugs in the access control code.
1350 Check that the pointer returned by @code{operator new} is non-null
1351 before attempting to modify the storage allocated. This check is
1352 normally unnecessary because the C++ standard specifies that
1353 @code{operator new} will only return @code{0} if it is declared
1354 @samp{throw()}, in which case the compiler will always check the
1355 return value even without this option. In all other cases, when
1356 @code{operator new} has a non-empty exception specification, memory
1357 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1358 @samp{new (nothrow)}.
1360 @item -fconserve-space
1361 @opindex fconserve-space
1362 Put uninitialized or runtime-initialized global variables into the
1363 common segment, as C does. This saves space in the executable at the
1364 cost of not diagnosing duplicate definitions. If you compile with this
1365 flag and your program mysteriously crashes after @code{main()} has
1366 completed, you may have an object that is being destroyed twice because
1367 two definitions were merged.
1369 This option is no longer useful on most targets, now that support has
1370 been added for putting variables into BSS without making them common.
1372 @item -fno-const-strings
1373 @opindex fno-const-strings
1374 Give string constants type @code{char *} instead of type @code{const
1375 char *}. By default, G++ uses type @code{const char *} as required by
1376 the standard. Even if you use @option{-fno-const-strings}, you cannot
1377 actually modify the value of a string constant.
1379 This option might be removed in a future release of G++. For maximum
1380 portability, you should structure your code so that it works with
1381 string constants that have type @code{const char *}.
1383 @item -fno-elide-constructors
1384 @opindex fno-elide-constructors
1385 The C++ standard allows an implementation to omit creating a temporary
1386 which is only used to initialize another object of the same type.
1387 Specifying this option disables that optimization, and forces G++ to
1388 call the copy constructor in all cases.
1390 @item -fno-enforce-eh-specs
1391 @opindex fno-enforce-eh-specs
1392 Don't check for violation of exception specifications at runtime. This
1393 option violates the C++ standard, but may be useful for reducing code
1394 size in production builds, much like defining @samp{NDEBUG}. The compiler
1395 will still optimize based on the exception specifications.
1398 @itemx -fno-for-scope
1400 @opindex fno-for-scope
1401 If @option{-ffor-scope} is specified, the scope of variables declared in
1402 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1403 as specified by the C++ standard.
1404 If @option{-fno-for-scope} is specified, the scope of variables declared in
1405 a @i{for-init-statement} extends to the end of the enclosing scope,
1406 as was the case in old versions of G++, and other (traditional)
1407 implementations of C++.
1409 The default if neither flag is given to follow the standard,
1410 but to allow and give a warning for old-style code that would
1411 otherwise be invalid, or have different behavior.
1413 @item -fno-gnu-keywords
1414 @opindex fno-gnu-keywords
1415 Do not recognize @code{typeof} as a keyword, so that code can use this
1416 word as an identifier. You can use the keyword @code{__typeof__} instead.
1417 @option{-ansi} implies @option{-fno-gnu-keywords}.
1419 @item -fno-implicit-templates
1420 @opindex fno-implicit-templates
1421 Never emit code for non-inline templates which are instantiated
1422 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1423 @xref{Template Instantiation}, for more information.
1425 @item -fno-implicit-inline-templates
1426 @opindex fno-implicit-inline-templates
1427 Don't emit code for implicit instantiations of inline templates, either.
1428 The default is to handle inlines differently so that compiles with and
1429 without optimization will need the same set of explicit instantiations.
1431 @item -fno-implement-inlines
1432 @opindex fno-implement-inlines
1433 To save space, do not emit out-of-line copies of inline functions
1434 controlled by @samp{#pragma implementation}. This will cause linker
1435 errors if these functions are not inlined everywhere they are called.
1437 @item -fms-extensions
1438 @opindex fms-extensions
1439 Disable pedantic warnings about constructs used in MFC, such as implicit
1440 int and getting a pointer to member function via non-standard syntax.
1442 @item -fno-nonansi-builtins
1443 @opindex fno-nonansi-builtins
1444 Disable built-in declarations of functions that are not mandated by
1445 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1446 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1448 @item -fno-operator-names
1449 @opindex fno-operator-names
1450 Do not treat the operator name keywords @code{and}, @code{bitand},
1451 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1452 synonyms as keywords.
1454 @item -fno-optional-diags
1455 @opindex fno-optional-diags
1456 Disable diagnostics that the standard says a compiler does not need to
1457 issue. Currently, the only such diagnostic issued by G++ is the one for
1458 a name having multiple meanings within a class.
1461 @opindex fpermissive
1462 Downgrade some diagnostics about nonconformant code from errors to
1463 warnings. Thus, using @option{-fpermissive} will allow some
1464 nonconforming code to compile.
1468 Enable automatic template instantiation at link time. This option also
1469 implies @option{-fno-implicit-templates}. @xref{Template
1470 Instantiation}, for more information.
1474 Disable generation of information about every class with virtual
1475 functions for use by the C++ runtime type identification features
1476 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1477 of the language, you can save some space by using this flag. Note that
1478 exception handling uses the same information, but it will generate it as
1483 Emit statistics about front-end processing at the end of the compilation.
1484 This information is generally only useful to the G++ development team.
1486 @item -ftemplate-depth-@var{n}
1487 @opindex ftemplate-depth
1488 Set the maximum instantiation depth for template classes to @var{n}.
1489 A limit on the template instantiation depth is needed to detect
1490 endless recursions during template class instantiation. ANSI/ISO C++
1491 conforming programs must not rely on a maximum depth greater than 17.
1493 @item -fno-threadsafe-statics
1494 @opindex fno-threadsafe-statics
1495 Do not emit the extra code to use the routines specified in the C++
1496 ABI for thread-safe initialization of local statics. You can use this
1497 option to reduce code size slightly in code that doesn't need to be
1500 @item -fuse-cxa-atexit
1501 @opindex fuse-cxa-atexit
1502 Register destructors for objects with static storage duration with the
1503 @code{__cxa_atexit} function rather than the @code{atexit} function.
1504 This option is required for fully standards-compliant handling of static
1505 destructors, but will only work if your C library supports
1506 @code{__cxa_atexit}.
1508 @item -fvisibility-inlines-hidden
1509 @opindex fvisibility-inlines-hidden
1510 Causes all inlined methods to be marked with
1511 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1512 appear in the export table of a DSO and do not require a PLT indirection
1513 when used within the DSO@. Enabling this option can have a dramatic effect
1514 on load and link times of a DSO as it massively reduces the size of the
1515 dynamic export table when the library makes heavy use of templates. While
1516 it can cause bloating through duplication of code within each DSO where
1517 it is used, often the wastage is less than the considerable space occupied
1518 by a long symbol name in the export table which is typical when using
1519 templates and namespaces. For even more savings, combine with the
1520 @option{-fvisibility=hidden} switch.
1524 Do not use weak symbol support, even if it is provided by the linker.
1525 By default, G++ will use weak symbols if they are available. This
1526 option exists only for testing, and should not be used by end-users;
1527 it will result in inferior code and has no benefits. This option may
1528 be removed in a future release of G++.
1532 Do not search for header files in the standard directories specific to
1533 C++, but do still search the other standard directories. (This option
1534 is used when building the C++ library.)
1537 In addition, these optimization, warning, and code generation options
1538 have meanings only for C++ programs:
1541 @item -fno-default-inline
1542 @opindex fno-default-inline
1543 Do not assume @samp{inline} for functions defined inside a class scope.
1544 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1545 functions will have linkage like inline functions; they just won't be
1548 @item -Wabi @r{(C++ only)}
1550 Warn when G++ generates code that is probably not compatible with the
1551 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1552 all such cases, there are probably some cases that are not warned about,
1553 even though G++ is generating incompatible code. There may also be
1554 cases where warnings are emitted even though the code that is generated
1557 You should rewrite your code to avoid these warnings if you are
1558 concerned about the fact that code generated by G++ may not be binary
1559 compatible with code generated by other compilers.
1561 The known incompatibilities at this point include:
1566 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1567 pack data into the same byte as a base class. For example:
1570 struct A @{ virtual void f(); int f1 : 1; @};
1571 struct B : public A @{ int f2 : 1; @};
1575 In this case, G++ will place @code{B::f2} into the same byte
1576 as@code{A::f1}; other compilers will not. You can avoid this problem
1577 by explicitly padding @code{A} so that its size is a multiple of the
1578 byte size on your platform; that will cause G++ and other compilers to
1579 layout @code{B} identically.
1582 Incorrect handling of tail-padding for virtual bases. G++ does not use
1583 tail padding when laying out virtual bases. For example:
1586 struct A @{ virtual void f(); char c1; @};
1587 struct B @{ B(); char c2; @};
1588 struct C : public A, public virtual B @{@};
1592 In this case, G++ will not place @code{B} into the tail-padding for
1593 @code{A}; other compilers will. You can avoid this problem by
1594 explicitly padding @code{A} so that its size is a multiple of its
1595 alignment (ignoring virtual base classes); that will cause G++ and other
1596 compilers to layout @code{C} identically.
1599 Incorrect handling of bit-fields with declared widths greater than that
1600 of their underlying types, when the bit-fields appear in a union. For
1604 union U @{ int i : 4096; @};
1608 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1609 union too small by the number of bits in an @code{int}.
1612 Empty classes can be placed at incorrect offsets. For example:
1622 struct C : public B, public A @{@};
1626 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1627 it should be placed at offset zero. G++ mistakenly believes that the
1628 @code{A} data member of @code{B} is already at offset zero.
1631 Names of template functions whose types involve @code{typename} or
1632 template template parameters can be mangled incorrectly.
1635 template <typename Q>
1636 void f(typename Q::X) @{@}
1638 template <template <typename> class Q>
1639 void f(typename Q<int>::X) @{@}
1643 Instantiations of these templates may be mangled incorrectly.
1647 @item -Wctor-dtor-privacy @r{(C++ only)}
1648 @opindex Wctor-dtor-privacy
1649 Warn when a class seems unusable because all the constructors or
1650 destructors in that class are private, and it has neither friends nor
1651 public static member functions.
1653 @item -Wnon-virtual-dtor @r{(C++ only)}
1654 @opindex Wnon-virtual-dtor
1655 Warn when a class appears to be polymorphic, thereby requiring a virtual
1656 destructor, yet it declares a non-virtual one.
1657 This warning is enabled by @option{-Wall}.
1659 @item -Wreorder @r{(C++ only)}
1661 @cindex reordering, warning
1662 @cindex warning for reordering of member initializers
1663 Warn when the order of member initializers given in the code does not
1664 match the order in which they must be executed. For instance:
1670 A(): j (0), i (1) @{ @}
1674 The compiler will rearrange the member initializers for @samp{i}
1675 and @samp{j} to match the declaration order of the members, emitting
1676 a warning to that effect. This warning is enabled by @option{-Wall}.
1679 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1682 @item -Weffc++ @r{(C++ only)}
1684 Warn about violations of the following style guidelines from Scott Meyers'
1685 @cite{Effective C++} book:
1689 Item 11: Define a copy constructor and an assignment operator for classes
1690 with dynamically allocated memory.
1693 Item 12: Prefer initialization to assignment in constructors.
1696 Item 14: Make destructors virtual in base classes.
1699 Item 15: Have @code{operator=} return a reference to @code{*this}.
1702 Item 23: Don't try to return a reference when you must return an object.
1706 Also warn about violations of the following style guidelines from
1707 Scott Meyers' @cite{More Effective C++} book:
1711 Item 6: Distinguish between prefix and postfix forms of increment and
1712 decrement operators.
1715 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1719 When selecting this option, be aware that the standard library
1720 headers do not obey all of these guidelines; use @samp{grep -v}
1721 to filter out those warnings.
1723 @item -Wno-deprecated @r{(C++ only)}
1724 @opindex Wno-deprecated
1725 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1727 @item -Wno-non-template-friend @r{(C++ only)}
1728 @opindex Wno-non-template-friend
1729 Disable warnings when non-templatized friend functions are declared
1730 within a template. Since the advent of explicit template specification
1731 support in G++, if the name of the friend is an unqualified-id (i.e.,
1732 @samp{friend foo(int)}), the C++ language specification demands that the
1733 friend declare or define an ordinary, nontemplate function. (Section
1734 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1735 could be interpreted as a particular specialization of a templatized
1736 function. Because this non-conforming behavior is no longer the default
1737 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1738 check existing code for potential trouble spots and is on by default.
1739 This new compiler behavior can be turned off with
1740 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1741 but disables the helpful warning.
1743 @item -Wold-style-cast @r{(C++ only)}
1744 @opindex Wold-style-cast
1745 Warn if an old-style (C-style) cast to a non-void type is used within
1746 a C++ program. The new-style casts (@samp{static_cast},
1747 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1748 unintended effects and much easier to search for.
1750 @item -Woverloaded-virtual @r{(C++ only)}
1751 @opindex Woverloaded-virtual
1752 @cindex overloaded virtual fn, warning
1753 @cindex warning for overloaded virtual fn
1754 Warn when a function declaration hides virtual functions from a
1755 base class. For example, in:
1762 struct B: public A @{
1767 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1775 will fail to compile.
1777 @item -Wno-pmf-conversions @r{(C++ only)}
1778 @opindex Wno-pmf-conversions
1779 Disable the diagnostic for converting a bound pointer to member function
1782 @item -Wsign-promo @r{(C++ only)}
1783 @opindex Wsign-promo
1784 Warn when overload resolution chooses a promotion from unsigned or
1785 enumerated type to a signed type, over a conversion to an unsigned type of
1786 the same size. Previous versions of G++ would try to preserve
1787 unsignedness, but the standard mandates the current behavior.
1789 @item -Wsynth @r{(C++ only)}
1791 @cindex warning for synthesized methods
1792 @cindex synthesized methods, warning
1793 Warn when G++'s synthesis behavior does not match that of cfront. For
1799 A& operator = (int);
1809 In this example, G++ will synthesize a default @samp{A& operator =
1810 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1813 @node Objective-C and Objective-C++ Dialect Options
1814 @section Options Controlling Objective-C and Objective-C++ Dialects
1816 @cindex compiler options, Objective-C and Objective-C++
1817 @cindex Objective-C and Objective-C++ options, command line
1818 @cindex options, Objective-C and Objective-C++
1819 (NOTE: This manual does not describe the Objective-C and Objective-C++
1820 languages themselves. See @xref{Standards,,Language Standards
1821 Supported by GCC}, for references.)
1823 This section describes the command-line options that are only meaningful
1824 for Objective-C and Objective-C++ programs, but you can also use most of
1825 the language-independent GNU compiler options.
1826 For example, you might compile a file @code{some_class.m} like this:
1829 gcc -g -fgnu-runtime -O -c some_class.m
1833 In this example, @option{-fgnu-runtime} is an option meant only for
1834 Objective-C and Objective-C++ programs; you can use the other options with
1835 any language supported by GCC@.
1837 Note that since Objective-C is an extension of the C language, Objective-C
1838 compilations may also use options specific to the C front-end (e.g.,
1839 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1840 C++-specific options (e.g., @option{-Wabi}).
1842 Here is a list of options that are @emph{only} for compiling Objective-C
1843 and Objective-C++ programs:
1846 @item -fconstant-string-class=@var{class-name}
1847 @opindex fconstant-string-class
1848 Use @var{class-name} as the name of the class to instantiate for each
1849 literal string specified with the syntax @code{@@"@dots{}"}. The default
1850 class name is @code{NXConstantString} if the GNU runtime is being used, and
1851 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1852 @option{-fconstant-cfstrings} option, if also present, will override the
1853 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1854 to be laid out as constant CoreFoundation strings.
1857 @opindex fgnu-runtime
1858 Generate object code compatible with the standard GNU Objective-C
1859 runtime. This is the default for most types of systems.
1861 @item -fnext-runtime
1862 @opindex fnext-runtime
1863 Generate output compatible with the NeXT runtime. This is the default
1864 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1865 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1868 @item -fno-nil-receivers
1869 @opindex fno-nil-receivers
1870 Assume that all Objective-C message dispatches (e.g.,
1871 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1872 is not @code{nil}. This allows for more efficient entry points in the runtime
1873 to be used. Currently, this option is only available in conjunction with
1874 the NeXT runtime on Mac OS X 10.3 and later.
1876 @item -fobjc-exceptions
1877 @opindex fobjc-exceptions
1878 Enable syntactic support for structured exception handling in Objective-C,
1879 similar to what is offered by C++ and Java. Currently, this option is only
1880 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1888 @@catch (AnObjCClass *exc) @{
1895 @@catch (AnotherClass *exc) @{
1898 @@catch (id allOthers) @{
1908 The @code{@@throw} statement may appear anywhere in an Objective-C or
1909 Objective-C++ program; when used inside of a @code{@@catch} block, the
1910 @code{@@throw} may appear without an argument (as shown above), in which case
1911 the object caught by the @code{@@catch} will be rethrown.
1913 Note that only (pointers to) Objective-C objects may be thrown and
1914 caught using this scheme. When an object is thrown, it will be caught
1915 by the nearest @code{@@catch} clause capable of handling objects of that type,
1916 analogously to how @code{catch} blocks work in C++ and Java. A
1917 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1918 any and all Objective-C exceptions not caught by previous @code{@@catch}
1921 The @code{@@finally} clause, if present, will be executed upon exit from the
1922 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1923 regardless of whether any exceptions are thrown, caught or rethrown
1924 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1925 of the @code{finally} clause in Java.
1927 There are several caveats to using the new exception mechanism:
1931 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1932 idioms provided by the @code{NSException} class, the new
1933 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1934 systems, due to additional functionality needed in the (NeXT) Objective-C
1938 As mentioned above, the new exceptions do not support handling
1939 types other than Objective-C objects. Furthermore, when used from
1940 Objective-C++, the Objective-C exception model does not interoperate with C++
1941 exceptions at this time. This means you cannot @code{@@throw} an exception
1942 from Objective-C and @code{catch} it in C++, or vice versa
1943 (i.e., @code{throw @dots{} @@catch}).
1946 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1947 blocks for thread-safe execution:
1950 @@synchronized (ObjCClass *guard) @{
1955 Upon entering the @code{@@synchronized} block, a thread of execution shall
1956 first check whether a lock has been placed on the corresponding @code{guard}
1957 object by another thread. If it has, the current thread shall wait until
1958 the other thread relinquishes its lock. Once @code{guard} becomes available,
1959 the current thread will place its own lock on it, execute the code contained in
1960 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1961 making @code{guard} available to other threads).
1963 Unlike Java, Objective-C does not allow for entire methods to be marked
1964 @code{@@synchronized}. Note that throwing exceptions out of
1965 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1966 to be unlocked properly.
1968 @item -freplace-objc-classes
1969 @opindex freplace-objc-classes
1970 Emit a special marker instructing @command{ld(1)} not to statically link in
1971 the resulting object file, and allow @command{dyld(1)} to load it in at
1972 run time instead. This is used in conjunction with the Fix-and-Continue
1973 debugging mode, where the object file in question may be recompiled and
1974 dynamically reloaded in the course of program execution, without the need
1975 to restart the program itself. Currently, Fix-and-Continue functionality
1976 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1981 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1982 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1983 compile time) with static class references that get initialized at load time,
1984 which improves run-time performance. Specifying the @option{-fzero-link} flag
1985 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1986 to be retained. This is useful in Zero-Link debugging mode, since it allows
1987 for individual class implementations to be modified during program execution.
1991 Dump interface declarations for all classes seen in the source file to a
1992 file named @file{@var{sourcename}.decl}.
1995 @opindex Wno-protocol
1996 If a class is declared to implement a protocol, a warning is issued for
1997 every method in the protocol that is not implemented by the class. The
1998 default behavior is to issue a warning for every method not explicitly
1999 implemented in the class, even if a method implementation is inherited
2000 from the superclass. If you use the @option{-Wno-protocol} option, then
2001 methods inherited from the superclass are considered to be implemented,
2002 and no warning is issued for them.
2006 Warn if multiple methods of different types for the same selector are
2007 found during compilation. The check is performed on the list of methods
2008 in the final stage of compilation. Additionally, a check is performed
2009 for each selector appearing in a @code{@@selector(@dots{})}
2010 expression, and a corresponding method for that selector has been found
2011 during compilation. Because these checks scan the method table only at
2012 the end of compilation, these warnings are not produced if the final
2013 stage of compilation is not reached, for example because an error is
2014 found during compilation, or because the @option{-fsyntax-only} option is
2017 @item -Wundeclared-selector
2018 @opindex Wundeclared-selector
2019 Warn if a @code{@@selector(@dots{})} expression referring to an
2020 undeclared selector is found. A selector is considered undeclared if no
2021 method with that name has been declared before the
2022 @code{@@selector(@dots{})} expression, either explicitly in an
2023 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2024 an @code{@@implementation} section. This option always performs its
2025 checks as soon as a @code{@@selector(@dots{})} expression is found,
2026 while @option{-Wselector} only performs its checks in the final stage of
2027 compilation. This also enforces the coding style convention
2028 that methods and selectors must be declared before being used.
2030 @item -print-objc-runtime-info
2031 @opindex print-objc-runtime-info
2032 Generate C header describing the largest structure that is passed by
2037 @node Language Independent Options
2038 @section Options to Control Diagnostic Messages Formatting
2039 @cindex options to control diagnostics formatting
2040 @cindex diagnostic messages
2041 @cindex message formatting
2043 Traditionally, diagnostic messages have been formatted irrespective of
2044 the output device's aspect (e.g.@: its width, @dots{}). The options described
2045 below can be used to control the diagnostic messages formatting
2046 algorithm, e.g.@: how many characters per line, how often source location
2047 information should be reported. Right now, only the C++ front end can
2048 honor these options. However it is expected, in the near future, that
2049 the remaining front ends would be able to digest them correctly.
2052 @item -fmessage-length=@var{n}
2053 @opindex fmessage-length
2054 Try to format error messages so that they fit on lines of about @var{n}
2055 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2056 the front ends supported by GCC@. If @var{n} is zero, then no
2057 line-wrapping will be done; each error message will appear on a single
2060 @opindex fdiagnostics-show-location
2061 @item -fdiagnostics-show-location=once
2062 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2063 reporter to emit @emph{once} source location information; that is, in
2064 case the message is too long to fit on a single physical line and has to
2065 be wrapped, the source location won't be emitted (as prefix) again,
2066 over and over, in subsequent continuation lines. This is the default
2069 @item -fdiagnostics-show-location=every-line
2070 Only meaningful in line-wrapping mode. Instructs the diagnostic
2071 messages reporter to emit the same source location information (as
2072 prefix) for physical lines that result from the process of breaking
2073 a message which is too long to fit on a single line.
2077 @node Warning Options
2078 @section Options to Request or Suppress Warnings
2079 @cindex options to control warnings
2080 @cindex warning messages
2081 @cindex messages, warning
2082 @cindex suppressing warnings
2084 Warnings are diagnostic messages that report constructions which
2085 are not inherently erroneous but which are risky or suggest there
2086 may have been an error.
2088 You can request many specific warnings with options beginning @samp{-W},
2089 for example @option{-Wimplicit} to request warnings on implicit
2090 declarations. Each of these specific warning options also has a
2091 negative form beginning @samp{-Wno-} to turn off warnings;
2092 for example, @option{-Wno-implicit}. This manual lists only one of the
2093 two forms, whichever is not the default.
2095 The following options control the amount and kinds of warnings produced
2096 by GCC; for further, language-specific options also refer to
2097 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2101 @cindex syntax checking
2103 @opindex fsyntax-only
2104 Check the code for syntax errors, but don't do anything beyond that.
2108 Issue all the warnings demanded by strict ISO C and ISO C++;
2109 reject all programs that use forbidden extensions, and some other
2110 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2111 version of the ISO C standard specified by any @option{-std} option used.
2113 Valid ISO C and ISO C++ programs should compile properly with or without
2114 this option (though a rare few will require @option{-ansi} or a
2115 @option{-std} option specifying the required version of ISO C)@. However,
2116 without this option, certain GNU extensions and traditional C and C++
2117 features are supported as well. With this option, they are rejected.
2119 @option{-pedantic} does not cause warning messages for use of the
2120 alternate keywords whose names begin and end with @samp{__}. Pedantic
2121 warnings are also disabled in the expression that follows
2122 @code{__extension__}. However, only system header files should use
2123 these escape routes; application programs should avoid them.
2124 @xref{Alternate Keywords}.
2126 Some users try to use @option{-pedantic} to check programs for strict ISO
2127 C conformance. They soon find that it does not do quite what they want:
2128 it finds some non-ISO practices, but not all---only those for which
2129 ISO C @emph{requires} a diagnostic, and some others for which
2130 diagnostics have been added.
2132 A feature to report any failure to conform to ISO C might be useful in
2133 some instances, but would require considerable additional work and would
2134 be quite different from @option{-pedantic}. We don't have plans to
2135 support such a feature in the near future.
2137 Where the standard specified with @option{-std} represents a GNU
2138 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2139 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2140 extended dialect is based. Warnings from @option{-pedantic} are given
2141 where they are required by the base standard. (It would not make sense
2142 for such warnings to be given only for features not in the specified GNU
2143 C dialect, since by definition the GNU dialects of C include all
2144 features the compiler supports with the given option, and there would be
2145 nothing to warn about.)
2147 @item -pedantic-errors
2148 @opindex pedantic-errors
2149 Like @option{-pedantic}, except that errors are produced rather than
2154 Inhibit all warning messages.
2158 Inhibit warning messages about the use of @samp{#import}.
2160 @item -Wchar-subscripts
2161 @opindex Wchar-subscripts
2162 Warn if an array subscript has type @code{char}. This is a common cause
2163 of error, as programmers often forget that this type is signed on some
2168 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2169 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2171 @item -Wfatal-errors
2172 @opindex Wfatal-errors
2173 This option causes the compiler to abort compilation on the first error
2174 occurred rather than trying to keep going and printing further error
2179 @opindex ffreestanding
2180 @opindex fno-builtin
2181 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2182 the arguments supplied have types appropriate to the format string
2183 specified, and that the conversions specified in the format string make
2184 sense. This includes standard functions, and others specified by format
2185 attributes (@pxref{Function Attributes}), in the @code{printf},
2186 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2187 not in the C standard) families (or other target-specific families).
2188 Which functions are checked without format attributes having been
2189 specified depends on the standard version selected, and such checks of
2190 functions without the attribute specified are disabled by
2191 @option{-ffreestanding} or @option{-fno-builtin}.
2193 The formats are checked against the format features supported by GNU
2194 libc version 2.2. These include all ISO C90 and C99 features, as well
2195 as features from the Single Unix Specification and some BSD and GNU
2196 extensions. Other library implementations may not support all these
2197 features; GCC does not support warning about features that go beyond a
2198 particular library's limitations. However, if @option{-pedantic} is used
2199 with @option{-Wformat}, warnings will be given about format features not
2200 in the selected standard version (but not for @code{strfmon} formats,
2201 since those are not in any version of the C standard). @xref{C Dialect
2202 Options,,Options Controlling C Dialect}.
2204 Since @option{-Wformat} also checks for null format arguments for
2205 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2207 @option{-Wformat} is included in @option{-Wall}. For more control over some
2208 aspects of format checking, the options @option{-Wformat-y2k},
2209 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2210 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2211 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2214 @opindex Wformat-y2k
2215 If @option{-Wformat} is specified, also warn about @code{strftime}
2216 formats which may yield only a two-digit year.
2218 @item -Wno-format-extra-args
2219 @opindex Wno-format-extra-args
2220 If @option{-Wformat} is specified, do not warn about excess arguments to a
2221 @code{printf} or @code{scanf} format function. The C standard specifies
2222 that such arguments are ignored.
2224 Where the unused arguments lie between used arguments that are
2225 specified with @samp{$} operand number specifications, normally
2226 warnings are still given, since the implementation could not know what
2227 type to pass to @code{va_arg} to skip the unused arguments. However,
2228 in the case of @code{scanf} formats, this option will suppress the
2229 warning if the unused arguments are all pointers, since the Single
2230 Unix Specification says that such unused arguments are allowed.
2232 @item -Wno-format-zero-length
2233 @opindex Wno-format-zero-length
2234 If @option{-Wformat} is specified, do not warn about zero-length formats.
2235 The C standard specifies that zero-length formats are allowed.
2237 @item -Wformat-nonliteral
2238 @opindex Wformat-nonliteral
2239 If @option{-Wformat} is specified, also warn if the format string is not a
2240 string literal and so cannot be checked, unless the format function
2241 takes its format arguments as a @code{va_list}.
2243 @item -Wformat-security
2244 @opindex Wformat-security
2245 If @option{-Wformat} is specified, also warn about uses of format
2246 functions that represent possible security problems. At present, this
2247 warns about calls to @code{printf} and @code{scanf} functions where the
2248 format string is not a string literal and there are no format arguments,
2249 as in @code{printf (foo);}. This may be a security hole if the format
2250 string came from untrusted input and contains @samp{%n}. (This is
2251 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2252 in future warnings may be added to @option{-Wformat-security} that are not
2253 included in @option{-Wformat-nonliteral}.)
2257 Enable @option{-Wformat} plus format checks not included in
2258 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2259 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2263 Warn about passing a null pointer for arguments marked as
2264 requiring a non-null value by the @code{nonnull} function attribute.
2266 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2267 can be disabled with the @option{-Wno-nonnull} option.
2269 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2271 Warn about uninitialized variables which are initialized with themselves.
2272 Note this option can only be used with the @option{-Wuninitialized} option,
2273 which in turn only works with @option{-O1} and above.
2275 For example, GCC will warn about @code{i} being uninitialized in the
2276 following snippet only when @option{-Winit-self} has been specified:
2287 @item -Wimplicit-int
2288 @opindex Wimplicit-int
2289 Warn when a declaration does not specify a type.
2291 @item -Wimplicit-function-declaration
2292 @itemx -Werror-implicit-function-declaration
2293 @opindex Wimplicit-function-declaration
2294 @opindex Werror-implicit-function-declaration
2295 Give a warning (or error) whenever a function is used before being
2296 declared. The form @option{-Wno-error-implicit-function-declaration}
2301 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2305 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2306 function with external linkage, returning int, taking either zero
2307 arguments, two, or three arguments of appropriate types.
2309 @item -Wmissing-braces
2310 @opindex Wmissing-braces
2311 Warn if an aggregate or union initializer is not fully bracketed. In
2312 the following example, the initializer for @samp{a} is not fully
2313 bracketed, but that for @samp{b} is fully bracketed.
2316 int a[2][2] = @{ 0, 1, 2, 3 @};
2317 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2320 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2321 @opindex Wmissing-include-dirs
2322 Warn if a user-supplied include directory does not exist.
2325 @opindex Wparentheses
2326 Warn if parentheses are omitted in certain contexts, such
2327 as when there is an assignment in a context where a truth value
2328 is expected, or when operators are nested whose precedence people
2329 often get confused about. Only the warning for an assignment used as
2330 a truth value is supported when compiling C++; the other warnings are
2331 only supported when compiling C@.
2333 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2334 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2335 interpretation from that of ordinary mathematical notation.
2337 Also warn about constructions where there may be confusion to which
2338 @code{if} statement an @code{else} branch belongs. Here is an example of
2353 In C, every @code{else} branch belongs to the innermost possible @code{if}
2354 statement, which in this example is @code{if (b)}. This is often not
2355 what the programmer expected, as illustrated in the above example by
2356 indentation the programmer chose. When there is the potential for this
2357 confusion, GCC will issue a warning when this flag is specified.
2358 To eliminate the warning, add explicit braces around the innermost
2359 @code{if} statement so there is no way the @code{else} could belong to
2360 the enclosing @code{if}. The resulting code would look like this:
2376 @item -Wsequence-point
2377 @opindex Wsequence-point
2378 Warn about code that may have undefined semantics because of violations
2379 of sequence point rules in the C standard.
2381 The C standard defines the order in which expressions in a C program are
2382 evaluated in terms of @dfn{sequence points}, which represent a partial
2383 ordering between the execution of parts of the program: those executed
2384 before the sequence point, and those executed after it. These occur
2385 after the evaluation of a full expression (one which is not part of a
2386 larger expression), after the evaluation of the first operand of a
2387 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2388 function is called (but after the evaluation of its arguments and the
2389 expression denoting the called function), and in certain other places.
2390 Other than as expressed by the sequence point rules, the order of
2391 evaluation of subexpressions of an expression is not specified. All
2392 these rules describe only a partial order rather than a total order,
2393 since, for example, if two functions are called within one expression
2394 with no sequence point between them, the order in which the functions
2395 are called is not specified. However, the standards committee have
2396 ruled that function calls do not overlap.
2398 It is not specified when between sequence points modifications to the
2399 values of objects take effect. Programs whose behavior depends on this
2400 have undefined behavior; the C standard specifies that ``Between the
2401 previous and next sequence point an object shall have its stored value
2402 modified at most once by the evaluation of an expression. Furthermore,
2403 the prior value shall be read only to determine the value to be
2404 stored.''. If a program breaks these rules, the results on any
2405 particular implementation are entirely unpredictable.
2407 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2408 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2409 diagnosed by this option, and it may give an occasional false positive
2410 result, but in general it has been found fairly effective at detecting
2411 this sort of problem in programs.
2413 The present implementation of this option only works for C programs. A
2414 future implementation may also work for C++ programs.
2416 The C standard is worded confusingly, therefore there is some debate
2417 over the precise meaning of the sequence point rules in subtle cases.
2418 Links to discussions of the problem, including proposed formal
2419 definitions, may be found on the GCC readings page, at
2420 @w{@uref{http://gcc.gnu.org/readings.html}}.
2423 @opindex Wreturn-type
2424 Warn whenever a function is defined with a return-type that defaults to
2425 @code{int}. Also warn about any @code{return} statement with no
2426 return-value in a function whose return-type is not @code{void}.
2428 For C, also warn if the return type of a function has a type qualifier
2429 such as @code{const}. Such a type qualifier has no effect, since the
2430 value returned by a function is not an lvalue. ISO C prohibits
2431 qualified @code{void} return types on function definitions, so such
2432 return types always receive a warning even without this option.
2434 For C++, a function without return type always produces a diagnostic
2435 message, even when @option{-Wno-return-type} is specified. The only
2436 exceptions are @samp{main} and functions defined in system headers.
2440 Warn whenever a @code{switch} statement has an index of enumerated type
2441 and lacks a @code{case} for one or more of the named codes of that
2442 enumeration. (The presence of a @code{default} label prevents this
2443 warning.) @code{case} labels outside the enumeration range also
2444 provoke warnings when this option is used.
2446 @item -Wswitch-default
2447 @opindex Wswitch-switch
2448 Warn whenever a @code{switch} statement does not have a @code{default}
2452 @opindex Wswitch-enum
2453 Warn whenever a @code{switch} statement has an index of enumerated type
2454 and lacks a @code{case} for one or more of the named codes of that
2455 enumeration. @code{case} labels outside the enumeration range also
2456 provoke warnings when this option is used.
2460 Warn if any trigraphs are encountered that might change the meaning of
2461 the program (trigraphs within comments are not warned about).
2463 @item -Wunused-function
2464 @opindex Wunused-function
2465 Warn whenever a static function is declared but not defined or a
2466 non\-inline static function is unused.
2468 @item -Wunused-label
2469 @opindex Wunused-label
2470 Warn whenever a label is declared but not used.
2472 To suppress this warning use the @samp{unused} attribute
2473 (@pxref{Variable Attributes}).
2475 @item -Wunused-parameter
2476 @opindex Wunused-parameter
2477 Warn whenever a function parameter is unused aside from its declaration.
2479 To suppress this warning use the @samp{unused} attribute
2480 (@pxref{Variable Attributes}).
2482 @item -Wunused-variable
2483 @opindex Wunused-variable
2484 Warn whenever a local variable or non-constant static variable is unused
2485 aside from its declaration
2487 To suppress this warning use the @samp{unused} attribute
2488 (@pxref{Variable Attributes}).
2490 @item -Wunused-value
2491 @opindex Wunused-value
2492 Warn whenever a statement computes a result that is explicitly not used.
2494 To suppress this warning cast the expression to @samp{void}.
2498 All the above @option{-Wunused} options combined.
2500 In order to get a warning about an unused function parameter, you must
2501 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2502 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2504 @item -Wuninitialized
2505 @opindex Wuninitialized
2506 Warn if an automatic variable is used without first being initialized or
2507 if a variable may be clobbered by a @code{setjmp} call.
2509 These warnings are possible only in optimizing compilation,
2510 because they require data flow information that is computed only
2511 when optimizing. If you don't specify @option{-O}, you simply won't
2514 If you want to warn about code which uses the uninitialized value of the
2515 variable in its own initializer, use the @option{-Winit-self} option.
2517 These warnings occur only for variables that are candidates for
2518 register allocation. Therefore, they do not occur for a variable that
2519 is declared @code{volatile}, or whose address is taken, or whose size
2520 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2521 structures, unions or arrays, even when they are in registers.
2523 Note that there may be no warning about a variable that is used only
2524 to compute a value that itself is never used, because such
2525 computations may be deleted by data flow analysis before the warnings
2528 These warnings are made optional because GCC is not smart
2529 enough to see all the reasons why the code might be correct
2530 despite appearing to have an error. Here is one example of how
2551 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2552 always initialized, but GCC doesn't know this. Here is
2553 another common case:
2558 if (change_y) save_y = y, y = new_y;
2560 if (change_y) y = save_y;
2565 This has no bug because @code{save_y} is used only if it is set.
2567 @cindex @code{longjmp} warnings
2568 This option also warns when a non-volatile automatic variable might be
2569 changed by a call to @code{longjmp}. These warnings as well are possible
2570 only in optimizing compilation.
2572 The compiler sees only the calls to @code{setjmp}. It cannot know
2573 where @code{longjmp} will be called; in fact, a signal handler could
2574 call it at any point in the code. As a result, you may get a warning
2575 even when there is in fact no problem because @code{longjmp} cannot
2576 in fact be called at the place which would cause a problem.
2578 Some spurious warnings can be avoided if you declare all the functions
2579 you use that never return as @code{noreturn}. @xref{Function
2582 @item -Wunknown-pragmas
2583 @opindex Wunknown-pragmas
2584 @cindex warning for unknown pragmas
2585 @cindex unknown pragmas, warning
2586 @cindex pragmas, warning of unknown
2587 Warn when a #pragma directive is encountered which is not understood by
2588 GCC@. If this command line option is used, warnings will even be issued
2589 for unknown pragmas in system header files. This is not the case if
2590 the warnings were only enabled by the @option{-Wall} command line option.
2592 @item -Wstrict-aliasing
2593 @opindex Wstrict-aliasing
2594 This option is only active when @option{-fstrict-aliasing} is active.
2595 It warns about code which might break the strict aliasing rules that the
2596 compiler is using for optimization. The warning does not catch all
2597 cases, but does attempt to catch the more common pitfalls. It is
2598 included in @option{-Wall}.
2600 @item -Wstrict-aliasing=2
2601 @opindex Wstrict-aliasing=2
2602 This option is only active when @option{-fstrict-aliasing} is active.
2603 It warns about all code which might break the strict aliasing rules that the
2604 compiler is using for optimization. This warning catches all cases, but
2605 it will also give a warning for some ambiguous cases that are safe.
2609 All of the above @samp{-W} options combined. This enables all the
2610 warnings about constructions that some users consider questionable, and
2611 that are easy to avoid (or modify to prevent the warning), even in
2612 conjunction with macros. This also enables some language-specific
2613 warnings described in @ref{C++ Dialect Options} and
2614 @ref{Objective-C and Objective-C++ Dialect Options}.
2617 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2618 Some of them warn about constructions that users generally do not
2619 consider questionable, but which occasionally you might wish to check
2620 for; others warn about constructions that are necessary or hard to avoid
2621 in some cases, and there is no simple way to modify the code to suppress
2628 (This option used to be called @option{-W}. The older name is still
2629 supported, but the newer name is more descriptive.) Print extra warning
2630 messages for these events:
2634 A function can return either with or without a value. (Falling
2635 off the end of the function body is considered returning without
2636 a value.) For example, this function would evoke such a
2650 An expression-statement or the left-hand side of a comma expression
2651 contains no side effects.
2652 To suppress the warning, cast the unused expression to void.
2653 For example, an expression such as @samp{x[i,j]} will cause a warning,
2654 but @samp{x[(void)i,j]} will not.
2657 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2660 Storage-class specifiers like @code{static} are not the first things in
2661 a declaration. According to the C Standard, this usage is obsolescent.
2664 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2668 A comparison between signed and unsigned values could produce an
2669 incorrect result when the signed value is converted to unsigned.
2670 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2673 An aggregate has an initializer which does not initialize all members.
2674 This warning can be independently controlled by
2675 @option{-Wmissing-field-initializers}.
2678 A function parameter is declared without a type specifier in K&R-style
2686 An empty body occurs in an @samp{if} or @samp{else} statement.
2689 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2690 @samp{>}, or @samp{>=}.
2693 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2696 Any of several floating-point events that often indicate errors, such as
2697 overflow, underflow, loss of precision, etc.
2699 @item @r{(C++ only)}
2700 An enumerator and a non-enumerator both appear in a conditional expression.
2702 @item @r{(C++ only)}
2703 A non-static reference or non-static @samp{const} member appears in a
2704 class without constructors.
2706 @item @r{(C++ only)}
2707 Ambiguous virtual bases.
2709 @item @r{(C++ only)}
2710 Subscripting an array which has been declared @samp{register}.
2712 @item @r{(C++ only)}
2713 Taking the address of a variable which has been declared @samp{register}.
2715 @item @r{(C++ only)}
2716 A base class is not initialized in a derived class' copy constructor.
2719 @item -Wno-div-by-zero
2720 @opindex Wno-div-by-zero
2721 @opindex Wdiv-by-zero
2722 Do not warn about compile-time integer division by zero. Floating point
2723 division by zero is not warned about, as it can be a legitimate way of
2724 obtaining infinities and NaNs.
2726 @item -Wsystem-headers
2727 @opindex Wsystem-headers
2728 @cindex warnings from system headers
2729 @cindex system headers, warnings from
2730 Print warning messages for constructs found in system header files.
2731 Warnings from system headers are normally suppressed, on the assumption
2732 that they usually do not indicate real problems and would only make the
2733 compiler output harder to read. Using this command line option tells
2734 GCC to emit warnings from system headers as if they occurred in user
2735 code. However, note that using @option{-Wall} in conjunction with this
2736 option will @emph{not} warn about unknown pragmas in system
2737 headers---for that, @option{-Wunknown-pragmas} must also be used.
2740 @opindex Wfloat-equal
2741 Warn if floating point values are used in equality comparisons.
2743 The idea behind this is that sometimes it is convenient (for the
2744 programmer) to consider floating-point values as approximations to
2745 infinitely precise real numbers. If you are doing this, then you need
2746 to compute (by analyzing the code, or in some other way) the maximum or
2747 likely maximum error that the computation introduces, and allow for it
2748 when performing comparisons (and when producing output, but that's a
2749 different problem). In particular, instead of testing for equality, you
2750 would check to see whether the two values have ranges that overlap; and
2751 this is done with the relational operators, so equality comparisons are
2754 @item -Wtraditional @r{(C only)}
2755 @opindex Wtraditional
2756 Warn about certain constructs that behave differently in traditional and
2757 ISO C@. Also warn about ISO C constructs that have no traditional C
2758 equivalent, and/or problematic constructs which should be avoided.
2762 Macro parameters that appear within string literals in the macro body.
2763 In traditional C macro replacement takes place within string literals,
2764 but does not in ISO C@.
2767 In traditional C, some preprocessor directives did not exist.
2768 Traditional preprocessors would only consider a line to be a directive
2769 if the @samp{#} appeared in column 1 on the line. Therefore
2770 @option{-Wtraditional} warns about directives that traditional C
2771 understands but would ignore because the @samp{#} does not appear as the
2772 first character on the line. It also suggests you hide directives like
2773 @samp{#pragma} not understood by traditional C by indenting them. Some
2774 traditional implementations would not recognize @samp{#elif}, so it
2775 suggests avoiding it altogether.
2778 A function-like macro that appears without arguments.
2781 The unary plus operator.
2784 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2785 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2786 constants.) Note, these suffixes appear in macros defined in the system
2787 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2788 Use of these macros in user code might normally lead to spurious
2789 warnings, however GCC's integrated preprocessor has enough context to
2790 avoid warning in these cases.
2793 A function declared external in one block and then used after the end of
2797 A @code{switch} statement has an operand of type @code{long}.
2800 A non-@code{static} function declaration follows a @code{static} one.
2801 This construct is not accepted by some traditional C compilers.
2804 The ISO type of an integer constant has a different width or
2805 signedness from its traditional type. This warning is only issued if
2806 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2807 typically represent bit patterns, are not warned about.
2810 Usage of ISO string concatenation is detected.
2813 Initialization of automatic aggregates.
2816 Identifier conflicts with labels. Traditional C lacks a separate
2817 namespace for labels.
2820 Initialization of unions. If the initializer is zero, the warning is
2821 omitted. This is done under the assumption that the zero initializer in
2822 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2823 initializer warnings and relies on default initialization to zero in the
2827 Conversions by prototypes between fixed/floating point values and vice
2828 versa. The absence of these prototypes when compiling with traditional
2829 C would cause serious problems. This is a subset of the possible
2830 conversion warnings, for the full set use @option{-Wconversion}.
2833 Use of ISO C style function definitions. This warning intentionally is
2834 @emph{not} issued for prototype declarations or variadic functions
2835 because these ISO C features will appear in your code when using
2836 libiberty's traditional C compatibility macros, @code{PARAMS} and
2837 @code{VPARAMS}. This warning is also bypassed for nested functions
2838 because that feature is already a GCC extension and thus not relevant to
2839 traditional C compatibility.
2842 @item -Wdeclaration-after-statement @r{(C only)}
2843 @opindex Wdeclaration-after-statement
2844 Warn when a declaration is found after a statement in a block. This
2845 construct, known from C++, was introduced with ISO C99 and is by default
2846 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2847 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2851 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2853 @item -Wendif-labels
2854 @opindex Wendif-labels
2855 Warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2859 Warn whenever a local variable shadows another local variable, parameter or
2860 global variable or whenever a built-in function is shadowed.
2862 @item -Wlarger-than-@var{len}
2863 @opindex Wlarger-than
2864 Warn whenever an object of larger than @var{len} bytes is defined.
2866 @item -Wpointer-arith
2867 @opindex Wpointer-arith
2868 Warn about anything that depends on the ``size of'' a function type or
2869 of @code{void}. GNU C assigns these types a size of 1, for
2870 convenience in calculations with @code{void *} pointers and pointers
2873 @item -Wbad-function-cast @r{(C only)}
2874 @opindex Wbad-function-cast
2875 Warn whenever a function call is cast to a non-matching type.
2876 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2880 Warn whenever a pointer is cast so as to remove a type qualifier from
2881 the target type. For example, warn if a @code{const char *} is cast
2882 to an ordinary @code{char *}.
2885 @opindex Wcast-align
2886 Warn whenever a pointer is cast such that the required alignment of the
2887 target is increased. For example, warn if a @code{char *} is cast to
2888 an @code{int *} on machines where integers can only be accessed at
2889 two- or four-byte boundaries.
2891 @item -Wwrite-strings
2892 @opindex Wwrite-strings
2893 When compiling C, give string constants the type @code{const
2894 char[@var{length}]} so that
2895 copying the address of one into a non-@code{const} @code{char *}
2896 pointer will get a warning; when compiling C++, warn about the
2897 deprecated conversion from string constants to @code{char *}.
2898 These warnings will help you find at
2899 compile time code that can try to write into a string constant, but
2900 only if you have been very careful about using @code{const} in
2901 declarations and prototypes. Otherwise, it will just be a nuisance;
2902 this is why we did not make @option{-Wall} request these warnings.
2905 @opindex Wconversion
2906 Warn if a prototype causes a type conversion that is different from what
2907 would happen to the same argument in the absence of a prototype. This
2908 includes conversions of fixed point to floating and vice versa, and
2909 conversions changing the width or signedness of a fixed point argument
2910 except when the same as the default promotion.
2912 Also, warn if a negative integer constant expression is implicitly
2913 converted to an unsigned type. For example, warn about the assignment
2914 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2915 casts like @code{(unsigned) -1}.
2917 @item -Wsign-compare
2918 @opindex Wsign-compare
2919 @cindex warning for comparison of signed and unsigned values
2920 @cindex comparison of signed and unsigned values, warning
2921 @cindex signed and unsigned values, comparison warning
2922 Warn when a comparison between signed and unsigned values could produce
2923 an incorrect result when the signed value is converted to unsigned.
2924 This warning is also enabled by @option{-Wextra}; to get the other warnings
2925 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2927 @item -Waggregate-return
2928 @opindex Waggregate-return
2929 Warn if any functions that return structures or unions are defined or
2930 called. (In languages where you can return an array, this also elicits
2933 @item -Wstrict-prototypes @r{(C only)}
2934 @opindex Wstrict-prototypes
2935 Warn if a function is declared or defined without specifying the
2936 argument types. (An old-style function definition is permitted without
2937 a warning if preceded by a declaration which specifies the argument
2940 @item -Wold-style-definition @r{(C only)}
2941 @opindex Wold-style-definition
2942 Warn if an old-style function definition is used. A warning is given
2943 even if there is a previous prototype.
2945 @item -Wmissing-prototypes @r{(C only)}
2946 @opindex Wmissing-prototypes
2947 Warn if a global function is defined without a previous prototype
2948 declaration. This warning is issued even if the definition itself
2949 provides a prototype. The aim is to detect global functions that fail
2950 to be declared in header files.
2952 @item -Wmissing-declarations @r{(C only)}
2953 @opindex Wmissing-declarations
2954 Warn if a global function is defined without a previous declaration.
2955 Do so even if the definition itself provides a prototype.
2956 Use this option to detect global functions that are not declared in
2959 @item -Wmissing-field-initializers
2960 @opindex Wmissing-field-initializers
2963 Warn if a structure's initializer has some fields missing. For
2964 example, the following code would cause such a warning, because
2965 @code{x.h} is implicitly zero:
2968 struct s @{ int f, g, h; @};
2969 struct s x = @{ 3, 4 @};
2972 This option does not warn about designated initializers, so the following
2973 modification would not trigger a warning:
2976 struct s @{ int f, g, h; @};
2977 struct s x = @{ .f = 3, .g = 4 @};
2980 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
2981 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
2983 @item -Wmissing-noreturn
2984 @opindex Wmissing-noreturn
2985 Warn about functions which might be candidates for attribute @code{noreturn}.
2986 Note these are only possible candidates, not absolute ones. Care should
2987 be taken to manually verify functions actually do not ever return before
2988 adding the @code{noreturn} attribute, otherwise subtle code generation
2989 bugs could be introduced. You will not get a warning for @code{main} in
2990 hosted C environments.
2992 @item -Wmissing-format-attribute
2993 @opindex Wmissing-format-attribute
2995 If @option{-Wformat} is enabled, also warn about functions which might be
2996 candidates for @code{format} attributes. Note these are only possible
2997 candidates, not absolute ones. GCC will guess that @code{format}
2998 attributes might be appropriate for any function that calls a function
2999 like @code{vprintf} or @code{vscanf}, but this might not always be the
3000 case, and some functions for which @code{format} attributes are
3001 appropriate may not be detected. This option has no effect unless
3002 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3004 @item -Wno-multichar
3005 @opindex Wno-multichar
3007 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3008 Usually they indicate a typo in the user's code, as they have
3009 implementation-defined values, and should not be used in portable code.
3011 @item -Wno-deprecated-declarations
3012 @opindex Wno-deprecated-declarations
3013 Do not warn about uses of functions, variables, and types marked as
3014 deprecated by using the @code{deprecated} attribute.
3015 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3016 @pxref{Type Attributes}.)
3020 Warn if a structure is given the packed attribute, but the packed
3021 attribute has no effect on the layout or size of the structure.
3022 Such structures may be mis-aligned for little benefit. For
3023 instance, in this code, the variable @code{f.x} in @code{struct bar}
3024 will be misaligned even though @code{struct bar} does not itself
3025 have the packed attribute:
3032 @} __attribute__((packed));
3042 Warn if padding is included in a structure, either to align an element
3043 of the structure or to align the whole structure. Sometimes when this
3044 happens it is possible to rearrange the fields of the structure to
3045 reduce the padding and so make the structure smaller.
3047 @item -Wredundant-decls
3048 @opindex Wredundant-decls
3049 Warn if anything is declared more than once in the same scope, even in
3050 cases where multiple declaration is valid and changes nothing.
3052 @item -Wnested-externs @r{(C only)}
3053 @opindex Wnested-externs
3054 Warn if an @code{extern} declaration is encountered within a function.
3056 @item -Wunreachable-code
3057 @opindex Wunreachable-code
3058 Warn if the compiler detects that code will never be executed.
3060 This option is intended to warn when the compiler detects that at
3061 least a whole line of source code will never be executed, because
3062 some condition is never satisfied or because it is after a
3063 procedure that never returns.
3065 It is possible for this option to produce a warning even though there
3066 are circumstances under which part of the affected line can be executed,
3067 so care should be taken when removing apparently-unreachable code.
3069 For instance, when a function is inlined, a warning may mean that the
3070 line is unreachable in only one inlined copy of the function.
3072 This option is not made part of @option{-Wall} because in a debugging
3073 version of a program there is often substantial code which checks
3074 correct functioning of the program and is, hopefully, unreachable
3075 because the program does work. Another common use of unreachable
3076 code is to provide behavior which is selectable at compile-time.
3080 Warn if a function can not be inlined and it was declared as inline.
3081 Even with this option, the compiler will not warn about failures to
3082 inline functions declared in system headers.
3084 The compiler uses a variety of heuristics to determine whether or not
3085 to inline a function. For example, the compiler takes into account
3086 the size of the function being inlined and the the amount of inlining
3087 that has already been done in the current function. Therefore,
3088 seemingly insignificant changes in the source program can cause the
3089 warnings produced by @option{-Winline} to appear or disappear.
3091 @item -Wno-invalid-offsetof @r{(C++ only)}
3092 @opindex Wno-invalid-offsetof
3093 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3094 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3095 to a non-POD type is undefined. In existing C++ implementations,
3096 however, @samp{offsetof} typically gives meaningful results even when
3097 applied to certain kinds of non-POD types. (Such as a simple
3098 @samp{struct} that fails to be a POD type only by virtue of having a
3099 constructor.) This flag is for users who are aware that they are
3100 writing nonportable code and who have deliberately chosen to ignore the
3103 The restrictions on @samp{offsetof} may be relaxed in a future version
3104 of the C++ standard.
3107 @opindex Winvalid-pch
3108 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3109 the search path but can't be used.
3113 @opindex Wno-long-long
3114 Warn if @samp{long long} type is used. This is default. To inhibit
3115 the warning messages, use @option{-Wno-long-long}. Flags
3116 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3117 only when @option{-pedantic} flag is used.
3119 @item -Wvariadic-macros
3120 @opindex Wvariadic-macros
3121 @opindex Wno-variadic-macros
3122 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3123 alternate syntax when in pedantic ISO C99 mode. This is default.
3124 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3126 @item -Wdisabled-optimization
3127 @opindex Wdisabled-optimization
3128 Warn if a requested optimization pass is disabled. This warning does
3129 not generally indicate that there is anything wrong with your code; it
3130 merely indicates that GCC's optimizers were unable to handle the code
3131 effectively. Often, the problem is that your code is too big or too
3132 complex; GCC will refuse to optimize programs when the optimization
3133 itself is likely to take inordinate amounts of time.
3137 Make all warnings into errors.
3140 @node Debugging Options
3141 @section Options for Debugging Your Program or GCC
3142 @cindex options, debugging
3143 @cindex debugging information options
3145 GCC has various special options that are used for debugging
3146 either your program or GCC:
3151 Produce debugging information in the operating system's native format
3152 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3155 On most systems that use stabs format, @option{-g} enables use of extra
3156 debugging information that only GDB can use; this extra information
3157 makes debugging work better in GDB but will probably make other debuggers
3159 refuse to read the program. If you want to control for certain whether
3160 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3161 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3163 GCC allows you to use @option{-g} with
3164 @option{-O}. The shortcuts taken by optimized code may occasionally
3165 produce surprising results: some variables you declared may not exist
3166 at all; flow of control may briefly move where you did not expect it;
3167 some statements may not be executed because they compute constant
3168 results or their values were already at hand; some statements may
3169 execute in different places because they were moved out of loops.
3171 Nevertheless it proves possible to debug optimized output. This makes
3172 it reasonable to use the optimizer for programs that might have bugs.
3174 The following options are useful when GCC is generated with the
3175 capability for more than one debugging format.
3179 Produce debugging information for use by GDB@. This means to use the
3180 most expressive format available (DWARF 2, stabs, or the native format
3181 if neither of those are supported), including GDB extensions if at all
3186 Produce debugging information in stabs format (if that is supported),
3187 without GDB extensions. This is the format used by DBX on most BSD
3188 systems. On MIPS, Alpha and System V Release 4 systems this option
3189 produces stabs debugging output which is not understood by DBX or SDB@.
3190 On System V Release 4 systems this option requires the GNU assembler.
3192 @item -feliminate-unused-debug-symbols
3193 @opindex feliminate-unused-debug-symbols
3194 Produce debugging information in stabs format (if that is supported),
3195 for only symbols that are actually used.
3199 Produce debugging information in stabs format (if that is supported),
3200 using GNU extensions understood only by the GNU debugger (GDB)@. The
3201 use of these extensions is likely to make other debuggers crash or
3202 refuse to read the program.
3206 Produce debugging information in COFF format (if that is supported).
3207 This is the format used by SDB on most System V systems prior to
3212 Produce debugging information in XCOFF format (if that is supported).
3213 This is the format used by the DBX debugger on IBM RS/6000 systems.
3217 Produce debugging information in XCOFF format (if that is supported),
3218 using GNU extensions understood only by the GNU debugger (GDB)@. The
3219 use of these extensions is likely to make other debuggers crash or
3220 refuse to read the program, and may cause assemblers other than the GNU
3221 assembler (GAS) to fail with an error.
3225 Produce debugging information in DWARF version 2 format (if that is
3226 supported). This is the format used by DBX on IRIX 6. With this
3227 option, GCC uses features of DWARF version 3 when they are useful;
3228 version 3 is upward compatible with version 2, but may still cause
3229 problems for older debuggers.
3233 Produce debugging information in VMS debug format (if that is
3234 supported). This is the format used by DEBUG on VMS systems.
3237 @itemx -ggdb@var{level}
3238 @itemx -gstabs@var{level}
3239 @itemx -gcoff@var{level}
3240 @itemx -gxcoff@var{level}
3241 @itemx -gvms@var{level}
3242 Request debugging information and also use @var{level} to specify how
3243 much information. The default level is 2.
3245 Level 1 produces minimal information, enough for making backtraces in
3246 parts of the program that you don't plan to debug. This includes
3247 descriptions of functions and external variables, but no information
3248 about local variables and no line numbers.
3250 Level 3 includes extra information, such as all the macro definitions
3251 present in the program. Some debuggers support macro expansion when
3252 you use @option{-g3}.
3254 @option{-gdwarf-2} does not accept a concatenated debug level, because
3255 GCC used to support an option @option{-gdwarf} that meant to generate
3256 debug information in version 1 of the DWARF format (which is very
3257 different from version 2), and it would have been too confusing. That
3258 debug format is long obsolete, but the option cannot be changed now.
3259 Instead use an additional @option{-g@var{level}} option to change the
3260 debug level for DWARF2.
3262 @item -feliminate-dwarf2-dups
3263 @opindex feliminate-dwarf2-dups
3264 Compress DWARF2 debugging information by eliminating duplicated
3265 information about each symbol. This option only makes sense when
3266 generating DWARF2 debugging information with @option{-gdwarf-2}.
3268 @cindex @command{prof}
3271 Generate extra code to write profile information suitable for the
3272 analysis program @command{prof}. You must use this option when compiling
3273 the source files you want data about, and you must also use it when
3276 @cindex @command{gprof}
3279 Generate extra code to write profile information suitable for the
3280 analysis program @command{gprof}. You must use this option when compiling
3281 the source files you want data about, and you must also use it when
3286 Makes the compiler print out each function name as it is compiled, and
3287 print some statistics about each pass when it finishes.
3290 @opindex ftime-report
3291 Makes the compiler print some statistics about the time consumed by each
3292 pass when it finishes.
3295 @opindex fmem-report
3296 Makes the compiler print some statistics about permanent memory
3297 allocation when it finishes.
3299 @item -fprofile-arcs
3300 @opindex fprofile-arcs
3301 Add code so that program flow @dfn{arcs} are instrumented. During
3302 execution the program records how many times each branch and call is
3303 executed and how many times it is taken or returns. When the compiled
3304 program exits it saves this data to a file called
3305 @file{@var{auxname}.gcda} for each source file. The data may be used for
3306 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3307 test coverage analysis (@option{-ftest-coverage}). Each object file's
3308 @var{auxname} is generated from the name of the output file, if
3309 explicitly specified and it is not the final executable, otherwise it is
3310 the basename of the source file. In both cases any suffix is removed
3311 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3312 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3317 Compile the source files with @option{-fprofile-arcs} plus optimization
3318 and code generation options. For test coverage analysis, use the
3319 additional @option{-ftest-coverage} option. You do not need to profile
3320 every source file in a program.
3323 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3324 (the latter implies the former).
3327 Run the program on a representative workload to generate the arc profile
3328 information. This may be repeated any number of times. You can run
3329 concurrent instances of your program, and provided that the file system
3330 supports locking, the data files will be correctly updated. Also
3331 @code{fork} calls are detected and correctly handled (double counting
3335 For profile-directed optimizations, compile the source files again with
3336 the same optimization and code generation options plus
3337 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3338 Control Optimization}).
3341 For test coverage analysis, use @command{gcov} to produce human readable
3342 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3343 @command{gcov} documentation for further information.
3347 With @option{-fprofile-arcs}, for each function of your program GCC
3348 creates a program flow graph, then finds a spanning tree for the graph.
3349 Only arcs that are not on the spanning tree have to be instrumented: the
3350 compiler adds code to count the number of times that these arcs are
3351 executed. When an arc is the only exit or only entrance to a block, the
3352 instrumentation code can be added to the block; otherwise, a new basic
3353 block must be created to hold the instrumentation code.
3355 @item -ftree-based-profiling
3356 @opindex ftree-based-profiling
3357 This option is used in addition to @option{-fprofile-arcs} or
3358 @option{-fbranch-probabilities} to control whether those optimizations
3359 are performed on a tree-based or rtl-based internal representation.
3360 If you use this option when compiling with @option{-fprofile-arcs},
3361 you must also use it when compiling later with @option{-fbranch-probabilities}.
3362 Currently the tree-based optimization is in an early stage of
3363 development, and this option is recommended only for those people
3364 working on improving it.
3367 @item -ftest-coverage
3368 @opindex ftest-coverage
3369 Produce a notes file that the @command{gcov} code-coverage utility
3370 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3371 show program coverage. Each source file's note file is called
3372 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3373 above for a description of @var{auxname} and instructions on how to
3374 generate test coverage data. Coverage data will match the source files
3375 more closely, if you do not optimize.
3377 @item -d@var{letters}
3378 @item -fdump-rtl-@var{pass}
3380 Says to make debugging dumps during compilation at times specified by
3381 @var{letters}. This is used for debugging the RTL-based passes of the
3382 compiler. The file names for most of the dumps are made by appending a
3383 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3384 from the name of the output file, if explicitly specified and it is not
3385 an executable, otherwise it is the basename of the source file.
3387 Most debug dumps can be enabled either passing a letter to the @option{-d}
3388 option, or with a long @option{-fdump-rtl} switch; here are the possible
3389 letters for use in @var{letters} and @var{pass}, and their meanings:
3394 Annotate the assembler output with miscellaneous debugging information.
3397 @itemx -fdump-rtl-bp
3399 @opindex fdump-rtl-bp
3400 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3403 @itemx -fdump-rtl-bbro
3405 @opindex fdump-rtl-bbro
3406 Dump after block reordering, to @file{@var{file}.30.bbro}.
3409 @itemx -fdump-rtl-combine
3411 @opindex fdump-rtl-combine
3412 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3415 @itemx -fdump-rtl-ce1
3416 @itemx -fdump-rtl-ce2
3418 @opindex fdump-rtl-ce1
3419 @opindex fdump-rtl-ce2
3420 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3421 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3422 and @option{-fdump-rtl-ce2} enable dumping after the second if
3423 conversion, to the file @file{@var{file}.18.ce2}.
3426 @itemx -fdump-rtl-btl
3427 @itemx -fdump-rtl-dbr
3429 @opindex fdump-rtl-btl
3430 @opindex fdump-rtl-dbr
3431 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3432 target load optimization, to to @file{@var{file}.31.btl}. @option{-dd}
3433 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3434 scheduling, to @file{@var{file}.36.dbr}.
3438 Dump all macro definitions, at the end of preprocessing, in addition to
3442 @itemx -fdump-rtl-ce3
3444 @opindex fdump-rtl-ce3
3445 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3448 @itemx -fdump-rtl-cfg
3449 @itemx -fdump-rtl-life
3451 @opindex fdump-rtl-cfg
3452 @opindex fdump-rtl-life
3453 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3454 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3455 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3456 to @file{@var{file}.16.life}.
3459 @itemx -fdump-rtl-greg
3461 @opindex fdump-rtl-greg
3462 Dump after global register allocation, to @file{@var{file}.23.greg}.
3465 @itemx -fdump-rtl-gcse
3466 @itemx -fdump-rtl-bypass
3468 @opindex fdump-rtl-gcse
3469 @opindex fdump-rtl-bypass
3470 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3471 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3472 enable dumping after jump bypassing and control flow optimizations, to
3473 @file{@var{file}.07.bypass}.
3476 @itemx -fdump-rtl-eh
3478 @opindex fdump-rtl-eh
3479 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3482 @itemx -fdump-rtl-sibling
3484 @opindex fdump-rtl-sibling
3485 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3488 @itemx -fdump-rtl-jump
3490 @opindex fdump-rtl-jump
3491 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3494 @itemx -fdump-rtl-stack
3496 @opindex fdump-rtl-stack
3497 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3500 @itemx -fdump-rtl-lreg
3502 @opindex fdump-rtl-lreg
3503 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3506 @itemx -fdump-rtl-loop
3507 @itemx -fdump-rtl-loop2
3509 @opindex fdump-rtl-loop
3510 @opindex fdump-rtl-loop2
3511 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3512 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3513 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3514 @file{@var{file}.13.loop2}.
3517 @itemx -fdump-rtl-sms
3519 @opindex fdump-rtl-sms
3520 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3523 @itemx -fdump-rtl-mach
3525 @opindex fdump-rtl-mach
3526 Dump after performing the machine dependent reorganization pass, to
3527 @file{@var{file}.35.mach}.
3530 @itemx -fdump-rtl-rnreg
3532 @opindex fdump-rtl-rnreg
3533 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3536 @itemx -fdump-rtl-regmove
3538 @opindex fdump-rtl-regmove
3539 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3542 @itemx -fdump-rtl-postreload
3544 @opindex fdump-rtl-postreload
3545 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3548 @itemx -fdump-rtl-expand
3550 @opindex fdump-rtl-expand
3551 Dump after RTL generation, to @file{@var{file}.00.expand}.
3554 @itemx -fdump-rtl-sched2
3556 @opindex fdump-rtl-sched2
3557 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3560 @itemx -fdump-rtl-cse
3562 @opindex fdump-rtl-cse
3563 Dump after CSE (including the jump optimization that sometimes follows
3564 CSE), to @file{@var{file}.04.cse}.
3567 @itemx -fdump-rtl-sched
3569 @opindex fdump-rtl-sched
3570 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3573 @itemx -fdump-rtl-cse2
3575 @opindex fdump-rtl-cse2
3576 Dump after the second CSE pass (including the jump optimization that
3577 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3580 @itemx -fdump-rtl-tracer
3582 @opindex fdump-rtl-tracer
3583 Dump after running tracer, to @file{@var{file}.12.tracer}.
3586 @itemx -fdump-rtl-vpt
3587 @itemx -fdump-rtl-vartrack
3589 @opindex fdump-rtl-vpt
3590 @opindex fdump-rtl-vartrack
3591 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3592 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3593 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3594 to @file{@var{file}.34.vartrack}.
3597 @itemx -fdump-rtl-flow2
3599 @opindex fdump-rtl-flow2
3600 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3603 @itemx -fdump-rtl-peephole2
3605 @opindex fdump-rtl-peephole2
3606 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3609 @itemx -fdump-rtl-web
3611 @opindex fdump-rtl-web
3612 Dump after live range splitting, to @file{@var{file}.14.web}.
3615 @itemx -fdump-rtl-all
3617 @opindex fdump-rtl-all
3618 Produce all the dumps listed above.
3622 Produce a core dump whenever an error occurs.
3626 Print statistics on memory usage, at the end of the run, to
3631 Annotate the assembler output with a comment indicating which
3632 pattern and alternative was used. The length of each instruction is
3637 Dump the RTL in the assembler output as a comment before each instruction.
3638 Also turns on @option{-dp} annotation.
3642 For each of the other indicated dump files (either with @option{-d} or
3643 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3644 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3648 Just generate RTL for a function instead of compiling it. Usually used
3649 with @samp{r} (@option{-fdump-rtl-expand}).
3653 Dump debugging information during parsing, to standard error.
3656 @item -fdump-unnumbered
3657 @opindex fdump-unnumbered
3658 When doing debugging dumps (see @option{-d} option above), suppress instruction
3659 numbers and line number note output. This makes it more feasible to
3660 use diff on debugging dumps for compiler invocations with different
3661 options, in particular with and without @option{-g}.
3663 @item -fdump-translation-unit @r{(C and C++ only)}
3664 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3665 @opindex fdump-translation-unit
3666 Dump a representation of the tree structure for the entire translation
3667 unit to a file. The file name is made by appending @file{.tu} to the
3668 source file name. If the @samp{-@var{options}} form is used, @var{options}
3669 controls the details of the dump as described for the
3670 @option{-fdump-tree} options.
3672 @item -fdump-class-hierarchy @r{(C++ only)}
3673 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3674 @opindex fdump-class-hierarchy
3675 Dump a representation of each class's hierarchy and virtual function
3676 table layout to a file. The file name is made by appending @file{.class}
3677 to the source file name. If the @samp{-@var{options}} form is used,
3678 @var{options} controls the details of the dump as described for the
3679 @option{-fdump-tree} options.
3681 @item -fdump-ipa-@var{switch}
3683 Control the dumping at various stages of inter-procedural analysis
3684 language tree to a file. The file name is generated by appending a switch
3685 specific suffix to the source file name. The following dumps are possible:
3689 Enables all inter-procedural analysis dumps; currently the only produced
3690 dump is the @samp{cgraph} dump.
3693 Dumps information about call-graph optimization, unused function removal,
3694 and inlining decisions.
3697 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3698 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3700 Control the dumping at various stages of processing the intermediate
3701 language tree to a file. The file name is generated by appending a switch
3702 specific suffix to the source file name. If the @samp{-@var{options}}
3703 form is used, @var{options} is a list of @samp{-} separated options that
3704 control the details of the dump. Not all options are applicable to all
3705 dumps, those which are not meaningful will be ignored. The following
3706 options are available
3710 Print the address of each node. Usually this is not meaningful as it
3711 changes according to the environment and source file. Its primary use
3712 is for tying up a dump file with a debug environment.
3714 Inhibit dumping of members of a scope or body of a function merely
3715 because that scope has been reached. Only dump such items when they
3716 are directly reachable by some other path. When dumping pretty-printed
3717 trees, this option inhibits dumping the bodies of control structures.
3719 Print a raw representation of the tree. By default, trees are
3720 pretty-printed into a C-like representation.
3722 Enable more detailed dumps (not honored by every dump option).
3724 Enable dumping various statistics about the pass (not honored by every dump
3727 Enable showing basic block boundaries (disabled in raw dumps).
3729 Enable showing virtual operands for every statement.
3731 Enable showing line numbers for statements.
3733 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3735 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3738 The following tree dumps are possible:
3742 Dump before any tree based optimization, to @file{@var{file}.original}.
3745 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3748 Dump after function inlining, to @file{@var{file}.inlined}.
3751 @opindex fdump-tree-gimple
3752 Dump each function before and after the gimplification pass to a file. The
3753 file name is made by appending @file{.gimple} to the source file name.
3756 @opindex fdump-tree-cfg
3757 Dump the control flow graph of each function to a file. The file name is
3758 made by appending @file{.cfg} to the source file name.
3761 @opindex fdump-tree-vcg
3762 Dump the control flow graph of each function to a file in VCG format. The
3763 file name is made by appending @file{.vcg} to the source file name. Note
3764 that if the file contains more than one function, the generated file cannot
3765 be used directly by VCG@. You will need to cut and paste each function's
3766 graph into its own separate file first.
3769 @opindex fdump-tree-ch
3770 Dump each function after copying loop headers. The file name is made by
3771 appending @file{.ch} to the source file name.
3774 @opindex fdump-tree-ssa
3775 Dump SSA related information to a file. The file name is made by appending
3776 @file{.ssa} to the source file name.
3779 @opindex fdump-tree-alias
3780 Dump aliasing information for each function. The file name is made by
3781 appending @file{.alias} to the source file name.
3784 @opindex fdump-tree-ccp
3785 Dump each function after CCP@. The file name is made by appending
3786 @file{.ccp} to the source file name.
3789 @opindex fdump-tree-pre
3790 Dump trees after partial redundancy elimination. The file name is made
3791 by appending @file{.pre} to the source file name.
3794 @opindex fdump-tree-fre
3795 Dump trees after full redundancy elimination. The file name is made
3796 by appending @file{.fre} to the source file name.
3799 @opindex fdump-tree-dce
3800 Dump each function after dead code elimination. The file name is made by
3801 appending @file{.dce} to the source file name.
3804 @opindex fdump-tree-mudflap
3805 Dump each function after adding mudflap instrumentation. The file name is
3806 made by appending @file{.mudflap} to the source file name.
3809 @opindex fdump-tree-sra
3810 Dump each function after performing scalar replacement of aggregates. The
3811 file name is made by appending @file{.sra} to the source file name.
3814 @opindex fdump-tree-dom
3815 Dump each function after applying dominator tree optimizations. The file
3816 name is made by appending @file{.dom} to the source file name.
3819 @opindex fdump-tree-dse
3820 Dump each function after applying dead store elimination. The file
3821 name is made by appending @file{.dse} to the source file name.
3824 @opindex fdump-tree-phiopt
3825 Dump each function after optimizing PHI nodes into straightline code. The file
3826 name is made by appending @file{.phiopt} to the source file name.
3829 @opindex fdump-tree-forwprop
3830 Dump each function after forward propagating single use variables. The file
3831 name is made by appending @file{.forwprop} to the source file name.
3834 @opindex fdump-tree-copyrename
3835 Dump each function after applying the copy rename optimization. The file
3836 name is made by appending @file{.copyrename} to the source file name.
3839 @opindex fdump-tree-nrv
3840 Dump each function after applying the named return value optimization on
3841 generic trees. The file name is made by appending @file{.nrv} to the source
3845 @opindex fdump-tree-vect
3846 Dump each function after applying vectorization of loops. The file name is
3847 made by appending @file{.vect} to the source file name.
3850 @opindex fdump-tree-all
3851 Enable all the available tree dumps with the flags provided in this option.
3854 @item -frandom-seed=@var{string}
3855 @opindex frandom-string
3856 This option provides a seed that GCC uses when it would otherwise use
3857 random numbers. It is used to generate certain symbol names
3858 that have to be different in every compiled file. It is also used to
3859 place unique stamps in coverage data files and the object files that
3860 produce them. You can use the @option{-frandom-seed} option to produce
3861 reproducibly identical object files.
3863 The @var{string} should be different for every file you compile.
3865 @item -fsched-verbose=@var{n}
3866 @opindex fsched-verbose
3867 On targets that use instruction scheduling, this option controls the
3868 amount of debugging output the scheduler prints. This information is
3869 written to standard error, unless @option{-dS} or @option{-dR} is
3870 specified, in which case it is output to the usual dump
3871 listing file, @file{.sched} or @file{.sched2} respectively. However
3872 for @var{n} greater than nine, the output is always printed to standard
3875 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3876 same information as @option{-dRS}. For @var{n} greater than one, it
3877 also output basic block probabilities, detailed ready list information
3878 and unit/insn info. For @var{n} greater than two, it includes RTL
3879 at abort point, control-flow and regions info. And for @var{n} over
3880 four, @option{-fsched-verbose} also includes dependence info.
3884 Store the usual ``temporary'' intermediate files permanently; place them
3885 in the current directory and name them based on the source file. Thus,
3886 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3887 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3888 preprocessed @file{foo.i} output file even though the compiler now
3889 normally uses an integrated preprocessor.
3893 Report the CPU time taken by each subprocess in the compilation
3894 sequence. For C source files, this is the compiler proper and assembler
3895 (plus the linker if linking is done). The output looks like this:
3902 The first number on each line is the ``user time'', that is time spent
3903 executing the program itself. The second number is ``system time'',
3904 time spent executing operating system routines on behalf of the program.
3905 Both numbers are in seconds.
3907 @item -fvar-tracking
3908 @opindex fvar-tracking
3909 Run variable tracking pass. It computes where variables are stored at each
3910 position in code. Better debugging information is then generated
3911 (if the debugging information format supports this information).
3913 It is enabled by default when compiling with optimization (@option{-Os},
3914 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
3915 the debug info format supports it.
3917 @item -print-file-name=@var{library}
3918 @opindex print-file-name
3919 Print the full absolute name of the library file @var{library} that
3920 would be used when linking---and don't do anything else. With this
3921 option, GCC does not compile or link anything; it just prints the
3924 @item -print-multi-directory
3925 @opindex print-multi-directory
3926 Print the directory name corresponding to the multilib selected by any
3927 other switches present in the command line. This directory is supposed
3928 to exist in @env{GCC_EXEC_PREFIX}.
3930 @item -print-multi-lib
3931 @opindex print-multi-lib
3932 Print the mapping from multilib directory names to compiler switches
3933 that enable them. The directory name is separated from the switches by
3934 @samp{;}, and each switch starts with an @samp{@@} instead of the
3935 @samp{-}, without spaces between multiple switches. This is supposed to
3936 ease shell-processing.
3938 @item -print-prog-name=@var{program}
3939 @opindex print-prog-name
3940 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
3942 @item -print-libgcc-file-name
3943 @opindex print-libgcc-file-name
3944 Same as @option{-print-file-name=libgcc.a}.
3946 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
3947 but you do want to link with @file{libgcc.a}. You can do
3950 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
3953 @item -print-search-dirs
3954 @opindex print-search-dirs
3955 Print the name of the configured installation directory and a list of
3956 program and library directories @command{gcc} will search---and don't do anything else.
3958 This is useful when @command{gcc} prints the error message
3959 @samp{installation problem, cannot exec cpp0: No such file or directory}.
3960 To resolve this you either need to put @file{cpp0} and the other compiler
3961 components where @command{gcc} expects to find them, or you can set the environment
3962 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
3963 Don't forget the trailing @samp{/}.
3964 @xref{Environment Variables}.
3967 @opindex dumpmachine
3968 Print the compiler's target machine (for example,
3969 @samp{i686-pc-linux-gnu})---and don't do anything else.
3972 @opindex dumpversion
3973 Print the compiler version (for example, @samp{3.0})---and don't do
3978 Print the compiler's built-in specs---and don't do anything else. (This
3979 is used when GCC itself is being built.) @xref{Spec Files}.
3981 @item -feliminate-unused-debug-types
3982 @opindex feliminate-unused-debug-types
3983 Normally, when producing DWARF2 output, GCC will emit debugging
3984 information for all types declared in a compilation
3985 unit, regardless of whether or not they are actually used
3986 in that compilation unit. Sometimes this is useful, such as
3987 if, in the debugger, you want to cast a value to a type that is
3988 not actually used in your program (but is declared). More often,
3989 however, this results in a significant amount of wasted space.
3990 With this option, GCC will avoid producing debug symbol output
3991 for types that are nowhere used in the source file being compiled.
3994 @node Optimize Options
3995 @section Options That Control Optimization
3996 @cindex optimize options
3997 @cindex options, optimization
3999 These options control various sorts of optimizations.
4001 Without any optimization option, the compiler's goal is to reduce the
4002 cost of compilation and to make debugging produce the expected
4003 results. Statements are independent: if you stop the program with a
4004 breakpoint between statements, you can then assign a new value to any
4005 variable or change the program counter to any other statement in the
4006 function and get exactly the results you would expect from the source
4009 Turning on optimization flags makes the compiler attempt to improve
4010 the performance and/or code size at the expense of compilation time
4011 and possibly the ability to debug the program.
4013 The compiler performs optimization based on the knowledge it has of
4014 the program. Optimization levels @option{-O2} and above, in
4015 particular, enable @emph{unit-at-a-time} mode, which allows the
4016 compiler to consider information gained from later functions in
4017 the file when compiling a function. Compiling multiple files at
4018 once to a single output file in @emph{unit-at-a-time} mode allows
4019 the compiler to use information gained from all of the files when
4020 compiling each of them.
4022 Not all optimizations are controlled directly by a flag. Only
4023 optimizations that have a flag are listed.
4030 Optimize. Optimizing compilation takes somewhat more time, and a lot
4031 more memory for a large function.
4033 With @option{-O}, the compiler tries to reduce code size and execution
4034 time, without performing any optimizations that take a great deal of
4037 @option{-O} turns on the following optimization flags:
4038 @gccoptlist{-fdefer-pop @gol
4039 -fmerge-constants @gol
4041 -floop-optimize @gol
4042 -fif-conversion @gol
4043 -fif-conversion2 @gol
4044 -fdelayed-branch @gol
4045 -fguess-branch-probability @gol
4048 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4049 where doing so does not interfere with debugging.
4053 Optimize even more. GCC performs nearly all supported optimizations
4054 that do not involve a space-speed tradeoff. The compiler does not
4055 perform loop unrolling or function inlining when you specify @option{-O2}.
4056 As compared to @option{-O}, this option increases both compilation time
4057 and the performance of the generated code.
4059 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4060 also turns on the following optimization flags:
4061 @gccoptlist{-fforce-mem @gol
4062 -foptimize-sibling-calls @gol
4063 -fstrength-reduce @gol
4064 -fcse-follow-jumps -fcse-skip-blocks @gol
4065 -frerun-cse-after-loop -frerun-loop-opt @gol
4066 -fgcse -fgcse-lm -fgcse-sm -fgcse-las @gol
4067 -fdelete-null-pointer-checks @gol
4068 -fexpensive-optimizations @gol
4070 -fschedule-insns -fschedule-insns2 @gol
4071 -fsched-interblock -fsched-spec @gol
4074 -freorder-blocks -freorder-functions @gol
4075 -fstrict-aliasing @gol
4076 -funit-at-a-time @gol
4077 -falign-functions -falign-jumps @gol
4078 -falign-loops -falign-labels @gol
4081 Please note the warning under @option{-fgcse} about
4082 invoking @option{-O2} on programs that use computed gotos.
4086 Optimize yet more. @option{-O3} turns on all optimizations specified by
4087 @option{-O2} and also turns on the @option{-finline-functions},
4088 @option{-fweb} and @option{-fgcse-after-reload} options.
4092 Do not optimize. This is the default.
4096 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4097 do not typically increase code size. It also performs further
4098 optimizations designed to reduce code size.
4100 @option{-Os} disables the following optimization flags:
4101 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4102 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4104 If you use multiple @option{-O} options, with or without level numbers,
4105 the last such option is the one that is effective.
4108 Options of the form @option{-f@var{flag}} specify machine-independent
4109 flags. Most flags have both positive and negative forms; the negative
4110 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4111 below, only one of the forms is listed---the one you typically will
4112 use. You can figure out the other form by either removing @samp{no-}
4115 The following options control specific optimizations. They are either
4116 activated by @option{-O} options or are related to ones that are. You
4117 can use the following flags in the rare cases when ``fine-tuning'' of
4118 optimizations to be performed is desired.
4121 @item -fno-default-inline
4122 @opindex fno-default-inline
4123 Do not make member functions inline by default merely because they are
4124 defined inside the class scope (C++ only). Otherwise, when you specify
4125 @w{@option{-O}}, member functions defined inside class scope are compiled
4126 inline by default; i.e., you don't need to add @samp{inline} in front of
4127 the member function name.
4129 @item -fno-defer-pop
4130 @opindex fno-defer-pop
4131 Always pop the arguments to each function call as soon as that function
4132 returns. For machines which must pop arguments after a function call,
4133 the compiler normally lets arguments accumulate on the stack for several
4134 function calls and pops them all at once.
4136 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4140 Force memory operands to be copied into registers before doing
4141 arithmetic on them. This produces better code by making all memory
4142 references potential common subexpressions. When they are not common
4143 subexpressions, instruction combination should eliminate the separate
4146 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4149 @opindex fforce-addr
4150 Force memory address constants to be copied into registers before
4151 doing arithmetic on them. This may produce better code just as
4152 @option{-fforce-mem} may.
4154 @item -fomit-frame-pointer
4155 @opindex fomit-frame-pointer
4156 Don't keep the frame pointer in a register for functions that
4157 don't need one. This avoids the instructions to save, set up and
4158 restore frame pointers; it also makes an extra register available
4159 in many functions. @strong{It also makes debugging impossible on
4162 On some machines, such as the VAX, this flag has no effect, because
4163 the standard calling sequence automatically handles the frame pointer
4164 and nothing is saved by pretending it doesn't exist. The
4165 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4166 whether a target machine supports this flag. @xref{Registers,,Register
4167 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4169 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4171 @item -foptimize-sibling-calls
4172 @opindex foptimize-sibling-calls
4173 Optimize sibling and tail recursive calls.
4175 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4179 Don't pay attention to the @code{inline} keyword. Normally this option
4180 is used to keep the compiler from expanding any functions inline.
4181 Note that if you are not optimizing, no functions can be expanded inline.
4183 @item -finline-functions
4184 @opindex finline-functions
4185 Integrate all simple functions into their callers. The compiler
4186 heuristically decides which functions are simple enough to be worth
4187 integrating in this way.
4189 If all calls to a given function are integrated, and the function is
4190 declared @code{static}, then the function is normally not output as
4191 assembler code in its own right.
4193 Enabled at level @option{-O3}.
4195 @item -finline-limit=@var{n}
4196 @opindex finline-limit
4197 By default, GCC limits the size of functions that can be inlined. This flag
4198 allows the control of this limit for functions that are explicitly marked as
4199 inline (i.e., marked with the inline keyword or defined within the class
4200 definition in c++). @var{n} is the size of functions that can be inlined in
4201 number of pseudo instructions (not counting parameter handling). The default
4202 value of @var{n} is 600.
4203 Increasing this value can result in more inlined code at
4204 the cost of compilation time and memory consumption. Decreasing usually makes
4205 the compilation faster and less code will be inlined (which presumably
4206 means slower programs). This option is particularly useful for programs that
4207 use inlining heavily such as those based on recursive templates with C++.
4209 Inlining is actually controlled by a number of parameters, which may be
4210 specified individually by using @option{--param @var{name}=@var{value}}.
4211 The @option{-finline-limit=@var{n}} option sets some of these parameters
4215 @item max-inline-insns-single
4216 is set to @var{n}/2.
4217 @item max-inline-insns-auto
4218 is set to @var{n}/2.
4219 @item min-inline-insns
4220 is set to 130 or @var{n}/4, whichever is smaller.
4221 @item max-inline-insns-rtl
4225 See below for a documentation of the individual
4226 parameters controlling inlining.
4228 @emph{Note:} pseudo instruction represents, in this particular context, an
4229 abstract measurement of function's size. In no way, it represents a count
4230 of assembly instructions and as such its exact meaning might change from one
4231 release to an another.
4233 @item -fkeep-inline-functions
4234 @opindex fkeep-inline-functions
4235 In C, emit @code{static} functions that are declared @code{inline}
4236 into the object file, even if the function has been inlined into all
4237 of its callers. This switch does not affect functions using the
4238 @code{extern inline} extension in GNU C@. In C++, emit any and all
4239 inline functions into the object file.
4241 @item -fkeep-static-consts
4242 @opindex fkeep-static-consts
4243 Emit variables declared @code{static const} when optimization isn't turned
4244 on, even if the variables aren't referenced.
4246 GCC enables this option by default. If you want to force the compiler to
4247 check if the variable was referenced, regardless of whether or not
4248 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4250 @item -fmerge-constants
4251 Attempt to merge identical constants (string constants and floating point
4252 constants) across compilation units.
4254 This option is the default for optimized compilation if the assembler and
4255 linker support it. Use @option{-fno-merge-constants} to inhibit this
4258 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4260 @item -fmerge-all-constants
4261 Attempt to merge identical constants and identical variables.
4263 This option implies @option{-fmerge-constants}. In addition to
4264 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4265 arrays or initialized constant variables with integral or floating point
4266 types. Languages like C or C++ require each non-automatic variable to
4267 have distinct location, so using this option will result in non-conforming
4270 @item -fmodulo-sched
4271 @opindex fmodulo-sched
4272 Perform swing modulo scheduling immediately before the first scheduling
4273 pass. This pass looks at innermost loops and reorders their
4274 instructions by overlapping different iterations.
4278 Use a graph coloring register allocator. Currently this option is meant
4279 only for testing. Users should not specify this option, since it is not
4280 yet ready for production use.
4282 @item -fno-branch-count-reg
4283 @opindex fno-branch-count-reg
4284 Do not use ``decrement and branch'' instructions on a count register,
4285 but instead generate a sequence of instructions that decrement a
4286 register, compare it against zero, then branch based upon the result.
4287 This option is only meaningful on architectures that support such
4288 instructions, which include x86, PowerPC, IA-64 and S/390.
4290 The default is @option{-fbranch-count-reg}, enabled when
4291 @option{-fstrength-reduce} is enabled.
4293 @item -fno-function-cse
4294 @opindex fno-function-cse
4295 Do not put function addresses in registers; make each instruction that
4296 calls a constant function contain the function's address explicitly.
4298 This option results in less efficient code, but some strange hacks
4299 that alter the assembler output may be confused by the optimizations
4300 performed when this option is not used.
4302 The default is @option{-ffunction-cse}
4304 @item -fno-zero-initialized-in-bss
4305 @opindex fno-zero-initialized-in-bss
4306 If the target supports a BSS section, GCC by default puts variables that
4307 are initialized to zero into BSS@. This can save space in the resulting
4310 This option turns off this behavior because some programs explicitly
4311 rely on variables going to the data section. E.g., so that the
4312 resulting executable can find the beginning of that section and/or make
4313 assumptions based on that.
4315 The default is @option{-fzero-initialized-in-bss}.
4317 @item -fbounds-check
4318 @opindex fbounds-check
4319 For front-ends that support it, generate additional code to check that
4320 indices used to access arrays are within the declared range. This is
4321 currently only supported by the Java and Fortran front-ends, where
4322 this option defaults to true and false respectively.
4324 @item -fmudflap -fmudflapth -fmudflapir
4328 @cindex bounds checking
4330 For front-ends that support it (C and C++), instrument all risky
4331 pointer/array dereferencing operations, some standard library
4332 string/heap functions, and some other associated constructs with
4333 range/validity tests. Modules so instrumented should be immune to
4334 buffer overflows, invalid heap use, and some other classes of C/C++
4335 programming errors. The instrumentation relies on a separate runtime
4336 library (@file{libmudflap}), which will be linked into a program if
4337 @option{-fmudflap} is given at link time. Run-time behavior of the
4338 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4339 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4342 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4343 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4344 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4345 instrumentation should ignore pointer reads. This produces less
4346 instrumentation (and therefore faster execution) and still provides
4347 some protection against outright memory corrupting writes, but allows
4348 erroneously read data to propagate within a program.
4350 @item -fstrength-reduce
4351 @opindex fstrength-reduce
4352 Perform the optimizations of loop strength reduction and
4353 elimination of iteration variables.
4355 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4357 @item -fthread-jumps
4358 @opindex fthread-jumps
4359 Perform optimizations where we check to see if a jump branches to a
4360 location where another comparison subsumed by the first is found. If
4361 so, the first branch is redirected to either the destination of the
4362 second branch or a point immediately following it, depending on whether
4363 the condition is known to be true or false.
4365 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4367 @item -fcse-follow-jumps
4368 @opindex fcse-follow-jumps
4369 In common subexpression elimination, scan through jump instructions
4370 when the target of the jump is not reached by any other path. For
4371 example, when CSE encounters an @code{if} statement with an
4372 @code{else} clause, CSE will follow the jump when the condition
4375 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4377 @item -fcse-skip-blocks
4378 @opindex fcse-skip-blocks
4379 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4380 follow jumps which conditionally skip over blocks. When CSE
4381 encounters a simple @code{if} statement with no else clause,
4382 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4383 body of the @code{if}.
4385 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4387 @item -frerun-cse-after-loop
4388 @opindex frerun-cse-after-loop
4389 Re-run common subexpression elimination after loop optimizations has been
4392 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4394 @item -frerun-loop-opt
4395 @opindex frerun-loop-opt
4396 Run the loop optimizer twice.
4398 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4402 Perform a global common subexpression elimination pass.
4403 This pass also performs global constant and copy propagation.
4405 @emph{Note:} When compiling a program using computed gotos, a GCC
4406 extension, you may get better runtime performance if you disable
4407 the global common subexpression elimination pass by adding
4408 @option{-fno-gcse} to the command line.
4410 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4414 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4415 attempt to move loads which are only killed by stores into themselves. This
4416 allows a loop containing a load/store sequence to be changed to a load outside
4417 the loop, and a copy/store within the loop.
4419 Enabled by default when gcse is enabled.
4423 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4424 global common subexpression elimination. This pass will attempt to move
4425 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4426 loops containing a load/store sequence can be changed to a load before
4427 the loop and a store after the loop.
4429 Enabled by default when gcse is enabled.
4433 When @option{-fgcse-las} is enabled, the global common subexpression
4434 elimination pass eliminates redundant loads that come after stores to the
4435 same memory location (both partial and full redundancies).
4437 Enabled by default when gcse is enabled.
4439 @item -fgcse-after-reload
4440 @opindex fgcse-after-reload
4441 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4442 pass is performed after reload. The purpose of this pass is to cleanup
4445 @item -floop-optimize
4446 @opindex floop-optimize
4447 Perform loop optimizations: move constant expressions out of loops, simplify
4448 exit test conditions and optionally do strength-reduction as well.
4450 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4452 @item -floop-optimize2
4453 @opindex floop-optimize2
4454 Perform loop optimizations using the new loop optimizer. The optimizations
4455 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4458 @item -fcrossjumping
4459 @opindex crossjumping
4460 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4461 resulting code may or may not perform better than without cross-jumping.
4463 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4465 @item -fif-conversion
4466 @opindex if-conversion
4467 Attempt to transform conditional jumps into branch-less equivalents. This
4468 include use of conditional moves, min, max, set flags and abs instructions, and
4469 some tricks doable by standard arithmetics. The use of conditional execution
4470 on chips where it is available is controlled by @code{if-conversion2}.
4472 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4474 @item -fif-conversion2
4475 @opindex if-conversion2
4476 Use conditional execution (where available) to transform conditional jumps into
4477 branch-less equivalents.
4479 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4481 @item -fdelete-null-pointer-checks
4482 @opindex fdelete-null-pointer-checks
4483 Use global dataflow analysis to identify and eliminate useless checks
4484 for null pointers. The compiler assumes that dereferencing a null
4485 pointer would have halted the program. If a pointer is checked after
4486 it has already been dereferenced, it cannot be null.
4488 In some environments, this assumption is not true, and programs can
4489 safely dereference null pointers. Use
4490 @option{-fno-delete-null-pointer-checks} to disable this optimization
4491 for programs which depend on that behavior.
4493 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4495 @item -fexpensive-optimizations
4496 @opindex fexpensive-optimizations
4497 Perform a number of minor optimizations that are relatively expensive.
4499 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4501 @item -foptimize-register-move
4503 @opindex foptimize-register-move
4505 Attempt to reassign register numbers in move instructions and as
4506 operands of other simple instructions in order to maximize the amount of
4507 register tying. This is especially helpful on machines with two-operand
4510 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4513 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4515 @item -fdelayed-branch
4516 @opindex fdelayed-branch
4517 If supported for the target machine, attempt to reorder instructions
4518 to exploit instruction slots available after delayed branch
4521 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4523 @item -fschedule-insns
4524 @opindex fschedule-insns
4525 If supported for the target machine, attempt to reorder instructions to
4526 eliminate execution stalls due to required data being unavailable. This
4527 helps machines that have slow floating point or memory load instructions
4528 by allowing other instructions to be issued until the result of the load
4529 or floating point instruction is required.
4531 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4533 @item -fschedule-insns2
4534 @opindex fschedule-insns2
4535 Similar to @option{-fschedule-insns}, but requests an additional pass of
4536 instruction scheduling after register allocation has been done. This is
4537 especially useful on machines with a relatively small number of
4538 registers and where memory load instructions take more than one cycle.
4540 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4542 @item -fno-sched-interblock
4543 @opindex fno-sched-interblock
4544 Don't schedule instructions across basic blocks. This is normally
4545 enabled by default when scheduling before register allocation, i.e.@:
4546 with @option{-fschedule-insns} or at @option{-O2} or higher.
4548 @item -fno-sched-spec
4549 @opindex fno-sched-spec
4550 Don't allow speculative motion of non-load instructions. This is normally
4551 enabled by default when scheduling before register allocation, i.e.@:
4552 with @option{-fschedule-insns} or at @option{-O2} or higher.
4554 @item -fsched-spec-load
4555 @opindex fsched-spec-load
4556 Allow speculative motion of some load instructions. This only makes
4557 sense when scheduling before register allocation, i.e.@: with
4558 @option{-fschedule-insns} or at @option{-O2} or higher.
4560 @item -fsched-spec-load-dangerous
4561 @opindex fsched-spec-load-dangerous
4562 Allow speculative motion of more load instructions. This only makes
4563 sense when scheduling before register allocation, i.e.@: with
4564 @option{-fschedule-insns} or at @option{-O2} or higher.
4566 @item -fsched-stalled-insns=@var{n}
4567 @opindex fsched-stalled-insns
4568 Define how many insns (if any) can be moved prematurely from the queue
4569 of stalled insns into the ready list, during the second scheduling pass.
4571 @item -fsched-stalled-insns-dep=@var{n}
4572 @opindex fsched-stalled-insns-dep
4573 Define how many insn groups (cycles) will be examined for a dependency
4574 on a stalled insn that is candidate for premature removal from the queue
4575 of stalled insns. Has an effect only during the second scheduling pass,
4576 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4578 @item -fsched2-use-superblocks
4579 @opindex fsched2-use-superblocks
4580 When scheduling after register allocation, do use superblock scheduling
4581 algorithm. Superblock scheduling allows motion across basic block boundaries
4582 resulting on faster schedules. This option is experimental, as not all machine
4583 descriptions used by GCC model the CPU closely enough to avoid unreliable
4584 results from the algorithm.
4586 This only makes sense when scheduling after register allocation, i.e.@: with
4587 @option{-fschedule-insns2} or at @option{-O2} or higher.
4589 @item -fsched2-use-traces
4590 @opindex fsched2-use-traces
4591 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4592 allocation and additionally perform code duplication in order to increase the
4593 size of superblocks using tracer pass. See @option{-ftracer} for details on
4596 This mode should produce faster but significantly longer programs. Also
4597 without @option{-fbranch-probabilities} the traces constructed may not
4598 match the reality and hurt the performance. This only makes
4599 sense when scheduling after register allocation, i.e.@: with
4600 @option{-fschedule-insns2} or at @option{-O2} or higher.
4602 @item -freschedule-modulo-scheduled-loops
4603 @opindex fscheduling-in-modulo-scheduled-loops
4604 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4605 we may want to prevent the later scheduling passes from changing its schedule, we use this
4606 option to control that.
4608 @item -fcaller-saves
4609 @opindex fcaller-saves
4610 Enable values to be allocated in registers that will be clobbered by
4611 function calls, by emitting extra instructions to save and restore the
4612 registers around such calls. Such allocation is done only when it
4613 seems to result in better code than would otherwise be produced.
4615 This option is always enabled by default on certain machines, usually
4616 those which have no call-preserved registers to use instead.
4618 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4621 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4622 enabled by default at @option{-O} and higher.
4625 Perform Full Redundancy Elimination (FRE) on trees. The difference
4626 between FRE and PRE is that FRE only considers expressions
4627 that are computed on all paths leading to the redundant computation.
4628 This analysis faster than PRE, though it exposes fewer redundancies.
4629 This flag is enabled by default at @option{-O} and higher.
4632 Perform sparse conditional constant propagation (CCP) on trees. This flag
4633 is enabled by default at @option{-O} and higher.
4636 Perform dead code elimination (DCE) on trees. This flag is enabled by
4637 default at @option{-O} and higher.
4639 @item -ftree-dominator-opts
4640 Perform dead code elimination (DCE) on trees. This flag is enabled by
4641 default at @option{-O} and higher.
4644 Perform loop header copying on trees. This is beneficial since it increases
4645 effectiveness of code motion optimizations. It also saves one jump. This flag
4646 is enabled by default at @option{-O} and higher. It is not enabled
4647 for @option{-Os}, since it usually increases code size.
4649 @item -ftree-loop-optimize
4650 Perform loop optimizations on trees. This flag is enabled by default
4651 at @option{-O} and higher.
4653 @item -ftree-loop-linear
4654 Perform linear loop transformations on tree. This flag can improve cache
4655 performance and allow further loop optimizations to take place.
4658 Perform loop invariant motion on trees. This pass moves only invartiants that
4659 would be hard to handle on rtl level (function calls, operations that expand to
4660 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4661 operands of conditions that are invariant out of the loop, so that we can use
4662 just trivial invariantness analysis in loop unswitching. The pass also includes
4666 Create a canonical counter for number of iterations in the loop for that
4667 determining number of iterations requires complicated analysis. Later
4668 optimizations then may determine the number easily. Useful especially
4669 in connection with unrolling.
4672 Perform induction variable optimizations (strength reduction, induction
4673 variable merging and induction variable elimination) on trees.
4676 Perform scalar replacement of aggregates. This pass replaces structure
4677 references with scalars to prevent committing structures to memory too
4678 early. This flag is enabled by default at @option{-O} and higher.
4680 @item -ftree-copyrename
4681 Perform copy renaming on trees. This pass attempts to rename compiler
4682 temporaries to other variables at copy locations, usually resulting in
4683 variable names which more closely resemble the original variables. This flag
4684 is enabled by default at @option{-O} and higher.
4687 Perform temporary expression replacement during the SSA->normal phase. Single
4688 use/single def temporaries are replaced at their use location with their
4689 defining expression. This results in non-GIMPLE code, but gives the expanders
4690 much more complex trees to work on resulting in better RTL generation. This is
4691 enabled by default at @option{-O} and higher.
4694 Perform live range splitting during the SSA->normal phase. Distinct live
4695 ranges of a variable are split into unique variables, allowing for better
4696 optimization later. This is enabled by default at @option{-O} and higher.
4698 @item -ftree-vectorize
4699 Perform loop vectorization on trees.
4703 Perform tail duplication to enlarge superblock size. This transformation
4704 simplifies the control flow of the function allowing other optimizations to do
4707 @item -funroll-loops
4708 @opindex funroll-loops
4709 Unroll loops whose number of iterations can be determined at compile
4710 time or upon entry to the loop. @option{-funroll-loops} implies both
4711 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4712 option makes code larger, and may or may not make it run faster.
4714 @item -funroll-all-loops
4715 @opindex funroll-all-loops
4716 Unroll all loops, even if their number of iterations is uncertain when
4717 the loop is entered. This usually makes programs run more slowly.
4718 @option{-funroll-all-loops} implies the same options as
4719 @option{-funroll-loops},
4721 @item -fsplit-ivs-in-unroller
4722 @opindex -fsplit-ivs-in-unroller
4723 Enables expressing of values of induction variables in later iterations
4724 of the unrolled loop using the value in the first iteration. This breaks
4725 long dependency chains, thus improving efficiency of the scheduling passes
4726 (for best results, @option{-fweb} should be used as well).
4728 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4729 same effect. However in cases the loop body is more complicated than
4730 a single basic block, this is not reliable. It also does not work at all
4731 on some of the architectures due to restrictions in the CSE pass.
4733 This optimization is enabled by default.
4735 @item -fvariable-expansion-in-unroller
4736 @opindex -fvariable-expansion-in-unroller
4737 With this option, the compiler will create multiple copies of some
4738 local variables when unrolling a loop which can result in superior code.
4740 @item -fprefetch-loop-arrays
4741 @opindex fprefetch-loop-arrays
4742 If supported by the target machine, generate instructions to prefetch
4743 memory to improve the performance of loops that access large arrays.
4745 These options may generate better or worse code; results are highly
4746 dependent on the structure of loops within the source code.
4749 @itemx -fno-peephole2
4750 @opindex fno-peephole
4751 @opindex fno-peephole2
4752 Disable any machine-specific peephole optimizations. The difference
4753 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4754 are implemented in the compiler; some targets use one, some use the
4755 other, a few use both.
4757 @option{-fpeephole} is enabled by default.
4758 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4760 @item -fno-guess-branch-probability
4761 @opindex fno-guess-branch-probability
4762 Do not guess branch probabilities using heuristics.
4764 GCC will use heuristics to guess branch probabilities if they are
4765 not provided by profiling feedback (@option{-fprofile-arcs}). These
4766 heuristics are based on the control flow graph. If some branch probabilities
4767 are specified by @samp{__builtin_expect}, then the heuristics will be
4768 used to guess branch probabilities for the rest of the control flow graph,
4769 taking the @samp{__builtin_expect} info into account. The interactions
4770 between the heuristics and @samp{__builtin_expect} can be complex, and in
4771 some cases, it may be useful to disable the heuristics so that the effects
4772 of @samp{__builtin_expect} are easier to understand.
4774 The default is @option{-fguess-branch-probability} at levels
4775 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4777 @item -freorder-blocks
4778 @opindex freorder-blocks
4779 Reorder basic blocks in the compiled function in order to reduce number of
4780 taken branches and improve code locality.
4782 Enabled at levels @option{-O2}, @option{-O3}.
4784 @item -freorder-blocks-and-partition
4785 @opindex freorder-blocks-and-partition
4786 In addition to reordering basic blocks in the compiled function, in order
4787 to reduce number of taken branches, partitions hot and cold basic blocks
4788 into separate sections of the assembly and .o files, to improve
4789 paging and cache locality performance.
4791 This optimization is automatically turned off in the presence of
4792 exception handling, for linkonce sections, for functions with a user-defined
4793 section attribute and on any architecture that does not support named
4796 @item -freorder-functions
4797 @opindex freorder-functions
4798 Reorder basic blocks in the compiled function in order to reduce number of
4799 taken branches and improve code locality. This is implemented by using special
4800 subsections @code{.text.hot} for most frequently executed functions and
4801 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4802 the linker so object file format must support named sections and linker must
4803 place them in a reasonable way.
4805 Also profile feedback must be available in to make this option effective. See
4806 @option{-fprofile-arcs} for details.
4808 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4810 @item -fstrict-aliasing
4811 @opindex fstrict-aliasing
4812 Allows the compiler to assume the strictest aliasing rules applicable to
4813 the language being compiled. For C (and C++), this activates
4814 optimizations based on the type of expressions. In particular, an
4815 object of one type is assumed never to reside at the same address as an
4816 object of a different type, unless the types are almost the same. For
4817 example, an @code{unsigned int} can alias an @code{int}, but not a
4818 @code{void*} or a @code{double}. A character type may alias any other
4821 Pay special attention to code like this:
4834 The practice of reading from a different union member than the one most
4835 recently written to (called ``type-punning'') is common. Even with
4836 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4837 is accessed through the union type. So, the code above will work as
4838 expected. However, this code might not:
4849 Every language that wishes to perform language-specific alias analysis
4850 should define a function that computes, given an @code{tree}
4851 node, an alias set for the node. Nodes in different alias sets are not
4852 allowed to alias. For an example, see the C front-end function
4853 @code{c_get_alias_set}.
4855 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4857 @item -falign-functions
4858 @itemx -falign-functions=@var{n}
4859 @opindex falign-functions
4860 Align the start of functions to the next power-of-two greater than
4861 @var{n}, skipping up to @var{n} bytes. For instance,
4862 @option{-falign-functions=32} aligns functions to the next 32-byte
4863 boundary, but @option{-falign-functions=24} would align to the next
4864 32-byte boundary only if this can be done by skipping 23 bytes or less.
4866 @option{-fno-align-functions} and @option{-falign-functions=1} are
4867 equivalent and mean that functions will not be aligned.
4869 Some assemblers only support this flag when @var{n} is a power of two;
4870 in that case, it is rounded up.
4872 If @var{n} is not specified or is zero, use a machine-dependent default.
4874 Enabled at levels @option{-O2}, @option{-O3}.
4876 @item -falign-labels
4877 @itemx -falign-labels=@var{n}
4878 @opindex falign-labels
4879 Align all branch targets to a power-of-two boundary, skipping up to
4880 @var{n} bytes like @option{-falign-functions}. This option can easily
4881 make code slower, because it must insert dummy operations for when the
4882 branch target is reached in the usual flow of the code.
4884 @option{-fno-align-labels} and @option{-falign-labels=1} are
4885 equivalent and mean that labels will not be aligned.
4887 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
4888 are greater than this value, then their values are used instead.
4890 If @var{n} is not specified or is zero, use a machine-dependent default
4891 which is very likely to be @samp{1}, meaning no alignment.
4893 Enabled at levels @option{-O2}, @option{-O3}.
4896 @itemx -falign-loops=@var{n}
4897 @opindex falign-loops
4898 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
4899 like @option{-falign-functions}. The hope is that the loop will be
4900 executed many times, which will make up for any execution of the dummy
4903 @option{-fno-align-loops} and @option{-falign-loops=1} are
4904 equivalent and mean that loops will not be aligned.
4906 If @var{n} is not specified or is zero, use a machine-dependent default.
4908 Enabled at levels @option{-O2}, @option{-O3}.
4911 @itemx -falign-jumps=@var{n}
4912 @opindex falign-jumps
4913 Align branch targets to a power-of-two boundary, for branch targets
4914 where the targets can only be reached by jumping, skipping up to @var{n}
4915 bytes like @option{-falign-functions}. In this case, no dummy operations
4918 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
4919 equivalent and mean that loops will not be aligned.
4921 If @var{n} is not specified or is zero, use a machine-dependent default.
4923 Enabled at levels @option{-O2}, @option{-O3}.
4925 @item -funit-at-a-time
4926 @opindex funit-at-a-time
4927 Parse the whole compilation unit before starting to produce code.
4928 This allows some extra optimizations to take place but consumes
4929 more memory (in general). There are some compatibility issues
4930 with @emph{unit-at-at-time} mode:
4933 enabling @emph{unit-at-a-time} mode may change the order
4934 in which functions, variables, and top-level @code{asm} statements
4935 are emitted, and will likely break code relying on some particular
4936 ordering. The majority of such top-level @code{asm} statements,
4937 though, can be replaced by @code{section} attributes.
4940 @emph{unit-at-a-time} mode removes unreferenced static variables
4941 and functions are removed. This may result in undefined references
4942 when an @code{asm} statement refers directly to variables or functions
4943 that are otherwise unused. In that case either the variable/function
4944 shall be listed as an operand of the @code{asm} statement operand or,
4945 in the case of top-level @code{asm} statements the attribute @code{used}
4946 shall be used on the declaration.
4949 Static functions now can use non-standard passing conventions that
4950 may break @code{asm} statements calling functions directly. Again,
4951 attribute @code{used} will prevent this behavior.
4954 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
4955 but this scheme may not be supported by future releases of GCC@.
4957 Enabled at levels @option{-O2}, @option{-O3}.
4961 Constructs webs as commonly used for register allocation purposes and assign
4962 each web individual pseudo register. This allows the register allocation pass
4963 to operate on pseudos directly, but also strengthens several other optimization
4964 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
4965 however, make debugging impossible, since variables will no longer stay in a
4968 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
4969 on targets where the default format for debugging information supports
4972 @item -fno-cprop-registers
4973 @opindex fno-cprop-registers
4974 After register allocation and post-register allocation instruction splitting,
4975 we perform a copy-propagation pass to try to reduce scheduling dependencies
4976 and occasionally eliminate the copy.
4978 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4980 @item -fprofile-generate
4981 @opindex fprofile-generate
4983 Enable options usually used for instrumenting application to produce
4984 profile useful for later recompilation with profile feedback based
4985 optimization. You must use @option{-fprofile-generate} both when
4986 compiling and when linking your program.
4988 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
4991 @opindex fprofile-use
4992 Enable profile feedback directed optimizations, and optimizations
4993 generally profitable only with profile feedback available.
4995 The following options are enabled: @code{-fbranch-probabilities},
4996 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5000 The following options control compiler behavior regarding floating
5001 point arithmetic. These options trade off between speed and
5002 correctness. All must be specifically enabled.
5006 @opindex ffloat-store
5007 Do not store floating point variables in registers, and inhibit other
5008 options that might change whether a floating point value is taken from a
5011 @cindex floating point precision
5012 This option prevents undesirable excess precision on machines such as
5013 the 68000 where the floating registers (of the 68881) keep more
5014 precision than a @code{double} is supposed to have. Similarly for the
5015 x86 architecture. For most programs, the excess precision does only
5016 good, but a few programs rely on the precise definition of IEEE floating
5017 point. Use @option{-ffloat-store} for such programs, after modifying
5018 them to store all pertinent intermediate computations into variables.
5022 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5023 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5024 @option{-fno-rounding-math} and @option{-fno-signaling-nans}.
5026 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5028 This option should never be turned on by any @option{-O} option since
5029 it can result in incorrect output for programs which depend on
5030 an exact implementation of IEEE or ISO rules/specifications for
5033 @item -fno-math-errno
5034 @opindex fno-math-errno
5035 Do not set ERRNO after calling math functions that are executed
5036 with a single instruction, e.g., sqrt. A program that relies on
5037 IEEE exceptions for math error handling may want to use this flag
5038 for speed while maintaining IEEE arithmetic compatibility.
5040 This option should never be turned on by any @option{-O} option since
5041 it can result in incorrect output for programs which depend on
5042 an exact implementation of IEEE or ISO rules/specifications for
5045 The default is @option{-fmath-errno}.
5047 @item -funsafe-math-optimizations
5048 @opindex funsafe-math-optimizations
5049 Allow optimizations for floating-point arithmetic that (a) assume
5050 that arguments and results are valid and (b) may violate IEEE or
5051 ANSI standards. When used at link-time, it may include libraries
5052 or startup files that change the default FPU control word or other
5053 similar optimizations.
5055 This option should never be turned on by any @option{-O} option since
5056 it can result in incorrect output for programs which depend on
5057 an exact implementation of IEEE or ISO rules/specifications for
5060 The default is @option{-fno-unsafe-math-optimizations}.
5062 @item -ffinite-math-only
5063 @opindex ffinite-math-only
5064 Allow optimizations for floating-point arithmetic that assume
5065 that arguments and results are not NaNs or +-Infs.
5067 This option should never be turned on by any @option{-O} option since
5068 it can result in incorrect output for programs which depend on
5069 an exact implementation of IEEE or ISO rules/specifications.
5071 The default is @option{-fno-finite-math-only}.
5073 @item -fno-trapping-math
5074 @opindex fno-trapping-math
5075 Compile code assuming that floating-point operations cannot generate
5076 user-visible traps. These traps include division by zero, overflow,
5077 underflow, inexact result and invalid operation. This option implies
5078 @option{-fno-signaling-nans}. Setting this option may allow faster
5079 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5081 This option should never be turned on by any @option{-O} option since
5082 it can result in incorrect output for programs which depend on
5083 an exact implementation of IEEE or ISO rules/specifications for
5086 The default is @option{-ftrapping-math}.
5088 @item -frounding-math
5089 @opindex frounding-math
5090 Disable transformations and optimizations that assume default floating
5091 point rounding behavior. This is round-to-zero for all floating point
5092 to integer conversions, and round-to-nearest for all other arithmetic
5093 truncations. This option should be specified for programs that change
5094 the FP rounding mode dynamically, or that may be executed with a
5095 non-default rounding mode. This option disables constant folding of
5096 floating point expressions at compile-time (which may be affected by
5097 rounding mode) and arithmetic transformations that are unsafe in the
5098 presence of sign-dependent rounding modes.
5100 The default is @option{-fno-rounding-math}.
5102 This option is experimental and does not currently guarantee to
5103 disable all GCC optimizations that are affected by rounding mode.
5104 Future versions of GCC may provide finer control of this setting
5105 using C99's @code{FENV_ACCESS} pragma. This command line option
5106 will be used to specify the default state for @code{FENV_ACCESS}.
5108 @item -fsignaling-nans
5109 @opindex fsignaling-nans
5110 Compile code assuming that IEEE signaling NaNs may generate user-visible
5111 traps during floating-point operations. Setting this option disables
5112 optimizations that may change the number of exceptions visible with
5113 signaling NaNs. This option implies @option{-ftrapping-math}.
5115 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5118 The default is @option{-fno-signaling-nans}.
5120 This option is experimental and does not currently guarantee to
5121 disable all GCC optimizations that affect signaling NaN behavior.
5123 @item -fsingle-precision-constant
5124 @opindex fsingle-precision-constant
5125 Treat floating point constant as single precision constant instead of
5126 implicitly converting it to double precision constant.
5131 The following options control optimizations that may improve
5132 performance, but are not enabled by any @option{-O} options. This
5133 section includes experimental options that may produce broken code.
5136 @item -fbranch-probabilities
5137 @opindex fbranch-probabilities
5138 After running a program compiled with @option{-fprofile-arcs}
5139 (@pxref{Debugging Options,, Options for Debugging Your Program or
5140 @command{gcc}}), you can compile it a second time using
5141 @option{-fbranch-probabilities}, to improve optimizations based on
5142 the number of times each branch was taken. When the program
5143 compiled with @option{-fprofile-arcs} exits it saves arc execution
5144 counts to a file called @file{@var{sourcename}.gcda} for each source
5145 file The information in this data file is very dependent on the
5146 structure of the generated code, so you must use the same source code
5147 and the same optimization options for both compilations.
5149 With @option{-fbranch-probabilities}, GCC puts a
5150 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5151 These can be used to improve optimization. Currently, they are only
5152 used in one place: in @file{reorg.c}, instead of guessing which path a
5153 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5154 exactly determine which path is taken more often.
5156 @item -fprofile-values
5157 @opindex fprofile-values
5158 If combined with @option{-fprofile-arcs}, it adds code so that some
5159 data about values of expressions in the program is gathered.
5161 With @option{-fbranch-probabilities}, it reads back the data gathered
5162 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5163 notes to instructions for their later usage in optimizations.
5165 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5169 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5170 a code to gather information about values of expressions.
5172 With @option{-fbranch-probabilities}, it reads back the data gathered
5173 and actually performs the optimizations based on them.
5174 Currently the optimizations include specialization of division operation
5175 using the knowledge about the value of the denominator.
5177 @item -fspeculative-prefetching
5178 @opindex fspeculative-prefetching
5179 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5180 a code to gather information about addresses of memory references in the
5183 With @option{-fbranch-probabilities}, it reads back the data gathered
5184 and issues prefetch instructions according to them. In addition to the opportunities
5185 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5186 memory access patterns---for example accesses to the data stored in linked
5187 list whose elements are usually allocated sequentially.
5189 In order to prevent issuing double prefetches, usage of
5190 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5192 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5194 @item -frename-registers
5195 @opindex frename-registers
5196 Attempt to avoid false dependencies in scheduled code by making use
5197 of registers left over after register allocation. This optimization
5198 will most benefit processors with lots of registers. Depending on the
5199 debug information format adopted by the target, however, it can
5200 make debugging impossible, since variables will no longer stay in
5201 a ``home register''.
5203 Not enabled by default at any level because it has known bugs.
5207 Use a graph coloring register allocator. Currently this option is meant
5208 for testing, so we are interested to hear about miscompilations with
5213 Perform tail duplication to enlarge superblock size. This transformation
5214 simplifies the control flow of the function allowing other optimizations to do
5217 Enabled with @option{-fprofile-use}.
5219 @item -funroll-loops
5220 @opindex funroll-loops
5221 Unroll loops whose number of iterations can be determined at compile time or
5222 upon entry to the loop. @option{-funroll-loops} implies
5223 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5224 (i.e.@: complete removal of loops with small constant number of iterations).
5225 This option makes code larger, and may or may not make it run faster.
5227 Enabled with @option{-fprofile-use}.
5229 @item -funroll-all-loops
5230 @opindex funroll-all-loops
5231 Unroll all loops, even if their number of iterations is uncertain when
5232 the loop is entered. This usually makes programs run more slowly.
5233 @option{-funroll-all-loops} implies the same options as
5234 @option{-funroll-loops}.
5237 @opindex fpeel-loops
5238 Peels the loops for that there is enough information that they do not
5239 roll much (from profile feedback). It also turns on complete loop peeling
5240 (i.e.@: complete removal of loops with small constant number of iterations).
5242 Enabled with @option{-fprofile-use}.
5244 @item -fmove-loop-invariants
5245 @opindex fmove-loop-invariants
5246 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5247 at level @option{-O1}
5249 @item -funswitch-loops
5250 @opindex funswitch-loops
5251 Move branches with loop invariant conditions out of the loop, with duplicates
5252 of the loop on both branches (modified according to result of the condition).
5254 @item -fprefetch-loop-arrays
5255 @opindex fprefetch-loop-arrays
5256 If supported by the target machine, generate instructions to prefetch
5257 memory to improve the performance of loops that access large arrays.
5259 Disabled at level @option{-Os}.
5261 @item -ffunction-sections
5262 @itemx -fdata-sections
5263 @opindex ffunction-sections
5264 @opindex fdata-sections
5265 Place each function or data item into its own section in the output
5266 file if the target supports arbitrary sections. The name of the
5267 function or the name of the data item determines the section's name
5270 Use these options on systems where the linker can perform optimizations
5271 to improve locality of reference in the instruction space. Most systems
5272 using the ELF object format and SPARC processors running Solaris 2 have
5273 linkers with such optimizations. AIX may have these optimizations in
5276 Only use these options when there are significant benefits from doing
5277 so. When you specify these options, the assembler and linker will
5278 create larger object and executable files and will also be slower.
5279 You will not be able to use @code{gprof} on all systems if you
5280 specify this option and you may have problems with debugging if
5281 you specify both this option and @option{-g}.
5283 @item -fbranch-target-load-optimize
5284 @opindex fbranch-target-load-optimize
5285 Perform branch target register load optimization before prologue / epilogue
5287 The use of target registers can typically be exposed only during reload,
5288 thus hoisting loads out of loops and doing inter-block scheduling needs
5289 a separate optimization pass.
5291 @item -fbranch-target-load-optimize2
5292 @opindex fbranch-target-load-optimize2
5293 Perform branch target register load optimization after prologue / epilogue
5296 @item -fbtr-bb-exclusive
5297 @opindex fbtr-bb-exclusive
5298 When performing branch target register load optimization, don't reuse
5299 branch target registers in within any basic block.
5301 @item --param @var{name}=@var{value}
5303 In some places, GCC uses various constants to control the amount of
5304 optimization that is done. For example, GCC will not inline functions
5305 that contain more that a certain number of instructions. You can
5306 control some of these constants on the command-line using the
5307 @option{--param} option.
5309 The names of specific parameters, and the meaning of the values, are
5310 tied to the internals of the compiler, and are subject to change
5311 without notice in future releases.
5313 In each case, the @var{value} is an integer. The allowable choices for
5314 @var{name} are given in the following table:
5317 @item sra-max-structure-size
5318 The maximum structure size, in bytes, at which the scalar replacement
5319 of aggregates (SRA) optimization will perform block copies. The
5320 default value, 0, implies that GCC will select the most appropriate
5323 @item sra-field-structure-ratio
5324 The threshold ratio (as a percentage) between instantiated fields and
5325 the complete structure size. We say that if the ratio of the number
5326 of bytes in instantiated fields to the number of bytes in the complete
5327 structure exceeds this parameter, then block copies are not used. The
5330 @item max-crossjump-edges
5331 The maximum number of incoming edges to consider for crossjumping.
5332 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5333 the number of edges incoming to each block. Increasing values mean
5334 more aggressive optimization, making the compile time increase with
5335 probably small improvement in executable size.
5337 @item min-crossjump-insns
5338 The minimum number of instructions which must be matched at the end
5339 of two blocks before crossjumping will be performed on them. This
5340 value is ignored in the case where all instructions in the block being
5341 crossjumped from are matched. The default value is 5.
5343 @item max-delay-slot-insn-search
5344 The maximum number of instructions to consider when looking for an
5345 instruction to fill a delay slot. If more than this arbitrary number of
5346 instructions is searched, the time savings from filling the delay slot
5347 will be minimal so stop searching. Increasing values mean more
5348 aggressive optimization, making the compile time increase with probably
5349 small improvement in executable run time.
5351 @item max-delay-slot-live-search
5352 When trying to fill delay slots, the maximum number of instructions to
5353 consider when searching for a block with valid live register
5354 information. Increasing this arbitrarily chosen value means more
5355 aggressive optimization, increasing the compile time. This parameter
5356 should be removed when the delay slot code is rewritten to maintain the
5359 @item max-gcse-memory
5360 The approximate maximum amount of memory that will be allocated in
5361 order to perform the global common subexpression elimination
5362 optimization. If more memory than specified is required, the
5363 optimization will not be done.
5365 @item max-gcse-passes
5366 The maximum number of passes of GCSE to run. The default is 1.
5368 @item max-pending-list-length
5369 The maximum number of pending dependencies scheduling will allow
5370 before flushing the current state and starting over. Large functions
5371 with few branches or calls can create excessively large lists which
5372 needlessly consume memory and resources.
5374 @item max-inline-insns-single
5375 Several parameters control the tree inliner used in gcc.
5376 This number sets the maximum number of instructions (counted in GCC's
5377 internal representation) in a single function that the tree inliner
5378 will consider for inlining. This only affects functions declared
5379 inline and methods implemented in a class declaration (C++).
5380 The default value is 500.
5382 @item max-inline-insns-auto
5383 When you use @option{-finline-functions} (included in @option{-O3}),
5384 a lot of functions that would otherwise not be considered for inlining
5385 by the compiler will be investigated. To those functions, a different
5386 (more restrictive) limit compared to functions declared inline can
5388 The default value is 120.
5390 @item large-function-insns
5391 The limit specifying really large functions. For functions greater than this
5392 limit inlining is constrained by @option{--param large-function-growth}.
5393 This parameter is useful primarily to avoid extreme compilation time caused by non-linear
5394 algorithms used by the backend.
5395 This parameter is ignored when @option{-funit-at-a-time} is not used.
5396 The default value is 3000.
5398 @item large-function-growth
5399 Specifies maximal growth of large function caused by inlining in percents.
5400 This parameter is ignored when @option{-funit-at-a-time} is not used.
5401 The default value is 200.
5403 @item inline-unit-growth
5404 Specifies maximal overall growth of the compilation unit caused by inlining.
5405 This parameter is ignored when @option{-funit-at-a-time} is not used.
5406 The default value is 150.
5408 @item max-inline-insns-recursive
5409 @itemx max-inline-insns-recursive-auto
5410 Specifies maximum number of instructions out-of-line copy of self recursive inline
5411 function can grow into by performing recursive inlining.
5413 For functions declared inline @option{--param max-inline-insns-recursive} is
5414 taken into acount. For function not declared inline, recursive inlining
5415 happens only when @option{-finline-functions} (included in @option{-O3}) is
5416 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5417 default value is 500.
5419 @item max-inline-recursive-depth
5420 @itemx max-inline-recursive-depth-auto
5421 Specifies maximum recursion depth used by the recursive inlining.
5423 For functions declared inline @option{--param max-inline-recursive-depth} is
5424 taken into acount. For function not declared inline, recursive inlining
5425 happens only when @option{-finline-functions} (included in @option{-O3}) is
5426 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5427 default value is 500.
5429 @item max-inline-insns-rtl
5430 For languages that use the RTL inliner (this happens at a later stage
5431 than tree inlining), you can set the maximum allowable size (counted
5432 in RTL instructions) for the RTL inliner with this parameter.
5433 The default value is 600.
5435 @item max-unrolled-insns
5436 The maximum number of instructions that a loop should have if that loop
5437 is unrolled, and if the loop is unrolled, it determines how many times
5438 the loop code is unrolled.
5440 @item max-average-unrolled-insns
5441 The maximum number of instructions biased by probabilities of their execution
5442 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5443 it determines how many times the loop code is unrolled.
5445 @item max-unroll-times
5446 The maximum number of unrollings of a single loop.
5448 @item max-peeled-insns
5449 The maximum number of instructions that a loop should have if that loop
5450 is peeled, and if the loop is peeled, it determines how many times
5451 the loop code is peeled.
5453 @item max-peel-times
5454 The maximum number of peelings of a single loop.
5456 @item max-completely-peeled-insns
5457 The maximum number of insns of a completely peeled loop.
5459 @item max-completely-peel-times
5460 The maximum number of iterations of a loop to be suitable for complete peeling.
5462 @item max-unswitch-insns
5463 The maximum number of insns of an unswitched loop.
5465 @item max-unswitch-level
5466 The maximum number of branches unswitched in a single loop.
5469 The minimum cost of an expensive expression in the loop invariant motion.
5471 @item iv-consider-all-candidates-bound
5472 Bound on number of candidates for induction variables below that
5473 all candidates are considered for each use in induction variable
5474 optimizations. Only the most relevant candidates are considered
5475 if there are more candidates, to avoid quadratic time complexity.
5477 @item iv-max-considered-uses
5478 The induction variable optimizations give up on loops that contain more
5479 induction variable uses.
5481 @item max-iterations-to-track
5483 The maximum number of iterations of a loop the brute force algorithm
5484 for analysis of # of iterations of the loop tries to evaluate.
5486 @item hot-bb-count-fraction
5487 Select fraction of the maximal count of repetitions of basic block in program
5488 given basic block needs to have to be considered hot.
5490 @item hot-bb-frequency-fraction
5491 Select fraction of the maximal frequency of executions of basic block in
5492 function given basic block needs to have to be considered hot
5494 @item tracer-dynamic-coverage
5495 @itemx tracer-dynamic-coverage-feedback
5497 This value is used to limit superblock formation once the given percentage of
5498 executed instructions is covered. This limits unnecessary code size
5501 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5502 feedback is available. The real profiles (as opposed to statically estimated
5503 ones) are much less balanced allowing the threshold to be larger value.
5505 @item tracer-max-code-growth
5506 Stop tail duplication once code growth has reached given percentage. This is
5507 rather hokey argument, as most of the duplicates will be eliminated later in
5508 cross jumping, so it may be set to much higher values than is the desired code
5511 @item tracer-min-branch-ratio
5513 Stop reverse growth when the reverse probability of best edge is less than this
5514 threshold (in percent).
5516 @item tracer-min-branch-ratio
5517 @itemx tracer-min-branch-ratio-feedback
5519 Stop forward growth if the best edge do have probability lower than this
5522 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5523 compilation for profile feedback and one for compilation without. The value
5524 for compilation with profile feedback needs to be more conservative (higher) in
5525 order to make tracer effective.
5527 @item max-cse-path-length
5529 Maximum number of basic blocks on path that cse considers. The default is 10.
5531 @item global-var-threshold
5533 Counts the number of function calls (@var{n}) and the number of
5534 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5535 single artificial variable will be created to represent all the
5536 call-clobbered variables at function call sites. This artificial
5537 variable will then be made to alias every call-clobbered variable.
5538 (done as @code{int * size_t} on the host machine; beware overflow).
5540 @item max-aliased-vops
5542 Maximum number of virtual operands allowed to represent aliases
5543 before triggering the alias grouping heuristic. Alias grouping
5544 reduces compile times and memory consumption needed for aliasing at
5545 the expense of precision loss in alias information.
5547 @item ggc-min-expand
5549 GCC uses a garbage collector to manage its own memory allocation. This
5550 parameter specifies the minimum percentage by which the garbage
5551 collector's heap should be allowed to expand between collections.
5552 Tuning this may improve compilation speed; it has no effect on code
5555 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5556 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5557 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5558 GCC is not able to calculate RAM on a particular platform, the lower
5559 bound of 30% is used. Setting this parameter and
5560 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5561 every opportunity. This is extremely slow, but can be useful for
5564 @item ggc-min-heapsize
5566 Minimum size of the garbage collector's heap before it begins bothering
5567 to collect garbage. The first collection occurs after the heap expands
5568 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5569 tuning this may improve compilation speed, and has no effect on code
5572 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5573 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5574 with a lower bound of 4096 (four megabytes) and an upper bound of
5575 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5576 particular platform, the lower bound is used. Setting this parameter
5577 very large effectively disables garbage collection. Setting this
5578 parameter and @option{ggc-min-expand} to zero causes a full collection
5579 to occur at every opportunity.
5581 @item max-reload-search-insns
5582 The maximum number of instruction reload should look backward for equivalent
5583 register. Increasing values mean more aggressive optimization, making the
5584 compile time increase with probably slightly better performance. The default
5587 @item max-cselib-memory-location
5588 The maximum number of memory locations cselib should take into acount.
5589 Increasing values mean more aggressive optimization, making the compile time
5590 increase with probably slightly better performance. The default value is 500.
5592 @item reorder-blocks-duplicate
5593 @itemx reorder-blocks-duplicate-feedback
5595 Used by basic block reordering pass to decide whether to use unconditional
5596 branch or duplicate the code on its destination. Code is duplicated when its
5597 estimated size is smaller than this value multiplied by the estimated size of
5598 unconditional jump in the hot spots of the program.
5600 The @option{reorder-block-duplicate-feedback} is used only when profile
5601 feedback is available and may be set to higher values than
5602 @option{reorder-block-duplicate} since information about the hot spots is more
5605 @item max-sched-region-blocks
5606 The maximum number of blocks in a region to be considered for
5607 interblock scheduling. The default value is 10.
5609 @item max-sched-region-insns
5610 The maximum number of insns in a region to be considered for
5611 interblock scheduling. The default value is 100.
5613 @item integer-share-limit
5614 Small integer constants can use a shared data structure, reducing the
5615 compiler's memory usage and increasing its speed. This sets the maximum
5616 value of a shared integer constant's. The default value is 256.
5621 @node Preprocessor Options
5622 @section Options Controlling the Preprocessor
5623 @cindex preprocessor options
5624 @cindex options, preprocessor
5626 These options control the C preprocessor, which is run on each C source
5627 file before actual compilation.
5629 If you use the @option{-E} option, nothing is done except preprocessing.
5630 Some of these options make sense only together with @option{-E} because
5631 they cause the preprocessor output to be unsuitable for actual
5636 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5637 and pass @var{option} directly through to the preprocessor. If
5638 @var{option} contains commas, it is split into multiple options at the
5639 commas. However, many options are modified, translated or interpreted
5640 by the compiler driver before being passed to the preprocessor, and
5641 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5642 interface is undocumented and subject to change, so whenever possible
5643 you should avoid using @option{-Wp} and let the driver handle the
5646 @item -Xpreprocessor @var{option}
5647 @opindex preprocessor
5648 Pass @var{option} as an option to the preprocessor. You can use this to
5649 supply system-specific preprocessor options which GCC does not know how to
5652 If you want to pass an option that takes an argument, you must use
5653 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5656 @include cppopts.texi
5658 @node Assembler Options
5659 @section Passing Options to the Assembler
5661 @c prevent bad page break with this line
5662 You can pass options to the assembler.
5665 @item -Wa,@var{option}
5667 Pass @var{option} as an option to the assembler. If @var{option}
5668 contains commas, it is split into multiple options at the commas.
5670 @item -Xassembler @var{option}
5672 Pass @var{option} as an option to the assembler. You can use this to
5673 supply system-specific assembler options which GCC does not know how to
5676 If you want to pass an option that takes an argument, you must use
5677 @option{-Xassembler} twice, once for the option and once for the argument.
5682 @section Options for Linking
5683 @cindex link options
5684 @cindex options, linking
5686 These options come into play when the compiler links object files into
5687 an executable output file. They are meaningless if the compiler is
5688 not doing a link step.
5692 @item @var{object-file-name}
5693 A file name that does not end in a special recognized suffix is
5694 considered to name an object file or library. (Object files are
5695 distinguished from libraries by the linker according to the file
5696 contents.) If linking is done, these object files are used as input
5705 If any of these options is used, then the linker is not run, and
5706 object file names should not be used as arguments. @xref{Overall
5710 @item -l@var{library}
5711 @itemx -l @var{library}
5713 Search the library named @var{library} when linking. (The second
5714 alternative with the library as a separate argument is only for
5715 POSIX compliance and is not recommended.)
5717 It makes a difference where in the command you write this option; the
5718 linker searches and processes libraries and object files in the order they
5719 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5720 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5721 to functions in @samp{z}, those functions may not be loaded.
5723 The linker searches a standard list of directories for the library,
5724 which is actually a file named @file{lib@var{library}.a}. The linker
5725 then uses this file as if it had been specified precisely by name.
5727 The directories searched include several standard system directories
5728 plus any that you specify with @option{-L}.
5730 Normally the files found this way are library files---archive files
5731 whose members are object files. The linker handles an archive file by
5732 scanning through it for members which define symbols that have so far
5733 been referenced but not defined. But if the file that is found is an
5734 ordinary object file, it is linked in the usual fashion. The only
5735 difference between using an @option{-l} option and specifying a file name
5736 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5737 and searches several directories.
5741 You need this special case of the @option{-l} option in order to
5742 link an Objective-C or Objective-C++ program.
5745 @opindex nostartfiles
5746 Do not use the standard system startup files when linking.
5747 The standard system libraries are used normally, unless @option{-nostdlib}
5748 or @option{-nodefaultlibs} is used.
5750 @item -nodefaultlibs
5751 @opindex nodefaultlibs
5752 Do not use the standard system libraries when linking.
5753 Only the libraries you specify will be passed to the linker.
5754 The standard startup files are used normally, unless @option{-nostartfiles}
5755 is used. The compiler may generate calls to @code{memcmp},
5756 @code{memset}, @code{memcpy} and @code{memmove}.
5757 These entries are usually resolved by entries in
5758 libc. These entry points should be supplied through some other
5759 mechanism when this option is specified.
5763 Do not use the standard system startup files or libraries when linking.
5764 No startup files and only the libraries you specify will be passed to
5765 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5766 @code{memcpy} and @code{memmove}.
5767 These entries are usually resolved by entries in
5768 libc. These entry points should be supplied through some other
5769 mechanism when this option is specified.
5771 @cindex @option{-lgcc}, use with @option{-nostdlib}
5772 @cindex @option{-nostdlib} and unresolved references
5773 @cindex unresolved references and @option{-nostdlib}
5774 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5775 @cindex @option{-nodefaultlibs} and unresolved references
5776 @cindex unresolved references and @option{-nodefaultlibs}
5777 One of the standard libraries bypassed by @option{-nostdlib} and
5778 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5779 that GCC uses to overcome shortcomings of particular machines, or special
5780 needs for some languages.
5781 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5782 Collection (GCC) Internals},
5783 for more discussion of @file{libgcc.a}.)
5784 In most cases, you need @file{libgcc.a} even when you want to avoid
5785 other standard libraries. In other words, when you specify @option{-nostdlib}
5786 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5787 This ensures that you have no unresolved references to internal GCC
5788 library subroutines. (For example, @samp{__main}, used to ensure C++
5789 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5790 GNU Compiler Collection (GCC) Internals}.)
5794 Produce a position independent executable on targets which support it.
5795 For predictable results, you must also specify the same set of options
5796 that were used to generate code (@option{-fpie}, @option{-fPIE},
5797 or model suboptions) when you specify this option.
5801 Remove all symbol table and relocation information from the executable.
5805 On systems that support dynamic linking, this prevents linking with the shared
5806 libraries. On other systems, this option has no effect.
5810 Produce a shared object which can then be linked with other objects to
5811 form an executable. Not all systems support this option. For predictable
5812 results, you must also specify the same set of options that were used to
5813 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5814 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5815 needs to build supplementary stub code for constructors to work. On
5816 multi-libbed systems, @samp{gcc -shared} must select the correct support
5817 libraries to link against. Failing to supply the correct flags may lead
5818 to subtle defects. Supplying them in cases where they are not necessary
5821 @item -shared-libgcc
5822 @itemx -static-libgcc
5823 @opindex shared-libgcc
5824 @opindex static-libgcc
5825 On systems that provide @file{libgcc} as a shared library, these options
5826 force the use of either the shared or static version respectively.
5827 If no shared version of @file{libgcc} was built when the compiler was
5828 configured, these options have no effect.
5830 There are several situations in which an application should use the
5831 shared @file{libgcc} instead of the static version. The most common
5832 of these is when the application wishes to throw and catch exceptions
5833 across different shared libraries. In that case, each of the libraries
5834 as well as the application itself should use the shared @file{libgcc}.
5836 Therefore, the G++ and GCJ drivers automatically add
5837 @option{-shared-libgcc} whenever you build a shared library or a main
5838 executable, because C++ and Java programs typically use exceptions, so
5839 this is the right thing to do.
5841 If, instead, you use the GCC driver to create shared libraries, you may
5842 find that they will not always be linked with the shared @file{libgcc}.
5843 If GCC finds, at its configuration time, that you have a non-GNU linker
5844 or a GNU linker that does not support option @option{--eh-frame-hdr},
5845 it will link the shared version of @file{libgcc} into shared libraries
5846 by default. Otherwise, it will take advantage of the linker and optimize
5847 away the linking with the shared version of @file{libgcc}, linking with
5848 the static version of libgcc by default. This allows exceptions to
5849 propagate through such shared libraries, without incurring relocation
5850 costs at library load time.
5852 However, if a library or main executable is supposed to throw or catch
5853 exceptions, you must link it using the G++ or GCJ driver, as appropriate
5854 for the languages used in the program, or using the option
5855 @option{-shared-libgcc}, such that it is linked with the shared
5860 Bind references to global symbols when building a shared object. Warn
5861 about any unresolved references (unless overridden by the link editor
5862 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
5865 @item -Xlinker @var{option}
5867 Pass @var{option} as an option to the linker. You can use this to
5868 supply system-specific linker options which GCC does not know how to
5871 If you want to pass an option that takes an argument, you must use
5872 @option{-Xlinker} twice, once for the option and once for the argument.
5873 For example, to pass @option{-assert definitions}, you must write
5874 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
5875 @option{-Xlinker "-assert definitions"}, because this passes the entire
5876 string as a single argument, which is not what the linker expects.
5878 @item -Wl,@var{option}
5880 Pass @var{option} as an option to the linker. If @var{option} contains
5881 commas, it is split into multiple options at the commas.
5883 @item -u @var{symbol}
5885 Pretend the symbol @var{symbol} is undefined, to force linking of
5886 library modules to define it. You can use @option{-u} multiple times with
5887 different symbols to force loading of additional library modules.
5890 @node Directory Options
5891 @section Options for Directory Search
5892 @cindex directory options
5893 @cindex options, directory search
5896 These options specify directories to search for header files, for
5897 libraries and for parts of the compiler:
5902 Add the directory @var{dir} to the head of the list of directories to be
5903 searched for header files. This can be used to override a system header
5904 file, substituting your own version, since these directories are
5905 searched before the system header file directories. However, you should
5906 not use this option to add directories that contain vendor-supplied
5907 system header files (use @option{-isystem} for that). If you use more than
5908 one @option{-I} option, the directories are scanned in left-to-right
5909 order; the standard system directories come after.
5911 If a standard system include directory, or a directory specified with
5912 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
5913 option will be ignored. The directory will still be searched but as a
5914 system directory at its normal position in the system include chain.
5915 This is to ensure that GCC's procedure to fix buggy system headers and
5916 the ordering for the include_next directive are not inadvertently changed.
5917 If you really need to change the search order for system directories,
5918 use the @option{-nostdinc} and/or @option{-isystem} options.
5920 @item -iquote@var{dir}
5922 Add the directory @var{dir} to the head of the list of directories to
5923 be searched for header files only for the case of @samp{#include
5924 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
5925 otherwise just like @option{-I}.
5929 Add directory @var{dir} to the list of directories to be searched
5932 @item -B@var{prefix}
5934 This option specifies where to find the executables, libraries,
5935 include files, and data files of the compiler itself.
5937 The compiler driver program runs one or more of the subprograms
5938 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
5939 @var{prefix} as a prefix for each program it tries to run, both with and
5940 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
5942 For each subprogram to be run, the compiler driver first tries the
5943 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
5944 was not specified, the driver tries two standard prefixes, which are
5945 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
5946 those results in a file name that is found, the unmodified program
5947 name is searched for using the directories specified in your
5948 @env{PATH} environment variable.
5950 The compiler will check to see if the path provided by the @option{-B}
5951 refers to a directory, and if necessary it will add a directory
5952 separator character at the end of the path.
5954 @option{-B} prefixes that effectively specify directory names also apply
5955 to libraries in the linker, because the compiler translates these
5956 options into @option{-L} options for the linker. They also apply to
5957 includes files in the preprocessor, because the compiler translates these
5958 options into @option{-isystem} options for the preprocessor. In this case,
5959 the compiler appends @samp{include} to the prefix.
5961 The run-time support file @file{libgcc.a} can also be searched for using
5962 the @option{-B} prefix, if needed. If it is not found there, the two
5963 standard prefixes above are tried, and that is all. The file is left
5964 out of the link if it is not found by those means.
5966 Another way to specify a prefix much like the @option{-B} prefix is to use
5967 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
5970 As a special kludge, if the path provided by @option{-B} is
5971 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
5972 9, then it will be replaced by @file{[dir/]include}. This is to help
5973 with boot-strapping the compiler.
5975 @item -specs=@var{file}
5977 Process @var{file} after the compiler reads in the standard @file{specs}
5978 file, in order to override the defaults that the @file{gcc} driver
5979 program uses when determining what switches to pass to @file{cc1},
5980 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
5981 @option{-specs=@var{file}} can be specified on the command line, and they
5982 are processed in order, from left to right.
5986 This option has been deprecated. Please use @option{-iquote} instead for
5987 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
5988 Any directories you specify with @option{-I} options before the @option{-I-}
5989 option are searched only for the case of @samp{#include "@var{file}"};
5990 they are not searched for @samp{#include <@var{file}>}.
5992 If additional directories are specified with @option{-I} options after
5993 the @option{-I-}, these directories are searched for all @samp{#include}
5994 directives. (Ordinarily @emph{all} @option{-I} directories are used
5997 In addition, the @option{-I-} option inhibits the use of the current
5998 directory (where the current input file came from) as the first search
5999 directory for @samp{#include "@var{file}"}. There is no way to
6000 override this effect of @option{-I-}. With @option{-I.} you can specify
6001 searching the directory which was current when the compiler was
6002 invoked. That is not exactly the same as what the preprocessor does
6003 by default, but it is often satisfactory.
6005 @option{-I-} does not inhibit the use of the standard system directories
6006 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6013 @section Specifying subprocesses and the switches to pass to them
6016 @command{gcc} is a driver program. It performs its job by invoking a
6017 sequence of other programs to do the work of compiling, assembling and
6018 linking. GCC interprets its command-line parameters and uses these to
6019 deduce which programs it should invoke, and which command-line options
6020 it ought to place on their command lines. This behavior is controlled
6021 by @dfn{spec strings}. In most cases there is one spec string for each
6022 program that GCC can invoke, but a few programs have multiple spec
6023 strings to control their behavior. The spec strings built into GCC can
6024 be overridden by using the @option{-specs=} command-line switch to specify
6027 @dfn{Spec files} are plaintext files that are used to construct spec
6028 strings. They consist of a sequence of directives separated by blank
6029 lines. The type of directive is determined by the first non-whitespace
6030 character on the line and it can be one of the following:
6033 @item %@var{command}
6034 Issues a @var{command} to the spec file processor. The commands that can
6038 @item %include <@var{file}>
6040 Search for @var{file} and insert its text at the current point in the
6043 @item %include_noerr <@var{file}>
6044 @cindex %include_noerr
6045 Just like @samp{%include}, but do not generate an error message if the include
6046 file cannot be found.
6048 @item %rename @var{old_name} @var{new_name}
6050 Rename the spec string @var{old_name} to @var{new_name}.
6054 @item *[@var{spec_name}]:
6055 This tells the compiler to create, override or delete the named spec
6056 string. All lines after this directive up to the next directive or
6057 blank line are considered to be the text for the spec string. If this
6058 results in an empty string then the spec will be deleted. (Or, if the
6059 spec did not exist, then nothing will happened.) Otherwise, if the spec
6060 does not currently exist a new spec will be created. If the spec does
6061 exist then its contents will be overridden by the text of this
6062 directive, unless the first character of that text is the @samp{+}
6063 character, in which case the text will be appended to the spec.
6065 @item [@var{suffix}]:
6066 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6067 and up to the next directive or blank line are considered to make up the
6068 spec string for the indicated suffix. When the compiler encounters an
6069 input file with the named suffix, it will processes the spec string in
6070 order to work out how to compile that file. For example:
6077 This says that any input file whose name ends in @samp{.ZZ} should be
6078 passed to the program @samp{z-compile}, which should be invoked with the
6079 command-line switch @option{-input} and with the result of performing the
6080 @samp{%i} substitution. (See below.)
6082 As an alternative to providing a spec string, the text that follows a
6083 suffix directive can be one of the following:
6086 @item @@@var{language}
6087 This says that the suffix is an alias for a known @var{language}. This is
6088 similar to using the @option{-x} command-line switch to GCC to specify a
6089 language explicitly. For example:
6096 Says that .ZZ files are, in fact, C++ source files.
6099 This causes an error messages saying:
6102 @var{name} compiler not installed on this system.
6106 GCC already has an extensive list of suffixes built into it.
6107 This directive will add an entry to the end of the list of suffixes, but
6108 since the list is searched from the end backwards, it is effectively
6109 possible to override earlier entries using this technique.
6113 GCC has the following spec strings built into it. Spec files can
6114 override these strings or create their own. Note that individual
6115 targets can also add their own spec strings to this list.
6118 asm Options to pass to the assembler
6119 asm_final Options to pass to the assembler post-processor
6120 cpp Options to pass to the C preprocessor
6121 cc1 Options to pass to the C compiler
6122 cc1plus Options to pass to the C++ compiler
6123 endfile Object files to include at the end of the link
6124 link Options to pass to the linker
6125 lib Libraries to include on the command line to the linker
6126 libgcc Decides which GCC support library to pass to the linker
6127 linker Sets the name of the linker
6128 predefines Defines to be passed to the C preprocessor
6129 signed_char Defines to pass to CPP to say whether @code{char} is signed
6131 startfile Object files to include at the start of the link
6134 Here is a small example of a spec file:
6140 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6143 This example renames the spec called @samp{lib} to @samp{old_lib} and
6144 then overrides the previous definition of @samp{lib} with a new one.
6145 The new definition adds in some extra command-line options before
6146 including the text of the old definition.
6148 @dfn{Spec strings} are a list of command-line options to be passed to their
6149 corresponding program. In addition, the spec strings can contain
6150 @samp{%}-prefixed sequences to substitute variable text or to
6151 conditionally insert text into the command line. Using these constructs
6152 it is possible to generate quite complex command lines.
6154 Here is a table of all defined @samp{%}-sequences for spec
6155 strings. Note that spaces are not generated automatically around the
6156 results of expanding these sequences. Therefore you can concatenate them
6157 together or combine them with constant text in a single argument.
6161 Substitute one @samp{%} into the program name or argument.
6164 Substitute the name of the input file being processed.
6167 Substitute the basename of the input file being processed.
6168 This is the substring up to (and not including) the last period
6169 and not including the directory.
6172 This is the same as @samp{%b}, but include the file suffix (text after
6176 Marks the argument containing or following the @samp{%d} as a
6177 temporary file name, so that that file will be deleted if GCC exits
6178 successfully. Unlike @samp{%g}, this contributes no text to the
6181 @item %g@var{suffix}
6182 Substitute a file name that has suffix @var{suffix} and is chosen
6183 once per compilation, and mark the argument in the same way as
6184 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6185 name is now chosen in a way that is hard to predict even when previously
6186 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6187 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6188 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6189 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6190 was simply substituted with a file name chosen once per compilation,
6191 without regard to any appended suffix (which was therefore treated
6192 just like ordinary text), making such attacks more likely to succeed.
6194 @item %u@var{suffix}
6195 Like @samp{%g}, but generates a new temporary file name even if
6196 @samp{%u@var{suffix}} was already seen.
6198 @item %U@var{suffix}
6199 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6200 new one if there is no such last file name. In the absence of any
6201 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6202 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6203 would involve the generation of two distinct file names, one
6204 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6205 simply substituted with a file name chosen for the previous @samp{%u},
6206 without regard to any appended suffix.
6208 @item %j@var{suffix}
6209 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6210 writable, and if save-temps is off; otherwise, substitute the name
6211 of a temporary file, just like @samp{%u}. This temporary file is not
6212 meant for communication between processes, but rather as a junk
6215 @item %|@var{suffix}
6216 @itemx %m@var{suffix}
6217 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6218 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6219 all. These are the two most common ways to instruct a program that it
6220 should read from standard input or write to standard output. If you
6221 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6222 construct: see for example @file{f/lang-specs.h}.
6224 @item %.@var{SUFFIX}
6225 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6226 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6227 terminated by the next space or %.
6230 Marks the argument containing or following the @samp{%w} as the
6231 designated output file of this compilation. This puts the argument
6232 into the sequence of arguments that @samp{%o} will substitute later.
6235 Substitutes the names of all the output files, with spaces
6236 automatically placed around them. You should write spaces
6237 around the @samp{%o} as well or the results are undefined.
6238 @samp{%o} is for use in the specs for running the linker.
6239 Input files whose names have no recognized suffix are not compiled
6240 at all, but they are included among the output files, so they will
6244 Substitutes the suffix for object files. Note that this is
6245 handled specially when it immediately follows @samp{%g, %u, or %U},
6246 because of the need for those to form complete file names. The
6247 handling is such that @samp{%O} is treated exactly as if it had already
6248 been substituted, except that @samp{%g, %u, and %U} do not currently
6249 support additional @var{suffix} characters following @samp{%O} as they would
6250 following, for example, @samp{.o}.
6253 Substitutes the standard macro predefinitions for the
6254 current target machine. Use this when running @code{cpp}.
6257 Like @samp{%p}, but puts @samp{__} before and after the name of each
6258 predefined macro, except for macros that start with @samp{__} or with
6259 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6263 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6264 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6265 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6269 Current argument is the name of a library or startup file of some sort.
6270 Search for that file in a standard list of directories and substitute
6271 the full name found.
6274 Print @var{str} as an error message. @var{str} is terminated by a newline.
6275 Use this when inconsistent options are detected.
6278 Substitute the contents of spec string @var{name} at this point.
6281 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6283 @item %x@{@var{option}@}
6284 Accumulate an option for @samp{%X}.
6287 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6291 Output the accumulated assembler options specified by @option{-Wa}.
6294 Output the accumulated preprocessor options specified by @option{-Wp}.
6297 Process the @code{asm} spec. This is used to compute the
6298 switches to be passed to the assembler.
6301 Process the @code{asm_final} spec. This is a spec string for
6302 passing switches to an assembler post-processor, if such a program is
6306 Process the @code{link} spec. This is the spec for computing the
6307 command line passed to the linker. Typically it will make use of the
6308 @samp{%L %G %S %D and %E} sequences.
6311 Dump out a @option{-L} option for each directory that GCC believes might
6312 contain startup files. If the target supports multilibs then the
6313 current multilib directory will be prepended to each of these paths.
6316 Output the multilib directory with directory separators replaced with
6317 @samp{_}. If multilib directories are not set, or the multilib directory is
6318 @file{.} then this option emits nothing.
6321 Process the @code{lib} spec. This is a spec string for deciding which
6322 libraries should be included on the command line to the linker.
6325 Process the @code{libgcc} spec. This is a spec string for deciding
6326 which GCC support library should be included on the command line to the linker.
6329 Process the @code{startfile} spec. This is a spec for deciding which
6330 object files should be the first ones passed to the linker. Typically
6331 this might be a file named @file{crt0.o}.
6334 Process the @code{endfile} spec. This is a spec string that specifies
6335 the last object files that will be passed to the linker.
6338 Process the @code{cpp} spec. This is used to construct the arguments
6339 to be passed to the C preprocessor.
6342 Process the @code{cc1} spec. This is used to construct the options to be
6343 passed to the actual C compiler (@samp{cc1}).
6346 Process the @code{cc1plus} spec. This is used to construct the options to be
6347 passed to the actual C++ compiler (@samp{cc1plus}).
6350 Substitute the variable part of a matched option. See below.
6351 Note that each comma in the substituted string is replaced by
6355 Remove all occurrences of @code{-S} from the command line. Note---this
6356 command is position dependent. @samp{%} commands in the spec string
6357 before this one will see @code{-S}, @samp{%} commands in the spec string
6358 after this one will not.
6360 @item %:@var{function}(@var{args})
6361 Call the named function @var{function}, passing it @var{args}.
6362 @var{args} is first processed as a nested spec string, then split
6363 into an argument vector in the usual fashion. The function returns
6364 a string which is processed as if it had appeared literally as part
6365 of the current spec.
6367 The following built-in spec functions are provided:
6370 @item @code{if-exists}
6371 The @code{if-exists} spec function takes one argument, an absolute
6372 pathname to a file. If the file exists, @code{if-exists} returns the
6373 pathname. Here is a small example of its usage:
6377 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6380 @item @code{if-exists-else}
6381 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6382 spec function, except that it takes two arguments. The first argument is
6383 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6384 returns the pathname. If it does not exist, it returns the second argument.
6385 This way, @code{if-exists-else} can be used to select one file or another,
6386 based on the existence of the first. Here is a small example of its usage:
6390 crt0%O%s %:if-exists(crti%O%s) \
6391 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6394 @item @code{replace-outfile}
6395 The @code{replace-outfile} spec function takes two arguments. It looks for the
6396 first argument in the outfiles array and replaces it with the second argument. Here
6397 is a small example of its usage:
6400 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6406 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6407 If that switch was not specified, this substitutes nothing. Note that
6408 the leading dash is omitted when specifying this option, and it is
6409 automatically inserted if the substitution is performed. Thus the spec
6410 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6411 and would output the command line option @option{-foo}.
6413 @item %W@{@code{S}@}
6414 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6417 @item %@{@code{S}*@}
6418 Substitutes all the switches specified to GCC whose names start
6419 with @code{-S}, but which also take an argument. This is used for
6420 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6421 GCC considers @option{-o foo} as being
6422 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6423 text, including the space. Thus two arguments would be generated.
6425 @item %@{@code{S}*&@code{T}*@}
6426 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6427 (the order of @code{S} and @code{T} in the spec is not significant).
6428 There can be any number of ampersand-separated variables; for each the
6429 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6431 @item %@{@code{S}:@code{X}@}
6432 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6434 @item %@{!@code{S}:@code{X}@}
6435 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6437 @item %@{@code{S}*:@code{X}@}
6438 Substitutes @code{X} if one or more switches whose names start with
6439 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6440 once, no matter how many such switches appeared. However, if @code{%*}
6441 appears somewhere in @code{X}, then @code{X} will be substituted once
6442 for each matching switch, with the @code{%*} replaced by the part of
6443 that switch that matched the @code{*}.
6445 @item %@{.@code{S}:@code{X}@}
6446 Substitutes @code{X}, if processing a file with suffix @code{S}.
6448 @item %@{!.@code{S}:@code{X}@}
6449 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6451 @item %@{@code{S}|@code{P}:@code{X}@}
6452 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6453 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6454 although they have a stronger binding than the @samp{|}. If @code{%*}
6455 appears in @code{X}, all of the alternatives must be starred, and only
6456 the first matching alternative is substituted.
6458 For example, a spec string like this:
6461 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6464 will output the following command-line options from the following input
6465 command-line options:
6470 -d fred.c -foo -baz -boggle
6471 -d jim.d -bar -baz -boggle
6474 @item %@{S:X; T:Y; :D@}
6476 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6477 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6478 be as many clauses as you need. This may be combined with @code{.},
6479 @code{!}, @code{|}, and @code{*} as needed.
6484 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6485 construct may contain other nested @samp{%} constructs or spaces, or
6486 even newlines. They are processed as usual, as described above.
6487 Trailing white space in @code{X} is ignored. White space may also
6488 appear anywhere on the left side of the colon in these constructs,
6489 except between @code{.} or @code{*} and the corresponding word.
6491 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6492 handled specifically in these constructs. If another value of
6493 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6494 @option{-W} switch is found later in the command line, the earlier
6495 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6496 just one letter, which passes all matching options.
6498 The character @samp{|} at the beginning of the predicate text is used to
6499 indicate that a command should be piped to the following command, but
6500 only if @option{-pipe} is specified.
6502 It is built into GCC which switches take arguments and which do not.
6503 (You might think it would be useful to generalize this to allow each
6504 compiler's spec to say which switches take arguments. But this cannot
6505 be done in a consistent fashion. GCC cannot even decide which input
6506 files have been specified without knowing which switches take arguments,
6507 and it must know which input files to compile in order to tell which
6510 GCC also knows implicitly that arguments starting in @option{-l} are to be
6511 treated as compiler output files, and passed to the linker in their
6512 proper position among the other output files.
6514 @c man begin OPTIONS
6516 @node Target Options
6517 @section Specifying Target Machine and Compiler Version
6518 @cindex target options
6519 @cindex cross compiling
6520 @cindex specifying machine version
6521 @cindex specifying compiler version and target machine
6522 @cindex compiler version, specifying
6523 @cindex target machine, specifying
6525 The usual way to run GCC is to run the executable called @file{gcc}, or
6526 @file{<machine>-gcc} when cross-compiling, or
6527 @file{<machine>-gcc-<version>} to run a version other than the one that
6528 was installed last. Sometimes this is inconvenient, so GCC provides
6529 options that will switch to another cross-compiler or version.
6532 @item -b @var{machine}
6534 The argument @var{machine} specifies the target machine for compilation.
6536 The value to use for @var{machine} is the same as was specified as the
6537 machine type when configuring GCC as a cross-compiler. For
6538 example, if a cross-compiler was configured with @samp{configure
6539 i386v}, meaning to compile for an 80386 running System V, then you
6540 would specify @option{-b i386v} to run that cross compiler.
6542 @item -V @var{version}
6544 The argument @var{version} specifies which version of GCC to run.
6545 This is useful when multiple versions are installed. For example,
6546 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6549 The @option{-V} and @option{-b} options work by running the
6550 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6551 use them if you can just run that directly.
6553 @node Submodel Options
6554 @section Hardware Models and Configurations
6555 @cindex submodel options
6556 @cindex specifying hardware config
6557 @cindex hardware models and configurations, specifying
6558 @cindex machine dependent options
6560 Earlier we discussed the standard option @option{-b} which chooses among
6561 different installed compilers for completely different target
6562 machines, such as VAX vs.@: 68000 vs.@: 80386.
6564 In addition, each of these target machine types can have its own
6565 special options, starting with @samp{-m}, to choose among various
6566 hardware models or configurations---for example, 68010 vs 68020,
6567 floating coprocessor or none. A single installed version of the
6568 compiler can compile for any model or configuration, according to the
6571 Some configurations of the compiler also support additional special
6572 options, usually for compatibility with other compilers on the same
6575 These options are defined by the macro @code{TARGET_SWITCHES} in the
6576 machine description. The default for the options is also defined by
6577 that macro, which enables you to change the defaults.
6579 @c This list is ordered alphanumerically by subsection name.
6580 @c It should be the same order and spelling as these options are listed
6581 @c in Machine Dependent Options
6589 * DEC Alpha Options::
6590 * DEC Alpha/VMS Options::
6594 * i386 and x86-64 Options::
6606 * RS/6000 and PowerPC Options::
6607 * S/390 and zSeries Options::
6610 * System V Options::
6611 * TMS320C3x/C4x Options::
6615 * Xstormy16 Options::
6621 @subsection ARC Options
6624 These options are defined for ARC implementations:
6629 Compile code for little endian mode. This is the default.
6633 Compile code for big endian mode.
6636 @opindex mmangle-cpu
6637 Prepend the name of the cpu to all public symbol names.
6638 In multiple-processor systems, there are many ARC variants with different
6639 instruction and register set characteristics. This flag prevents code
6640 compiled for one cpu to be linked with code compiled for another.
6641 No facility exists for handling variants that are ``almost identical''.
6642 This is an all or nothing option.
6644 @item -mcpu=@var{cpu}
6646 Compile code for ARC variant @var{cpu}.
6647 Which variants are supported depend on the configuration.
6648 All variants support @option{-mcpu=base}, this is the default.
6650 @item -mtext=@var{text-section}
6651 @itemx -mdata=@var{data-section}
6652 @itemx -mrodata=@var{readonly-data-section}
6656 Put functions, data, and readonly data in @var{text-section},
6657 @var{data-section}, and @var{readonly-data-section} respectively
6658 by default. This can be overridden with the @code{section} attribute.
6659 @xref{Variable Attributes}.
6664 @subsection ARM Options
6667 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6671 @item -mabi=@var{name}
6673 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6674 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6677 @opindex mapcs-frame
6678 Generate a stack frame that is compliant with the ARM Procedure Call
6679 Standard for all functions, even if this is not strictly necessary for
6680 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6681 with this option will cause the stack frames not to be generated for
6682 leaf functions. The default is @option{-mno-apcs-frame}.
6686 This is a synonym for @option{-mapcs-frame}.
6689 @c not currently implemented
6690 @item -mapcs-stack-check
6691 @opindex mapcs-stack-check
6692 Generate code to check the amount of stack space available upon entry to
6693 every function (that actually uses some stack space). If there is
6694 insufficient space available then either the function
6695 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6696 called, depending upon the amount of stack space required. The run time
6697 system is required to provide these functions. The default is
6698 @option{-mno-apcs-stack-check}, since this produces smaller code.
6700 @c not currently implemented
6702 @opindex mapcs-float
6703 Pass floating point arguments using the float point registers. This is
6704 one of the variants of the APCS@. This option is recommended if the
6705 target hardware has a floating point unit or if a lot of floating point
6706 arithmetic is going to be performed by the code. The default is
6707 @option{-mno-apcs-float}, since integer only code is slightly increased in
6708 size if @option{-mapcs-float} is used.
6710 @c not currently implemented
6711 @item -mapcs-reentrant
6712 @opindex mapcs-reentrant
6713 Generate reentrant, position independent code. The default is
6714 @option{-mno-apcs-reentrant}.
6717 @item -mthumb-interwork
6718 @opindex mthumb-interwork
6719 Generate code which supports calling between the ARM and Thumb
6720 instruction sets. Without this option the two instruction sets cannot
6721 be reliably used inside one program. The default is
6722 @option{-mno-thumb-interwork}, since slightly larger code is generated
6723 when @option{-mthumb-interwork} is specified.
6725 @item -mno-sched-prolog
6726 @opindex mno-sched-prolog
6727 Prevent the reordering of instructions in the function prolog, or the
6728 merging of those instruction with the instructions in the function's
6729 body. This means that all functions will start with a recognizable set
6730 of instructions (or in fact one of a choice from a small set of
6731 different function prologues), and this information can be used to
6732 locate the start if functions inside an executable piece of code. The
6733 default is @option{-msched-prolog}.
6736 @opindex mhard-float
6737 Generate output containing floating point instructions. This is the
6741 @opindex msoft-float
6742 Generate output containing library calls for floating point.
6743 @strong{Warning:} the requisite libraries are not available for all ARM
6744 targets. Normally the facilities of the machine's usual C compiler are
6745 used, but this cannot be done directly in cross-compilation. You must make
6746 your own arrangements to provide suitable library functions for
6749 @option{-msoft-float} changes the calling convention in the output file;
6750 therefore, it is only useful if you compile @emph{all} of a program with
6751 this option. In particular, you need to compile @file{libgcc.a}, the
6752 library that comes with GCC, with @option{-msoft-float} in order for
6755 @item -mfloat-abi=@var{name}
6757 Specifies which ABI to use for floating point values. Permissible values
6758 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6760 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6761 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6762 of floating point instructions, but still uses the soft-float calling
6765 @item -mlittle-endian
6766 @opindex mlittle-endian
6767 Generate code for a processor running in little-endian mode. This is
6768 the default for all standard configurations.
6771 @opindex mbig-endian
6772 Generate code for a processor running in big-endian mode; the default is
6773 to compile code for a little-endian processor.
6775 @item -mwords-little-endian
6776 @opindex mwords-little-endian
6777 This option only applies when generating code for big-endian processors.
6778 Generate code for a little-endian word order but a big-endian byte
6779 order. That is, a byte order of the form @samp{32107654}. Note: this
6780 option should only be used if you require compatibility with code for
6781 big-endian ARM processors generated by versions of the compiler prior to
6784 @item -mcpu=@var{name}
6786 This specifies the name of the target ARM processor. GCC uses this name
6787 to determine what kind of instructions it can emit when generating
6788 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6789 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6790 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6791 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6792 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6793 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
6794 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6795 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6796 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
6797 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
6798 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
6799 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
6800 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
6801 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
6804 @itemx -mtune=@var{name}
6806 This option is very similar to the @option{-mcpu=} option, except that
6807 instead of specifying the actual target processor type, and hence
6808 restricting which instructions can be used, it specifies that GCC should
6809 tune the performance of the code as if the target were of the type
6810 specified in this option, but still choosing the instructions that it
6811 will generate based on the cpu specified by a @option{-mcpu=} option.
6812 For some ARM implementations better performance can be obtained by using
6815 @item -march=@var{name}
6817 This specifies the name of the target ARM architecture. GCC uses this
6818 name to determine what kind of instructions it can emit when generating
6819 assembly code. This option can be used in conjunction with or instead
6820 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6821 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6822 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6823 @samp{iwmmxt}, @samp{ep9312}.
6825 @item -mfpu=@var{name}
6826 @itemx -mfpe=@var{number}
6827 @itemx -mfp=@var{number}
6831 This specifies what floating point hardware (or hardware emulation) is
6832 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6833 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6834 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6835 with older versions of GCC@.
6837 If @option{-msoft-float} is specified this specifies the format of
6838 floating point values.
6840 @item -mstructure-size-boundary=@var{n}
6841 @opindex mstructure-size-boundary
6842 The size of all structures and unions will be rounded up to a multiple
6843 of the number of bits set by this option. Permissible values are 8, 32
6844 and 64. The default value varies for different toolchains. For the COFF
6845 targeted toolchain the default value is 8. A value of 64 is only allowed
6846 if the underlying ABI supports it.
6848 Specifying the larger number can produce faster, more efficient code, but
6849 can also increase the size of the program. Different values are potentially
6850 incompatible. Code compiled with one value cannot necessarily expect to
6851 work with code or libraries compiled with another value, if they exchange
6852 information using structures or unions.
6854 @item -mabort-on-noreturn
6855 @opindex mabort-on-noreturn
6856 Generate a call to the function @code{abort} at the end of a
6857 @code{noreturn} function. It will be executed if the function tries to
6861 @itemx -mno-long-calls
6862 @opindex mlong-calls
6863 @opindex mno-long-calls
6864 Tells the compiler to perform function calls by first loading the
6865 address of the function into a register and then performing a subroutine
6866 call on this register. This switch is needed if the target function
6867 will lie outside of the 64 megabyte addressing range of the offset based
6868 version of subroutine call instruction.
6870 Even if this switch is enabled, not all function calls will be turned
6871 into long calls. The heuristic is that static functions, functions
6872 which have the @samp{short-call} attribute, functions that are inside
6873 the scope of a @samp{#pragma no_long_calls} directive and functions whose
6874 definitions have already been compiled within the current compilation
6875 unit, will not be turned into long calls. The exception to this rule is
6876 that weak function definitions, functions with the @samp{long-call}
6877 attribute or the @samp{section} attribute, and functions that are within
6878 the scope of a @samp{#pragma long_calls} directive, will always be
6879 turned into long calls.
6881 This feature is not enabled by default. Specifying
6882 @option{-mno-long-calls} will restore the default behavior, as will
6883 placing the function calls within the scope of a @samp{#pragma
6884 long_calls_off} directive. Note these switches have no effect on how
6885 the compiler generates code to handle function calls via function
6888 @item -mnop-fun-dllimport
6889 @opindex mnop-fun-dllimport
6890 Disable support for the @code{dllimport} attribute.
6892 @item -msingle-pic-base
6893 @opindex msingle-pic-base
6894 Treat the register used for PIC addressing as read-only, rather than
6895 loading it in the prologue for each function. The run-time system is
6896 responsible for initializing this register with an appropriate value
6897 before execution begins.
6899 @item -mpic-register=@var{reg}
6900 @opindex mpic-register
6901 Specify the register to be used for PIC addressing. The default is R10
6902 unless stack-checking is enabled, when R9 is used.
6904 @item -mcirrus-fix-invalid-insns
6905 @opindex mcirrus-fix-invalid-insns
6906 @opindex mno-cirrus-fix-invalid-insns
6907 Insert NOPs into the instruction stream to in order to work around
6908 problems with invalid Maverick instruction combinations. This option
6909 is only valid if the @option{-mcpu=ep9312} option has been used to
6910 enable generation of instructions for the Cirrus Maverick floating
6911 point co-processor. This option is not enabled by default, since the
6912 problem is only present in older Maverick implementations. The default
6913 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
6916 @item -mpoke-function-name
6917 @opindex mpoke-function-name
6918 Write the name of each function into the text section, directly
6919 preceding the function prologue. The generated code is similar to this:
6923 .ascii "arm_poke_function_name", 0
6926 .word 0xff000000 + (t1 - t0)
6927 arm_poke_function_name
6929 stmfd sp!, @{fp, ip, lr, pc@}
6933 When performing a stack backtrace, code can inspect the value of
6934 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
6935 location @code{pc - 12} and the top 8 bits are set, then we know that
6936 there is a function name embedded immediately preceding this location
6937 and has length @code{((pc[-3]) & 0xff000000)}.
6941 Generate code for the 16-bit Thumb instruction set. The default is to
6942 use the 32-bit ARM instruction set.
6945 @opindex mtpcs-frame
6946 Generate a stack frame that is compliant with the Thumb Procedure Call
6947 Standard for all non-leaf functions. (A leaf function is one that does
6948 not call any other functions.) The default is @option{-mno-tpcs-frame}.
6950 @item -mtpcs-leaf-frame
6951 @opindex mtpcs-leaf-frame
6952 Generate a stack frame that is compliant with the Thumb Procedure Call
6953 Standard for all leaf functions. (A leaf function is one that does
6954 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
6956 @item -mcallee-super-interworking
6957 @opindex mcallee-super-interworking
6958 Gives all externally visible functions in the file being compiled an ARM
6959 instruction set header which switches to Thumb mode before executing the
6960 rest of the function. This allows these functions to be called from
6961 non-interworking code.
6963 @item -mcaller-super-interworking
6964 @opindex mcaller-super-interworking
6965 Allows calls via function pointers (including virtual functions) to
6966 execute correctly regardless of whether the target code has been
6967 compiled for interworking or not. There is a small overhead in the cost
6968 of executing a function pointer if this option is enabled.
6973 @subsection AVR Options
6976 These options are defined for AVR implementations:
6979 @item -mmcu=@var{mcu}
6981 Specify ATMEL AVR instruction set or MCU type.
6983 Instruction set avr1 is for the minimal AVR core, not supported by the C
6984 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
6985 attiny11, attiny12, attiny15, attiny28).
6987 Instruction set avr2 (default) is for the classic AVR core with up to
6988 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
6989 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
6990 at90c8534, at90s8535).
6992 Instruction set avr3 is for the classic AVR core with up to 128K program
6993 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
6995 Instruction set avr4 is for the enhanced AVR core with up to 8K program
6996 memory space (MCU types: atmega8, atmega83, atmega85).
6998 Instruction set avr5 is for the enhanced AVR core with up to 128K program
6999 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7000 atmega64, atmega128, at43usb355, at94k).
7004 Output instruction sizes to the asm file.
7006 @item -minit-stack=@var{N}
7007 @opindex minit-stack
7008 Specify the initial stack address, which may be a symbol or numeric value,
7009 @samp{__stack} is the default.
7011 @item -mno-interrupts
7012 @opindex mno-interrupts
7013 Generated code is not compatible with hardware interrupts.
7014 Code size will be smaller.
7016 @item -mcall-prologues
7017 @opindex mcall-prologues
7018 Functions prologues/epilogues expanded as call to appropriate
7019 subroutines. Code size will be smaller.
7021 @item -mno-tablejump
7022 @opindex mno-tablejump
7023 Do not generate tablejump insns which sometimes increase code size.
7026 @opindex mtiny-stack
7027 Change only the low 8 bits of the stack pointer.
7031 Assume int to be 8 bit integer. This affects the sizes of all types: A
7032 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7033 and long long will be 4 bytes. Please note that this option does not
7034 comply to the C standards, but it will provide you with smaller code
7039 @subsection CRIS Options
7040 @cindex CRIS Options
7042 These options are defined specifically for the CRIS ports.
7045 @item -march=@var{architecture-type}
7046 @itemx -mcpu=@var{architecture-type}
7049 Generate code for the specified architecture. The choices for
7050 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7051 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7052 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7055 @item -mtune=@var{architecture-type}
7057 Tune to @var{architecture-type} everything applicable about the generated
7058 code, except for the ABI and the set of available instructions. The
7059 choices for @var{architecture-type} are the same as for
7060 @option{-march=@var{architecture-type}}.
7062 @item -mmax-stack-frame=@var{n}
7063 @opindex mmax-stack-frame
7064 Warn when the stack frame of a function exceeds @var{n} bytes.
7066 @item -melinux-stacksize=@var{n}
7067 @opindex melinux-stacksize
7068 Only available with the @samp{cris-axis-aout} target. Arranges for
7069 indications in the program to the kernel loader that the stack of the
7070 program should be set to @var{n} bytes.
7076 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7077 @option{-march=v3} and @option{-march=v8} respectively.
7079 @item -mmul-bug-workaround
7080 @itemx -mno-mul-bug-workaround
7081 @opindex mmul-bug-workaround
7082 @opindex mno-mul-bug-workaround
7083 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7084 models where it applies. This option is active by default.
7088 Enable CRIS-specific verbose debug-related information in the assembly
7089 code. This option also has the effect to turn off the @samp{#NO_APP}
7090 formatted-code indicator to the assembler at the beginning of the
7095 Do not use condition-code results from previous instruction; always emit
7096 compare and test instructions before use of condition codes.
7098 @item -mno-side-effects
7099 @opindex mno-side-effects
7100 Do not emit instructions with side-effects in addressing modes other than
7104 @itemx -mno-stack-align
7106 @itemx -mno-data-align
7107 @itemx -mconst-align
7108 @itemx -mno-const-align
7109 @opindex mstack-align
7110 @opindex mno-stack-align
7111 @opindex mdata-align
7112 @opindex mno-data-align
7113 @opindex mconst-align
7114 @opindex mno-const-align
7115 These options (no-options) arranges (eliminate arrangements) for the
7116 stack-frame, individual data and constants to be aligned for the maximum
7117 single data access size for the chosen CPU model. The default is to
7118 arrange for 32-bit alignment. ABI details such as structure layout are
7119 not affected by these options.
7127 Similar to the stack- data- and const-align options above, these options
7128 arrange for stack-frame, writable data and constants to all be 32-bit,
7129 16-bit or 8-bit aligned. The default is 32-bit alignment.
7131 @item -mno-prologue-epilogue
7132 @itemx -mprologue-epilogue
7133 @opindex mno-prologue-epilogue
7134 @opindex mprologue-epilogue
7135 With @option{-mno-prologue-epilogue}, the normal function prologue and
7136 epilogue that sets up the stack-frame are omitted and no return
7137 instructions or return sequences are generated in the code. Use this
7138 option only together with visual inspection of the compiled code: no
7139 warnings or errors are generated when call-saved registers must be saved,
7140 or storage for local variable needs to be allocated.
7146 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7147 instruction sequences that load addresses for functions from the PLT part
7148 of the GOT rather than (traditional on other architectures) calls to the
7149 PLT@. The default is @option{-mgotplt}.
7153 Legacy no-op option only recognized with the cris-axis-aout target.
7157 Legacy no-op option only recognized with the cris-axis-elf and
7158 cris-axis-linux-gnu targets.
7162 Only recognized with the cris-axis-aout target, where it selects a
7163 GNU/linux-like multilib, include files and instruction set for
7168 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7172 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7173 to link with input-output functions from a simulator library. Code,
7174 initialized data and zero-initialized data are allocated consecutively.
7178 Like @option{-sim}, but pass linker options to locate initialized data at
7179 0x40000000 and zero-initialized data at 0x80000000.
7182 @node Darwin Options
7183 @subsection Darwin Options
7184 @cindex Darwin options
7186 These options are defined for all architectures running the Darwin operating
7189 FSF GCC on Darwin does not create ``fat'' object files; it will create
7190 an object file for the single architecture that it was built to
7191 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7192 @option{-arch} options are used; it does so by running the compiler or
7193 linker multiple times and joining the results together with
7196 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7197 @samp{i686}) is determined by the flags that specify the ISA
7198 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7199 @option{-force_cpusubtype_ALL} option can be used to override this.
7201 The Darwin tools vary in their behaviour when presented with an ISA
7202 mismatch. The assembler, @file{as}, will only permit instructions to
7203 be used that are valid for the subtype of the file it is generating,
7204 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7205 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7206 and print an error if asked to create a shared library with a less
7207 restrictive subtype than its input files (for instance, trying to put
7208 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7209 for executables, @file{ld}, will quietly give the executable the most
7210 restrictive subtype of any of its input files.
7215 Add the framework directory @var{dir} to the head of the list of
7216 directories to be searched for header files. These directories are
7217 interleaved with those specified by @option{-I} options and are
7218 scanned in a left-to-right order.
7220 A framework directory is a directory with frameworks in it. A
7221 framework is a directory with a @samp{"Headers"} and/or
7222 @samp{"PrivateHeaders"} directory contained directly in it that ends
7223 in @samp{".framework"}. The name of a framework is the name of this
7224 directory excluding the @samp{".framework"}. Headers associated with
7225 the framework are found in one of those two directories, with
7226 @samp{"Headers"} being searched first. A subframework is a framework
7227 directory that is in a framework's @samp{"Frameworks"} directory.
7228 Includes of subframework headers can only appear in a header of a
7229 framework that contains the subframework, or in a sibling subframework
7230 header. Two subframeworks are siblings if they occur in the same
7231 framework. A subframework should not have the same name as a
7232 framework, a warning will be issued if this is violated. Currently a
7233 subframework cannot have subframeworks, in the future, the mechanism
7234 may be extended to support this. The standard frameworks can be found
7235 in @samp{"/System/Library/Frameworks"} and
7236 @samp{"/Library/Frameworks"}. An example include looks like
7237 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7238 the name of the framework and header.h is found in the
7239 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7243 Emit debugging information for symbols that are used. For STABS
7244 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7245 This is by default ON@.
7249 Emit debugging information for all symbols and types.
7251 @item -mone-byte-bool
7252 @opindex -mone-byte-bool
7253 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7254 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7255 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7256 option has no effect on x86.
7258 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7259 to generate code that is not binary compatible with code generated
7260 without that switch. Using this switch may require recompiling all
7261 other modules in a program, including system libraries. Use this
7262 switch to conform to a non-default data model.
7264 @item -mfix-and-continue
7265 @itemx -ffix-and-continue
7266 @itemx -findirect-data
7267 @opindex mfix-and-continue
7268 @opindex ffix-and-continue
7269 @opindex findirect-data
7270 Generate code suitable for fast turn around development. Needed to
7271 enable gdb to dynamically load @code{.o} files into already running
7272 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7273 are provided for backwards compatibility.
7277 Loads all members of static archive libraries.
7278 See man ld(1) for more information.
7280 @item -arch_errors_fatal
7281 @opindex arch_errors_fatal
7282 Cause the errors having to do with files that have the wrong architecture
7286 @opindex bind_at_load
7287 Causes the output file to be marked such that the dynamic linker will
7288 bind all undefined references when the file is loaded or launched.
7292 Produce a Mach-o bundle format file.
7293 See man ld(1) for more information.
7295 @item -bundle_loader @var{executable}
7296 @opindex bundle_loader
7297 This option specifies the @var{executable} that will be loading the build
7298 output file being linked. See man ld(1) for more information.
7301 @opindex -dynamiclib
7302 When passed this option, GCC will produce a dynamic library instead of
7303 an executable when linking, using the Darwin @file{libtool} command.
7305 @item -force_cpusubtype_ALL
7306 @opindex -force_cpusubtype_ALL
7307 This causes GCC's output file to have the @var{ALL} subtype, instead of
7308 one controlled by the @option{-mcpu} or @option{-march} option.
7310 @item -allowable_client @var{client_name}
7312 @itemx -compatibility_version
7313 @itemx -current_version
7315 @itemx -dependency-file
7317 @itemx -dylinker_install_name
7319 @itemx -exported_symbols_list
7321 @itemx -flat_namespace
7322 @itemx -force_flat_namespace
7323 @itemx -headerpad_max_install_names
7326 @itemx -install_name
7327 @itemx -keep_private_externs
7328 @itemx -multi_module
7329 @itemx -multiply_defined
7330 @itemx -multiply_defined_unused
7332 @itemx -no_dead_strip_inits_and_terms
7333 @itemx -nofixprebinding
7336 @itemx -noseglinkedit
7337 @itemx -pagezero_size
7339 @itemx -prebind_all_twolevel_modules
7340 @itemx -private_bundle
7341 @itemx -read_only_relocs
7343 @itemx -sectobjectsymbols
7347 @itemx -sectobjectsymbols
7350 @itemx -segs_read_only_addr
7351 @itemx -segs_read_write_addr
7352 @itemx -seg_addr_table
7353 @itemx -seg_addr_table_filename
7356 @itemx -segs_read_only_addr
7357 @itemx -segs_read_write_addr
7358 @itemx -single_module
7361 @itemx -sub_umbrella
7362 @itemx -twolevel_namespace
7365 @itemx -unexported_symbols_list
7366 @itemx -weak_reference_mismatches
7369 @opindex allowable_client
7370 @opindex client_name
7371 @opindex compatibility_version
7372 @opindex current_version
7374 @opindex dependency-file
7376 @opindex dylinker_install_name
7378 @opindex exported_symbols_list
7380 @opindex flat_namespace
7381 @opindex force_flat_namespace
7382 @opindex headerpad_max_install_names
7385 @opindex install_name
7386 @opindex keep_private_externs
7387 @opindex multi_module
7388 @opindex multiply_defined
7389 @opindex multiply_defined_unused
7391 @opindex no_dead_strip_inits_and_terms
7392 @opindex nofixprebinding
7393 @opindex nomultidefs
7395 @opindex noseglinkedit
7396 @opindex pagezero_size
7398 @opindex prebind_all_twolevel_modules
7399 @opindex private_bundle
7400 @opindex read_only_relocs
7402 @opindex sectobjectsymbols
7406 @opindex sectobjectsymbols
7409 @opindex segs_read_only_addr
7410 @opindex segs_read_write_addr
7411 @opindex seg_addr_table
7412 @opindex seg_addr_table_filename
7413 @opindex seglinkedit
7415 @opindex segs_read_only_addr
7416 @opindex segs_read_write_addr
7417 @opindex single_module
7419 @opindex sub_library
7420 @opindex sub_umbrella
7421 @opindex twolevel_namespace
7424 @opindex unexported_symbols_list
7425 @opindex weak_reference_mismatches
7426 @opindex whatsloaded
7428 These options are passed to the Darwin linker. The Darwin linker man page
7429 describes them in detail.
7432 @node DEC Alpha Options
7433 @subsection DEC Alpha Options
7435 These @samp{-m} options are defined for the DEC Alpha implementations:
7438 @item -mno-soft-float
7440 @opindex mno-soft-float
7441 @opindex msoft-float
7442 Use (do not use) the hardware floating-point instructions for
7443 floating-point operations. When @option{-msoft-float} is specified,
7444 functions in @file{libgcc.a} will be used to perform floating-point
7445 operations. Unless they are replaced by routines that emulate the
7446 floating-point operations, or compiled in such a way as to call such
7447 emulations routines, these routines will issue floating-point
7448 operations. If you are compiling for an Alpha without floating-point
7449 operations, you must ensure that the library is built so as not to call
7452 Note that Alpha implementations without floating-point operations are
7453 required to have floating-point registers.
7458 @opindex mno-fp-regs
7459 Generate code that uses (does not use) the floating-point register set.
7460 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7461 register set is not used, floating point operands are passed in integer
7462 registers as if they were integers and floating-point results are passed
7463 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7464 so any function with a floating-point argument or return value called by code
7465 compiled with @option{-mno-fp-regs} must also be compiled with that
7468 A typical use of this option is building a kernel that does not use,
7469 and hence need not save and restore, any floating-point registers.
7473 The Alpha architecture implements floating-point hardware optimized for
7474 maximum performance. It is mostly compliant with the IEEE floating
7475 point standard. However, for full compliance, software assistance is
7476 required. This option generates code fully IEEE compliant code
7477 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7478 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7479 defined during compilation. The resulting code is less efficient but is
7480 able to correctly support denormalized numbers and exceptional IEEE
7481 values such as not-a-number and plus/minus infinity. Other Alpha
7482 compilers call this option @option{-ieee_with_no_inexact}.
7484 @item -mieee-with-inexact
7485 @opindex mieee-with-inexact
7486 This is like @option{-mieee} except the generated code also maintains
7487 the IEEE @var{inexact-flag}. Turning on this option causes the
7488 generated code to implement fully-compliant IEEE math. In addition to
7489 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7490 macro. On some Alpha implementations the resulting code may execute
7491 significantly slower than the code generated by default. Since there is
7492 very little code that depends on the @var{inexact-flag}, you should
7493 normally not specify this option. Other Alpha compilers call this
7494 option @option{-ieee_with_inexact}.
7496 @item -mfp-trap-mode=@var{trap-mode}
7497 @opindex mfp-trap-mode
7498 This option controls what floating-point related traps are enabled.
7499 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7500 The trap mode can be set to one of four values:
7504 This is the default (normal) setting. The only traps that are enabled
7505 are the ones that cannot be disabled in software (e.g., division by zero
7509 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7513 Like @samp{su}, but the instructions are marked to be safe for software
7514 completion (see Alpha architecture manual for details).
7517 Like @samp{su}, but inexact traps are enabled as well.
7520 @item -mfp-rounding-mode=@var{rounding-mode}
7521 @opindex mfp-rounding-mode
7522 Selects the IEEE rounding mode. Other Alpha compilers call this option
7523 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7528 Normal IEEE rounding mode. Floating point numbers are rounded towards
7529 the nearest machine number or towards the even machine number in case
7533 Round towards minus infinity.
7536 Chopped rounding mode. Floating point numbers are rounded towards zero.
7539 Dynamic rounding mode. A field in the floating point control register
7540 (@var{fpcr}, see Alpha architecture reference manual) controls the
7541 rounding mode in effect. The C library initializes this register for
7542 rounding towards plus infinity. Thus, unless your program modifies the
7543 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7546 @item -mtrap-precision=@var{trap-precision}
7547 @opindex mtrap-precision
7548 In the Alpha architecture, floating point traps are imprecise. This
7549 means without software assistance it is impossible to recover from a
7550 floating trap and program execution normally needs to be terminated.
7551 GCC can generate code that can assist operating system trap handlers
7552 in determining the exact location that caused a floating point trap.
7553 Depending on the requirements of an application, different levels of
7554 precisions can be selected:
7558 Program precision. This option is the default and means a trap handler
7559 can only identify which program caused a floating point exception.
7562 Function precision. The trap handler can determine the function that
7563 caused a floating point exception.
7566 Instruction precision. The trap handler can determine the exact
7567 instruction that caused a floating point exception.
7570 Other Alpha compilers provide the equivalent options called
7571 @option{-scope_safe} and @option{-resumption_safe}.
7573 @item -mieee-conformant
7574 @opindex mieee-conformant
7575 This option marks the generated code as IEEE conformant. You must not
7576 use this option unless you also specify @option{-mtrap-precision=i} and either
7577 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7578 is to emit the line @samp{.eflag 48} in the function prologue of the
7579 generated assembly file. Under DEC Unix, this has the effect that
7580 IEEE-conformant math library routines will be linked in.
7582 @item -mbuild-constants
7583 @opindex mbuild-constants
7584 Normally GCC examines a 32- or 64-bit integer constant to
7585 see if it can construct it from smaller constants in two or three
7586 instructions. If it cannot, it will output the constant as a literal and
7587 generate code to load it from the data segment at runtime.
7589 Use this option to require GCC to construct @emph{all} integer constants
7590 using code, even if it takes more instructions (the maximum is six).
7592 You would typically use this option to build a shared library dynamic
7593 loader. Itself a shared library, it must relocate itself in memory
7594 before it can find the variables and constants in its own data segment.
7600 Select whether to generate code to be assembled by the vendor-supplied
7601 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7619 Indicate whether GCC should generate code to use the optional BWX,
7620 CIX, FIX and MAX instruction sets. The default is to use the instruction
7621 sets supported by the CPU type specified via @option{-mcpu=} option or that
7622 of the CPU on which GCC was built if none was specified.
7627 @opindex mfloat-ieee
7628 Generate code that uses (does not use) VAX F and G floating point
7629 arithmetic instead of IEEE single and double precision.
7631 @item -mexplicit-relocs
7632 @itemx -mno-explicit-relocs
7633 @opindex mexplicit-relocs
7634 @opindex mno-explicit-relocs
7635 Older Alpha assemblers provided no way to generate symbol relocations
7636 except via assembler macros. Use of these macros does not allow
7637 optimal instruction scheduling. GNU binutils as of version 2.12
7638 supports a new syntax that allows the compiler to explicitly mark
7639 which relocations should apply to which instructions. This option
7640 is mostly useful for debugging, as GCC detects the capabilities of
7641 the assembler when it is built and sets the default accordingly.
7645 @opindex msmall-data
7646 @opindex mlarge-data
7647 When @option{-mexplicit-relocs} is in effect, static data is
7648 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7649 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7650 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7651 16-bit relocations off of the @code{$gp} register. This limits the
7652 size of the small data area to 64KB, but allows the variables to be
7653 directly accessed via a single instruction.
7655 The default is @option{-mlarge-data}. With this option the data area
7656 is limited to just below 2GB@. Programs that require more than 2GB of
7657 data must use @code{malloc} or @code{mmap} to allocate the data in the
7658 heap instead of in the program's data segment.
7660 When generating code for shared libraries, @option{-fpic} implies
7661 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7665 @opindex msmall-text
7666 @opindex mlarge-text
7667 When @option{-msmall-text} is used, the compiler assumes that the
7668 code of the entire program (or shared library) fits in 4MB, and is
7669 thus reachable with a branch instruction. When @option{-msmall-data}
7670 is used, the compiler can assume that all local symbols share the
7671 same @code{$gp} value, and thus reduce the number of instructions
7672 required for a function call from 4 to 1.
7674 The default is @option{-mlarge-text}.
7676 @item -mcpu=@var{cpu_type}
7678 Set the instruction set and instruction scheduling parameters for
7679 machine type @var{cpu_type}. You can specify either the @samp{EV}
7680 style name or the corresponding chip number. GCC supports scheduling
7681 parameters for the EV4, EV5 and EV6 family of processors and will
7682 choose the default values for the instruction set from the processor
7683 you specify. If you do not specify a processor type, GCC will default
7684 to the processor on which the compiler was built.
7686 Supported values for @var{cpu_type} are
7692 Schedules as an EV4 and has no instruction set extensions.
7696 Schedules as an EV5 and has no instruction set extensions.
7700 Schedules as an EV5 and supports the BWX extension.
7705 Schedules as an EV5 and supports the BWX and MAX extensions.
7709 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7713 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7716 @item -mtune=@var{cpu_type}
7718 Set only the instruction scheduling parameters for machine type
7719 @var{cpu_type}. The instruction set is not changed.
7721 @item -mmemory-latency=@var{time}
7722 @opindex mmemory-latency
7723 Sets the latency the scheduler should assume for typical memory
7724 references as seen by the application. This number is highly
7725 dependent on the memory access patterns used by the application
7726 and the size of the external cache on the machine.
7728 Valid options for @var{time} are
7732 A decimal number representing clock cycles.
7738 The compiler contains estimates of the number of clock cycles for
7739 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7740 (also called Dcache, Scache, and Bcache), as well as to main memory.
7741 Note that L3 is only valid for EV5.
7746 @node DEC Alpha/VMS Options
7747 @subsection DEC Alpha/VMS Options
7749 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7752 @item -mvms-return-codes
7753 @opindex mvms-return-codes
7754 Return VMS condition codes from main. The default is to return POSIX
7755 style condition (e.g.@ error) codes.
7759 @subsection FRV Options
7766 Only use the first 32 general purpose registers.
7771 Use all 64 general purpose registers.
7776 Use only the first 32 floating point registers.
7781 Use all 64 floating point registers
7784 @opindex mhard-float
7786 Use hardware instructions for floating point operations.
7789 @opindex msoft-float
7791 Use library routines for floating point operations.
7796 Dynamically allocate condition code registers.
7801 Do not try to dynamically allocate condition code registers, only
7802 use @code{icc0} and @code{fcc0}.
7807 Change ABI to use double word insns.
7812 Do not use double word instructions.
7817 Use floating point double instructions.
7822 Do not use floating point double instructions.
7827 Use media instructions.
7832 Do not use media instructions.
7837 Use multiply and add/subtract instructions.
7842 Do not use multiply and add/subtract instructions.
7847 Select the FDPIC ABI, that uses function descriptors to represent
7848 pointers to functions. Without any PIC/PIE-related options, it
7849 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
7850 assumes GOT entries and small data are within a 12-bit range from the
7851 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
7852 are computed with 32 bits.
7855 @opindex minline-plt
7857 Enable inlining of PLT entries in function calls to functions that are
7858 not known to bind locally. It has no effect without @option{-mfdpic}.
7859 It's enabled by default if optimizing for speed and compiling for
7860 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
7861 optimization option such as @option{-O3} or above is present in the
7867 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
7868 that is known to be in read-only sections. It's enabled by default,
7869 except for @option{-fpic} or @option{-fpie}: even though it may help
7870 make the global offset table smaller, it trades 1 instruction for 4.
7871 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
7872 one of which may be shared by multiple symbols, and it avoids the need
7873 for a GOT entry for the referenced symbol, so it's more likely to be a
7874 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
7876 @item -multilib-library-pic
7877 @opindex multilib-library-pic
7879 Link with the (library, not FD) pic libraries. It's implied by
7880 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
7881 @option{-fpic} without @option{-mfdpic}. You should never have to use
7887 Follow the EABI requirement of always creating a frame pointer whenever
7888 a stack frame is allocated. This option is enabled by default and can
7889 be disabled with @option{-mno-linked-fp}.
7892 @opindex mlong-calls
7894 Use indirect addressing to call functions outside the current
7895 compilation unit. This allows the functions to be placed anywhere
7896 within the 32-bit address space.
7898 @item -malign-labels
7899 @opindex malign-labels
7901 Try to align labels to an 8-byte boundary by inserting nops into the
7902 previous packet. This option only has an effect when VLIW packing
7903 is enabled. It doesn't create new packets; it merely adds nops to
7907 @opindex mlibrary-pic
7909 Generate position-independent EABI code.
7914 Use only the first four media accumulator registers.
7919 Use all eight media accumulator registers.
7924 Pack VLIW instructions.
7929 Do not pack VLIW instructions.
7934 Do not mark ABI switches in e_flags.
7939 Enable the use of conditional-move instructions (default).
7941 This switch is mainly for debugging the compiler and will likely be removed
7942 in a future version.
7944 @item -mno-cond-move
7945 @opindex mno-cond-move
7947 Disable the use of conditional-move instructions.
7949 This switch is mainly for debugging the compiler and will likely be removed
7950 in a future version.
7955 Enable the use of conditional set instructions (default).
7957 This switch is mainly for debugging the compiler and will likely be removed
7958 in a future version.
7963 Disable the use of conditional set instructions.
7965 This switch is mainly for debugging the compiler and will likely be removed
7966 in a future version.
7971 Enable the use of conditional execution (default).
7973 This switch is mainly for debugging the compiler and will likely be removed
7974 in a future version.
7976 @item -mno-cond-exec
7977 @opindex mno-cond-exec
7979 Disable the use of conditional execution.
7981 This switch is mainly for debugging the compiler and will likely be removed
7982 in a future version.
7985 @opindex mvliw-branch
7987 Run a pass to pack branches into VLIW instructions (default).
7989 This switch is mainly for debugging the compiler and will likely be removed
7990 in a future version.
7992 @item -mno-vliw-branch
7993 @opindex mno-vliw-branch
7995 Do not run a pass to pack branches into VLIW instructions.
7997 This switch is mainly for debugging the compiler and will likely be removed
7998 in a future version.
8000 @item -mmulti-cond-exec
8001 @opindex mmulti-cond-exec
8003 Enable optimization of @code{&&} and @code{||} in conditional execution
8006 This switch is mainly for debugging the compiler and will likely be removed
8007 in a future version.
8009 @item -mno-multi-cond-exec
8010 @opindex mno-multi-cond-exec
8012 Disable optimization of @code{&&} and @code{||} in conditional execution.
8014 This switch is mainly for debugging the compiler and will likely be removed
8015 in a future version.
8017 @item -mnested-cond-exec
8018 @opindex mnested-cond-exec
8020 Enable nested conditional execution optimizations (default).
8022 This switch is mainly for debugging the compiler and will likely be removed
8023 in a future version.
8025 @item -mno-nested-cond-exec
8026 @opindex mno-nested-cond-exec
8028 Disable nested conditional execution optimizations.
8030 This switch is mainly for debugging the compiler and will likely be removed
8031 in a future version.
8033 @item -mtomcat-stats
8034 @opindex mtomcat-stats
8036 Cause gas to print out tomcat statistics.
8038 @item -mcpu=@var{cpu}
8041 Select the processor type for which to generate code. Possible values are
8042 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8043 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8047 @node H8/300 Options
8048 @subsection H8/300 Options
8050 These @samp{-m} options are defined for the H8/300 implementations:
8055 Shorten some address references at link time, when possible; uses the
8056 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8057 ld, Using ld}, for a fuller description.
8061 Generate code for the H8/300H@.
8065 Generate code for the H8S@.
8069 Generate code for the H8S and H8/300H in the normal mode. This switch
8070 must be used either with @option{-mh} or @option{-ms}.
8074 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8078 Make @code{int} data 32 bits by default.
8082 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8083 The default for the H8/300H and H8S is to align longs and floats on 4
8085 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8086 This option has no effect on the H8/300.
8090 @subsection HPPA Options
8091 @cindex HPPA Options
8093 These @samp{-m} options are defined for the HPPA family of computers:
8096 @item -march=@var{architecture-type}
8098 Generate code for the specified architecture. The choices for
8099 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8100 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8101 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8102 architecture option for your machine. Code compiled for lower numbered
8103 architectures will run on higher numbered architectures, but not the
8107 @itemx -mpa-risc-1-1
8108 @itemx -mpa-risc-2-0
8109 @opindex mpa-risc-1-0
8110 @opindex mpa-risc-1-1
8111 @opindex mpa-risc-2-0
8112 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8115 @opindex mbig-switch
8116 Generate code suitable for big switch tables. Use this option only if
8117 the assembler/linker complain about out of range branches within a switch
8120 @item -mjump-in-delay
8121 @opindex mjump-in-delay
8122 Fill delay slots of function calls with unconditional jump instructions
8123 by modifying the return pointer for the function call to be the target
8124 of the conditional jump.
8126 @item -mdisable-fpregs
8127 @opindex mdisable-fpregs
8128 Prevent floating point registers from being used in any manner. This is
8129 necessary for compiling kernels which perform lazy context switching of
8130 floating point registers. If you use this option and attempt to perform
8131 floating point operations, the compiler will abort.
8133 @item -mdisable-indexing
8134 @opindex mdisable-indexing
8135 Prevent the compiler from using indexing address modes. This avoids some
8136 rather obscure problems when compiling MIG generated code under MACH@.
8138 @item -mno-space-regs
8139 @opindex mno-space-regs
8140 Generate code that assumes the target has no space registers. This allows
8141 GCC to generate faster indirect calls and use unscaled index address modes.
8143 Such code is suitable for level 0 PA systems and kernels.
8145 @item -mfast-indirect-calls
8146 @opindex mfast-indirect-calls
8147 Generate code that assumes calls never cross space boundaries. This
8148 allows GCC to emit code which performs faster indirect calls.
8150 This option will not work in the presence of shared libraries or nested
8153 @item -mfixed-range=@var{register-range}
8154 @opindex mfixed-range
8155 Generate code treating the given register range as fixed registers.
8156 A fixed register is one that the register allocator can not use. This is
8157 useful when compiling kernel code. A register range is specified as
8158 two registers separated by a dash. Multiple register ranges can be
8159 specified separated by a comma.
8161 @item -mlong-load-store
8162 @opindex mlong-load-store
8163 Generate 3-instruction load and store sequences as sometimes required by
8164 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8167 @item -mportable-runtime
8168 @opindex mportable-runtime
8169 Use the portable calling conventions proposed by HP for ELF systems.
8173 Enable the use of assembler directives only GAS understands.
8175 @item -mschedule=@var{cpu-type}
8177 Schedule code according to the constraints for the machine type
8178 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8179 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8180 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8181 proper scheduling option for your machine. The default scheduling is
8185 @opindex mlinker-opt
8186 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8187 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8188 linkers in which they give bogus error messages when linking some programs.
8191 @opindex msoft-float
8192 Generate output containing library calls for floating point.
8193 @strong{Warning:} the requisite libraries are not available for all HPPA
8194 targets. Normally the facilities of the machine's usual C compiler are
8195 used, but this cannot be done directly in cross-compilation. You must make
8196 your own arrangements to provide suitable library functions for
8197 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8198 does provide software floating point support.
8200 @option{-msoft-float} changes the calling convention in the output file;
8201 therefore, it is only useful if you compile @emph{all} of a program with
8202 this option. In particular, you need to compile @file{libgcc.a}, the
8203 library that comes with GCC, with @option{-msoft-float} in order for
8208 Generate the predefine, @code{_SIO}, for server IO@. The default is
8209 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8210 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8211 options are available under HP-UX and HI-UX@.
8215 Use GNU ld specific options. This passes @option{-shared} to ld when
8216 building a shared library. It is the default when GCC is configured,
8217 explicitly or implicitly, with the GNU linker. This option does not
8218 have any affect on which ld is called, it only changes what parameters
8219 are passed to that ld. The ld that is called is determined by the
8220 @option{--with-ld} configure option, GCC's program search path, and
8221 finally by the user's @env{PATH}. The linker used by GCC can be printed
8222 using @samp{which `gcc -print-prog-name=ld`}.
8226 Use HP ld specific options. This passes @option{-b} to ld when building
8227 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8228 links. It is the default when GCC is configured, explicitly or
8229 implicitly, with the HP linker. This option does not have any affect on
8230 which ld is called, it only changes what parameters are passed to that
8231 ld. The ld that is called is determined by the @option{--with-ld}
8232 configure option, GCC's program search path, and finally by the user's
8233 @env{PATH}. The linker used by GCC can be printed using @samp{which
8234 `gcc -print-prog-name=ld`}.
8237 @opindex mno-long-calls
8238 Generate code that uses long call sequences. This ensures that a call
8239 is always able to reach linker generated stubs. The default is to generate
8240 long calls only when the distance from the call site to the beginning
8241 of the function or translation unit, as the case may be, exceeds a
8242 predefined limit set by the branch type being used. The limits for
8243 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8244 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8247 Distances are measured from the beginning of functions when using the
8248 @option{-ffunction-sections} option, or when using the @option{-mgas}
8249 and @option{-mno-portable-runtime} options together under HP-UX with
8252 It is normally not desirable to use this option as it will degrade
8253 performance. However, it may be useful in large applications,
8254 particularly when partial linking is used to build the application.
8256 The types of long calls used depends on the capabilities of the
8257 assembler and linker, and the type of code being generated. The
8258 impact on systems that support long absolute calls, and long pic
8259 symbol-difference or pc-relative calls should be relatively small.
8260 However, an indirect call is used on 32-bit ELF systems in pic code
8261 and it is quite long.
8263 @item -munix=@var{unix-std}
8265 Generate compiler predefines and select a startfile for the specified
8266 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8267 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8268 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8269 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8270 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8273 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8274 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8275 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8276 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8277 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8278 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8280 It is @emph{important} to note that this option changes the interfaces
8281 for various library routines. It also affects the operational behavior
8282 of the C library. Thus, @emph{extreme} care is needed in using this
8285 Library code that is intended to operate with more than one UNIX
8286 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8287 as appropriate. Most GNU software doesn't provide this capability.
8291 Suppress the generation of link options to search libdld.sl when the
8292 @option{-static} option is specified on HP-UX 10 and later.
8296 The HP-UX implementation of setlocale in libc has a dependency on
8297 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8298 when the @option{-static} option is specified, special link options
8299 are needed to resolve this dependency.
8301 On HP-UX 10 and later, the GCC driver adds the necessary options to
8302 link with libdld.sl when the @option{-static} option is specified.
8303 This causes the resulting binary to be dynamic. On the 64-bit port,
8304 the linkers generate dynamic binaries by default in any case. The
8305 @option{-nolibdld} option can be used to prevent the GCC driver from
8306 adding these link options.
8310 Add support for multithreading with the @dfn{dce thread} library
8311 under HP-UX@. This option sets flags for both the preprocessor and
8315 @node i386 and x86-64 Options
8316 @subsection Intel 386 and AMD x86-64 Options
8317 @cindex i386 Options
8318 @cindex x86-64 Options
8319 @cindex Intel 386 Options
8320 @cindex AMD x86-64 Options
8322 These @samp{-m} options are defined for the i386 and x86-64 family of
8326 @item -mtune=@var{cpu-type}
8328 Tune to @var{cpu-type} everything applicable about the generated code, except
8329 for the ABI and the set of available instructions. The choices for
8333 Original Intel's i386 CPU@.
8335 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8337 Intel Pentium CPU with no MMX support.
8339 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8340 @item i686, pentiumpro
8341 Intel PentiumPro CPU@.
8343 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8344 @item pentium3, pentium3m
8345 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8348 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8349 support. Used by Centrino notebooks.
8350 @item pentium4, pentium4m
8351 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8353 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8356 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8357 SSE2 and SSE3 instruction set support.
8359 AMD K6 CPU with MMX instruction set support.
8361 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8362 @item athlon, athlon-tbird
8363 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8365 @item athlon-4, athlon-xp, athlon-mp
8366 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8367 instruction set support.
8368 @item k8, opteron, athlon64, athlon-fx
8369 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8370 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8372 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8375 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8376 instruction set support.
8378 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8379 implemented for this chip.)
8381 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8382 implemented for this chip.)
8385 While picking a specific @var{cpu-type} will schedule things appropriately
8386 for that particular chip, the compiler will not generate any code that
8387 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8390 @item -march=@var{cpu-type}
8392 Generate instructions for the machine type @var{cpu-type}. The choices
8393 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8394 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8396 @item -mcpu=@var{cpu-type}
8398 A deprecated synonym for @option{-mtune}.
8407 @opindex mpentiumpro
8408 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8409 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8410 These synonyms are deprecated.
8412 @item -mfpmath=@var{unit}
8414 Generate floating point arithmetics for selected unit @var{unit}. The choices
8419 Use the standard 387 floating point coprocessor present majority of chips and
8420 emulated otherwise. Code compiled with this option will run almost everywhere.
8421 The temporary results are computed in 80bit precision instead of precision
8422 specified by the type resulting in slightly different results compared to most
8423 of other chips. See @option{-ffloat-store} for more detailed description.
8425 This is the default choice for i386 compiler.
8428 Use scalar floating point instructions present in the SSE instruction set.
8429 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8430 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8431 instruction set supports only single precision arithmetics, thus the double and
8432 extended precision arithmetics is still done using 387. Later version, present
8433 only in Pentium4 and the future AMD x86-64 chips supports double precision
8436 For i387 you need to use @option{-march=@var{cpu-type}}, @option{-msse} or
8437 @option{-msse2} switches to enable SSE extensions and make this option
8438 effective. For x86-64 compiler, these extensions are enabled by default.
8440 The resulting code should be considerably faster in the majority of cases and avoid
8441 the numerical instability problems of 387 code, but may break some existing
8442 code that expects temporaries to be 80bit.
8444 This is the default choice for the x86-64 compiler.
8447 Attempt to utilize both instruction sets at once. This effectively double the
8448 amount of available registers and on chips with separate execution units for
8449 387 and SSE the execution resources too. Use this option with care, as it is
8450 still experimental, because the GCC register allocator does not model separate
8451 functional units well resulting in instable performance.
8454 @item -masm=@var{dialect}
8455 @opindex masm=@var{dialect}
8456 Output asm instructions using selected @var{dialect}. Supported choices are
8457 @samp{intel} or @samp{att} (the default one).
8462 @opindex mno-ieee-fp
8463 Control whether or not the compiler uses IEEE floating point
8464 comparisons. These handle correctly the case where the result of a
8465 comparison is unordered.
8468 @opindex msoft-float
8469 Generate output containing library calls for floating point.
8470 @strong{Warning:} the requisite libraries are not part of GCC@.
8471 Normally the facilities of the machine's usual C compiler are used, but
8472 this can't be done directly in cross-compilation. You must make your
8473 own arrangements to provide suitable library functions for
8476 On machines where a function returns floating point results in the 80387
8477 register stack, some floating point opcodes may be emitted even if
8478 @option{-msoft-float} is used.
8480 @item -mno-fp-ret-in-387
8481 @opindex mno-fp-ret-in-387
8482 Do not use the FPU registers for return values of functions.
8484 The usual calling convention has functions return values of types
8485 @code{float} and @code{double} in an FPU register, even if there
8486 is no FPU@. The idea is that the operating system should emulate
8489 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8490 in ordinary CPU registers instead.
8492 @item -mno-fancy-math-387
8493 @opindex mno-fancy-math-387
8494 Some 387 emulators do not support the @code{sin}, @code{cos} and
8495 @code{sqrt} instructions for the 387. Specify this option to avoid
8496 generating those instructions. This option is the default on FreeBSD,
8497 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8498 indicates that the target cpu will always have an FPU and so the
8499 instruction will not need emulation. As of revision 2.6.1, these
8500 instructions are not generated unless you also use the
8501 @option{-funsafe-math-optimizations} switch.
8503 @item -malign-double
8504 @itemx -mno-align-double
8505 @opindex malign-double
8506 @opindex mno-align-double
8507 Control whether GCC aligns @code{double}, @code{long double}, and
8508 @code{long long} variables on a two word boundary or a one word
8509 boundary. Aligning @code{double} variables on a two word boundary will
8510 produce code that runs somewhat faster on a @samp{Pentium} at the
8511 expense of more memory.
8513 @strong{Warning:} if you use the @option{-malign-double} switch,
8514 structures containing the above types will be aligned differently than
8515 the published application binary interface specifications for the 386
8516 and will not be binary compatible with structures in code compiled
8517 without that switch.
8519 @item -m96bit-long-double
8520 @itemx -m128bit-long-double
8521 @opindex m96bit-long-double
8522 @opindex m128bit-long-double
8523 These switches control the size of @code{long double} type. The i386
8524 application binary interface specifies the size to be 96 bits,
8525 so @option{-m96bit-long-double} is the default in 32 bit mode.
8527 Modern architectures (Pentium and newer) would prefer @code{long double}
8528 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8529 conforming to the ABI, this would not be possible. So specifying a
8530 @option{-m128bit-long-double} will align @code{long double}
8531 to a 16 byte boundary by padding the @code{long double} with an additional
8534 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8535 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8537 Notice that neither of these options enable any extra precision over the x87
8538 standard of 80 bits for a @code{long double}.
8540 @strong{Warning:} if you override the default value for your target ABI, the
8541 structures and arrays containing @code{long double} variables will change
8542 their size as well as function calling convention for function taking
8543 @code{long double} will be modified. Hence they will not be binary
8544 compatible with arrays or structures in code compiled without that switch.
8548 @itemx -mno-svr3-shlib
8549 @opindex msvr3-shlib
8550 @opindex mno-svr3-shlib
8551 Control whether GCC places uninitialized local variables into the
8552 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8553 into @code{bss}. These options are meaningful only on System V Release 3.
8557 Use a different function-calling convention, in which functions that
8558 take a fixed number of arguments return with the @code{ret} @var{num}
8559 instruction, which pops their arguments while returning. This saves one
8560 instruction in the caller since there is no need to pop the arguments
8563 You can specify that an individual function is called with this calling
8564 sequence with the function attribute @samp{stdcall}. You can also
8565 override the @option{-mrtd} option by using the function attribute
8566 @samp{cdecl}. @xref{Function Attributes}.
8568 @strong{Warning:} this calling convention is incompatible with the one
8569 normally used on Unix, so you cannot use it if you need to call
8570 libraries compiled with the Unix compiler.
8572 Also, you must provide function prototypes for all functions that
8573 take variable numbers of arguments (including @code{printf});
8574 otherwise incorrect code will be generated for calls to those
8577 In addition, seriously incorrect code will result if you call a
8578 function with too many arguments. (Normally, extra arguments are
8579 harmlessly ignored.)
8581 @item -mregparm=@var{num}
8583 Control how many registers are used to pass integer arguments. By
8584 default, no registers are used to pass arguments, and at most 3
8585 registers can be used. You can control this behavior for a specific
8586 function by using the function attribute @samp{regparm}.
8587 @xref{Function Attributes}.
8589 @strong{Warning:} if you use this switch, and
8590 @var{num} is nonzero, then you must build all modules with the same
8591 value, including any libraries. This includes the system libraries and
8594 @item -mpreferred-stack-boundary=@var{num}
8595 @opindex mpreferred-stack-boundary
8596 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8597 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8598 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8599 size (@option{-Os}), in which case the default is the minimum correct
8600 alignment (4 bytes for x86, and 8 bytes for x86-64).
8602 On Pentium and PentiumPro, @code{double} and @code{long double} values
8603 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8604 suffer significant run time performance penalties. On Pentium III, the
8605 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8606 penalties if it is not 16 byte aligned.
8608 To ensure proper alignment of this values on the stack, the stack boundary
8609 must be as aligned as that required by any value stored on the stack.
8610 Further, every function must be generated such that it keeps the stack
8611 aligned. Thus calling a function compiled with a higher preferred
8612 stack boundary from a function compiled with a lower preferred stack
8613 boundary will most likely misalign the stack. It is recommended that
8614 libraries that use callbacks always use the default setting.
8616 This extra alignment does consume extra stack space, and generally
8617 increases code size. Code that is sensitive to stack space usage, such
8618 as embedded systems and operating system kernels, may want to reduce the
8619 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8637 These switches enable or disable the use of built-in functions that allow
8638 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8641 @xref{X86 Built-in Functions}, for details of the functions enabled
8642 and disabled by these switches.
8644 To have SSE/SSE2 instructions generated automatically from floating-point
8645 code, see @option{-mfpmath=sse}.
8648 @itemx -mno-push-args
8650 @opindex mno-push-args
8651 Use PUSH operations to store outgoing parameters. This method is shorter
8652 and usually equally fast as method using SUB/MOV operations and is enabled
8653 by default. In some cases disabling it may improve performance because of
8654 improved scheduling and reduced dependencies.
8656 @item -maccumulate-outgoing-args
8657 @opindex maccumulate-outgoing-args
8658 If enabled, the maximum amount of space required for outgoing arguments will be
8659 computed in the function prologue. This is faster on most modern CPUs
8660 because of reduced dependencies, improved scheduling and reduced stack usage
8661 when preferred stack boundary is not equal to 2. The drawback is a notable
8662 increase in code size. This switch implies @option{-mno-push-args}.
8666 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8667 on thread-safe exception handling must compile and link all code with the
8668 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8669 @option{-D_MT}; when linking, it links in a special thread helper library
8670 @option{-lmingwthrd} which cleans up per thread exception handling data.
8672 @item -mno-align-stringops
8673 @opindex mno-align-stringops
8674 Do not align destination of inlined string operations. This switch reduces
8675 code size and improves performance in case the destination is already aligned,
8676 but GCC doesn't know about it.
8678 @item -minline-all-stringops
8679 @opindex minline-all-stringops
8680 By default GCC inlines string operations only when destination is known to be
8681 aligned at least to 4 byte boundary. This enables more inlining, increase code
8682 size, but may improve performance of code that depends on fast memcpy, strlen
8683 and memset for short lengths.
8685 @item -momit-leaf-frame-pointer
8686 @opindex momit-leaf-frame-pointer
8687 Don't keep the frame pointer in a register for leaf functions. This
8688 avoids the instructions to save, set up and restore frame pointers and
8689 makes an extra register available in leaf functions. The option
8690 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8691 which might make debugging harder.
8693 @item -mtls-direct-seg-refs
8694 @itemx -mno-tls-direct-seg-refs
8695 @opindex mtls-direct-seg-refs
8696 Controls whether TLS variables may be accessed with offsets from the
8697 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8698 or whether the thread base pointer must be added. Whether or not this
8699 is legal depends on the operating system, and whether it maps the
8700 segment to cover the entire TLS area.
8702 For systems that use GNU libc, the default is on.
8705 These @samp{-m} switches are supported in addition to the above
8706 on AMD x86-64 processors in 64-bit environments.
8713 Generate code for a 32-bit or 64-bit environment.
8714 The 32-bit environment sets int, long and pointer to 32 bits and
8715 generates code that runs on any i386 system.
8716 The 64-bit environment sets int to 32 bits and long and pointer
8717 to 64 bits and generates code for AMD's x86-64 architecture.
8720 @opindex no-red-zone
8721 Do not use a so called red zone for x86-64 code. The red zone is mandated
8722 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8723 stack pointer that will not be modified by signal or interrupt handlers
8724 and therefore can be used for temporary data without adjusting the stack
8725 pointer. The flag @option{-mno-red-zone} disables this red zone.
8727 @item -mcmodel=small
8728 @opindex mcmodel=small
8729 Generate code for the small code model: the program and its symbols must
8730 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8731 Programs can be statically or dynamically linked. This is the default
8734 @item -mcmodel=kernel
8735 @opindex mcmodel=kernel
8736 Generate code for the kernel code model. The kernel runs in the
8737 negative 2 GB of the address space.
8738 This model has to be used for Linux kernel code.
8740 @item -mcmodel=medium
8741 @opindex mcmodel=medium
8742 Generate code for the medium model: The program is linked in the lower 2
8743 GB of the address space but symbols can be located anywhere in the
8744 address space. Programs can be statically or dynamically linked, but
8745 building of shared libraries are not supported with the medium model.
8747 @item -mcmodel=large
8748 @opindex mcmodel=large
8749 Generate code for the large model: This model makes no assumptions
8750 about addresses and sizes of sections. Currently GCC does not implement
8755 @subsection IA-64 Options
8756 @cindex IA-64 Options
8758 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8762 @opindex mbig-endian
8763 Generate code for a big endian target. This is the default for HP-UX@.
8765 @item -mlittle-endian
8766 @opindex mlittle-endian
8767 Generate code for a little endian target. This is the default for AIX5
8774 Generate (or don't) code for the GNU assembler. This is the default.
8775 @c Also, this is the default if the configure option @option{--with-gnu-as}
8782 Generate (or don't) code for the GNU linker. This is the default.
8783 @c Also, this is the default if the configure option @option{--with-gnu-ld}
8788 Generate code that does not use a global pointer register. The result
8789 is not position independent code, and violates the IA-64 ABI@.
8791 @item -mvolatile-asm-stop
8792 @itemx -mno-volatile-asm-stop
8793 @opindex mvolatile-asm-stop
8794 @opindex mno-volatile-asm-stop
8795 Generate (or don't) a stop bit immediately before and after volatile asm
8800 Generate code that works around Itanium B step errata.
8802 @item -mregister-names
8803 @itemx -mno-register-names
8804 @opindex mregister-names
8805 @opindex mno-register-names
8806 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
8807 the stacked registers. This may make assembler output more readable.
8813 Disable (or enable) optimizations that use the small data section. This may
8814 be useful for working around optimizer bugs.
8817 @opindex mconstant-gp
8818 Generate code that uses a single constant global pointer value. This is
8819 useful when compiling kernel code.
8823 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
8824 This is useful when compiling firmware code.
8826 @item -minline-float-divide-min-latency
8827 @opindex minline-float-divide-min-latency
8828 Generate code for inline divides of floating point values
8829 using the minimum latency algorithm.
8831 @item -minline-float-divide-max-throughput
8832 @opindex minline-float-divide-max-throughput
8833 Generate code for inline divides of floating point values
8834 using the maximum throughput algorithm.
8836 @item -minline-int-divide-min-latency
8837 @opindex minline-int-divide-min-latency
8838 Generate code for inline divides of integer values
8839 using the minimum latency algorithm.
8841 @item -minline-int-divide-max-throughput
8842 @opindex minline-int-divide-max-throughput
8843 Generate code for inline divides of integer values
8844 using the maximum throughput algorithm.
8846 @item -mno-dwarf2-asm
8848 @opindex mno-dwarf2-asm
8849 @opindex mdwarf2-asm
8850 Don't (or do) generate assembler code for the DWARF2 line number debugging
8851 info. This may be useful when not using the GNU assembler.
8853 @item -mfixed-range=@var{register-range}
8854 @opindex mfixed-range
8855 Generate code treating the given register range as fixed registers.
8856 A fixed register is one that the register allocator can not use. This is
8857 useful when compiling kernel code. A register range is specified as
8858 two registers separated by a dash. Multiple register ranges can be
8859 specified separated by a comma.
8861 @item -mearly-stop-bits
8862 @itemx -mno-early-stop-bits
8863 @opindex mearly-stop-bits
8864 @opindex mno-early-stop-bits
8865 Allow stop bits to be placed earlier than immediately preceding the
8866 instruction that triggered the stop bit. This can improve instruction
8867 scheduling, but does not always do so.
8870 @node M32R/D Options
8871 @subsection M32R/D Options
8872 @cindex M32R/D options
8874 These @option{-m} options are defined for Renesas M32R/D architectures:
8879 Generate code for the M32R/2@.
8883 Generate code for the M32R/X@.
8887 Generate code for the M32R@. This is the default.
8890 @opindex mmodel=small
8891 Assume all objects live in the lower 16MB of memory (so that their addresses
8892 can be loaded with the @code{ld24} instruction), and assume all subroutines
8893 are reachable with the @code{bl} instruction.
8894 This is the default.
8896 The addressability of a particular object can be set with the
8897 @code{model} attribute.
8899 @item -mmodel=medium
8900 @opindex mmodel=medium
8901 Assume objects may be anywhere in the 32-bit address space (the compiler
8902 will generate @code{seth/add3} instructions to load their addresses), and
8903 assume all subroutines are reachable with the @code{bl} instruction.
8906 @opindex mmodel=large
8907 Assume objects may be anywhere in the 32-bit address space (the compiler
8908 will generate @code{seth/add3} instructions to load their addresses), and
8909 assume subroutines may not be reachable with the @code{bl} instruction
8910 (the compiler will generate the much slower @code{seth/add3/jl}
8911 instruction sequence).
8914 @opindex msdata=none
8915 Disable use of the small data area. Variables will be put into
8916 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
8917 @code{section} attribute has been specified).
8918 This is the default.
8920 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
8921 Objects may be explicitly put in the small data area with the
8922 @code{section} attribute using one of these sections.
8925 @opindex msdata=sdata
8926 Put small global and static data in the small data area, but do not
8927 generate special code to reference them.
8931 Put small global and static data in the small data area, and generate
8932 special instructions to reference them.
8936 @cindex smaller data references
8937 Put global and static objects less than or equal to @var{num} bytes
8938 into the small data or bss sections instead of the normal data or bss
8939 sections. The default value of @var{num} is 8.
8940 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
8941 for this option to have any effect.
8943 All modules should be compiled with the same @option{-G @var{num}} value.
8944 Compiling with different values of @var{num} may or may not work; if it
8945 doesn't the linker will give an error message---incorrect code will not be
8950 Makes the M32R specific code in the compiler display some statistics
8951 that might help in debugging programs.
8954 @opindex malign-loops
8955 Align all loops to a 32-byte boundary.
8957 @item -mno-align-loops
8958 @opindex mno-align-loops
8959 Do not enforce a 32-byte alignment for loops. This is the default.
8961 @item -missue-rate=@var{number}
8962 @opindex missue-rate=@var{number}
8963 Issue @var{number} instructions per cycle. @var{number} can only be 1
8966 @item -mbranch-cost=@var{number}
8967 @opindex mbranch-cost=@var{number}
8968 @var{number} can only be 1 or 2. If it is 1 then branches will be
8969 preferred over conditional code, if it is 2, then the opposite will
8972 @item -mflush-trap=@var{number}
8973 @opindex mflush-trap=@var{number}
8974 Specifies the trap number to use to flush the cache. The default is
8975 12. Valid numbers are between 0 and 15 inclusive.
8977 @item -mno-flush-trap
8978 @opindex mno-flush-trap
8979 Specifies that the cache cannot be flushed by using a trap.
8981 @item -mflush-func=@var{name}
8982 @opindex mflush-func=@var{name}
8983 Specifies the name of the operating system function to call to flush
8984 the cache. The default is @emph{_flush_cache}, but a function call
8985 will only be used if a trap is not available.
8987 @item -mno-flush-func
8988 @opindex mno-flush-func
8989 Indicates that there is no OS function for flushing the cache.
8993 @node M680x0 Options
8994 @subsection M680x0 Options
8995 @cindex M680x0 options
8997 These are the @samp{-m} options defined for the 68000 series. The default
8998 values for these options depends on which style of 68000 was selected when
8999 the compiler was configured; the defaults for the most common choices are
9007 Generate output for a 68000. This is the default
9008 when the compiler is configured for 68000-based systems.
9010 Use this option for microcontrollers with a 68000 or EC000 core,
9011 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9017 Generate output for a 68020. This is the default
9018 when the compiler is configured for 68020-based systems.
9022 Generate output containing 68881 instructions for floating point.
9023 This is the default for most 68020 systems unless @option{--nfp} was
9024 specified when the compiler was configured.
9028 Generate output for a 68030. This is the default when the compiler is
9029 configured for 68030-based systems.
9033 Generate output for a 68040. This is the default when the compiler is
9034 configured for 68040-based systems.
9036 This option inhibits the use of 68881/68882 instructions that have to be
9037 emulated by software on the 68040. Use this option if your 68040 does not
9038 have code to emulate those instructions.
9042 Generate output for a 68060. This is the default when the compiler is
9043 configured for 68060-based systems.
9045 This option inhibits the use of 68020 and 68881/68882 instructions that
9046 have to be emulated by software on the 68060. Use this option if your 68060
9047 does not have code to emulate those instructions.
9051 Generate output for a CPU32. This is the default
9052 when the compiler is configured for CPU32-based systems.
9054 Use this option for microcontrollers with a
9055 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9056 68336, 68340, 68341, 68349 and 68360.
9060 Generate output for a 520X ``coldfire'' family cpu. This is the default
9061 when the compiler is configured for 520X-based systems.
9063 Use this option for microcontroller with a 5200 core, including
9064 the MCF5202, MCF5203, MCF5204 and MCF5202.
9069 Generate output for a 68040, without using any of the new instructions.
9070 This results in code which can run relatively efficiently on either a
9071 68020/68881 or a 68030 or a 68040. The generated code does use the
9072 68881 instructions that are emulated on the 68040.
9076 Generate output for a 68060, without using any of the new instructions.
9077 This results in code which can run relatively efficiently on either a
9078 68020/68881 or a 68030 or a 68040. The generated code does use the
9079 68881 instructions that are emulated on the 68060.
9082 @opindex msoft-float
9083 Generate output containing library calls for floating point.
9084 @strong{Warning:} the requisite libraries are not available for all m68k
9085 targets. Normally the facilities of the machine's usual C compiler are
9086 used, but this can't be done directly in cross-compilation. You must
9087 make your own arrangements to provide suitable library functions for
9088 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9089 @samp{m68k-*-coff} do provide software floating point support.
9093 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9094 Additionally, parameters passed on the stack are also aligned to a
9095 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9098 @opindex mnobitfield
9099 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9100 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9104 Do use the bit-field instructions. The @option{-m68020} option implies
9105 @option{-mbitfield}. This is the default if you use a configuration
9106 designed for a 68020.
9110 Use a different function-calling convention, in which functions
9111 that take a fixed number of arguments return with the @code{rtd}
9112 instruction, which pops their arguments while returning. This
9113 saves one instruction in the caller since there is no need to pop
9114 the arguments there.
9116 This calling convention is incompatible with the one normally
9117 used on Unix, so you cannot use it if you need to call libraries
9118 compiled with the Unix compiler.
9120 Also, you must provide function prototypes for all functions that
9121 take variable numbers of arguments (including @code{printf});
9122 otherwise incorrect code will be generated for calls to those
9125 In addition, seriously incorrect code will result if you call a
9126 function with too many arguments. (Normally, extra arguments are
9127 harmlessly ignored.)
9129 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9130 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9133 @itemx -mno-align-int
9135 @opindex mno-align-int
9136 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9137 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9138 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9139 Aligning variables on 32-bit boundaries produces code that runs somewhat
9140 faster on processors with 32-bit busses at the expense of more memory.
9142 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9143 align structures containing the above types differently than
9144 most published application binary interface specifications for the m68k.
9148 Use the pc-relative addressing mode of the 68000 directly, instead of
9149 using a global offset table. At present, this option implies @option{-fpic},
9150 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9151 not presently supported with @option{-mpcrel}, though this could be supported for
9152 68020 and higher processors.
9154 @item -mno-strict-align
9155 @itemx -mstrict-align
9156 @opindex mno-strict-align
9157 @opindex mstrict-align
9158 Do not (do) assume that unaligned memory references will be handled by
9162 Generate code that allows the data segment to be located in a different
9163 area of memory from the text segment. This allows for execute in place in
9164 an environment without virtual memory management. This option implies
9168 Generate code that assumes that the data segment follows the text segment.
9169 This is the default.
9171 @item -mid-shared-library
9172 Generate code that supports shared libraries via the library ID method.
9173 This allows for execute in place and shared libraries in an environment
9174 without virtual memory management. This option implies @option{-fPIC}.
9176 @item -mno-id-shared-library
9177 Generate code that doesn't assume ID based shared libraries are being used.
9178 This is the default.
9180 @item -mshared-library-id=n
9181 Specified the identification number of the ID based shared library being
9182 compiled. Specifying a value of 0 will generate more compact code, specifying
9183 other values will force the allocation of that number to the current
9184 library but is no more space or time efficient than omitting this option.
9188 @node M68hc1x Options
9189 @subsection M68hc1x Options
9190 @cindex M68hc1x options
9192 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9193 microcontrollers. The default values for these options depends on
9194 which style of microcontroller was selected when the compiler was configured;
9195 the defaults for the most common choices are given below.
9202 Generate output for a 68HC11. This is the default
9203 when the compiler is configured for 68HC11-based systems.
9209 Generate output for a 68HC12. This is the default
9210 when the compiler is configured for 68HC12-based systems.
9216 Generate output for a 68HCS12.
9219 @opindex mauto-incdec
9220 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9227 Enable the use of 68HC12 min and max instructions.
9230 @itemx -mno-long-calls
9231 @opindex mlong-calls
9232 @opindex mno-long-calls
9233 Treat all calls as being far away (near). If calls are assumed to be
9234 far away, the compiler will use the @code{call} instruction to
9235 call a function and the @code{rtc} instruction for returning.
9239 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9241 @item -msoft-reg-count=@var{count}
9242 @opindex msoft-reg-count
9243 Specify the number of pseudo-soft registers which are used for the
9244 code generation. The maximum number is 32. Using more pseudo-soft
9245 register may or may not result in better code depending on the program.
9246 The default is 4 for 68HC11 and 2 for 68HC12.
9251 @subsection MCore Options
9252 @cindex MCore options
9254 These are the @samp{-m} options defined for the Motorola M*Core
9262 @opindex mno-hardlit
9263 Inline constants into the code stream if it can be done in two
9264 instructions or less.
9270 Use the divide instruction. (Enabled by default).
9272 @item -mrelax-immediate
9273 @itemx -mno-relax-immediate
9274 @opindex mrelax-immediate
9275 @opindex mno-relax-immediate
9276 Allow arbitrary sized immediates in bit operations.
9278 @item -mwide-bitfields
9279 @itemx -mno-wide-bitfields
9280 @opindex mwide-bitfields
9281 @opindex mno-wide-bitfields
9282 Always treat bit-fields as int-sized.
9284 @item -m4byte-functions
9285 @itemx -mno-4byte-functions
9286 @opindex m4byte-functions
9287 @opindex mno-4byte-functions
9288 Force all functions to be aligned to a four byte boundary.
9290 @item -mcallgraph-data
9291 @itemx -mno-callgraph-data
9292 @opindex mcallgraph-data
9293 @opindex mno-callgraph-data
9294 Emit callgraph information.
9297 @itemx -mno-slow-bytes
9298 @opindex mslow-bytes
9299 @opindex mno-slow-bytes
9300 Prefer word access when reading byte quantities.
9302 @item -mlittle-endian
9304 @opindex mlittle-endian
9305 @opindex mbig-endian
9306 Generate code for a little endian target.
9312 Generate code for the 210 processor.
9316 @subsection MIPS Options
9317 @cindex MIPS options
9323 Generate big-endian code.
9327 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9330 @item -march=@var{arch}
9332 Generate code that will run on @var{arch}, which can be the name of a
9333 generic MIPS ISA, or the name of a particular processor.
9335 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9336 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9337 The processor names are:
9338 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9340 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9341 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9345 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9346 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9347 The special value @samp{from-abi} selects the
9348 most compatible architecture for the selected ABI (that is,
9349 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9351 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9352 (for example, @samp{-march=r2k}). Prefixes are optional, and
9353 @samp{vr} may be written @samp{r}.
9355 GCC defines two macros based on the value of this option. The first
9356 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9357 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9358 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9359 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9360 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9362 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9363 above. In other words, it will have the full prefix and will not
9364 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9365 the macro names the resolved architecture (either @samp{"mips1"} or
9366 @samp{"mips3"}). It names the default architecture when no
9367 @option{-march} option is given.
9369 @item -mtune=@var{arch}
9371 Optimize for @var{arch}. Among other things, this option controls
9372 the way instructions are scheduled, and the perceived cost of arithmetic
9373 operations. The list of @var{arch} values is the same as for
9376 When this option is not used, GCC will optimize for the processor
9377 specified by @option{-march}. By using @option{-march} and
9378 @option{-mtune} together, it is possible to generate code that will
9379 run on a family of processors, but optimize the code for one
9380 particular member of that family.
9382 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9383 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9384 @samp{-march} ones described above.
9388 Equivalent to @samp{-march=mips1}.
9392 Equivalent to @samp{-march=mips2}.
9396 Equivalent to @samp{-march=mips3}.
9400 Equivalent to @samp{-march=mips4}.
9404 Equivalent to @samp{-march=mips32}.
9408 Equivalent to @samp{-march=mips32r2}.
9412 Equivalent to @samp{-march=mips64}.
9418 Use (do not use) the MIPS16 ISA@.
9430 Generate code for the given ABI@.
9432 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9433 generates 64-bit code when you select a 64-bit architecture, but you
9434 can use @option{-mgp32} to get 32-bit code instead.
9436 For information about the O64 ABI, see
9437 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9440 @itemx -mno-abicalls
9442 @opindex mno-abicalls
9443 Generate (do not generate) SVR4-style position-independent code.
9444 @option{-mabicalls} is the default for SVR4-based systems.
9450 Lift (do not lift) the usual restrictions on the size of the global
9453 GCC normally uses a single instruction to load values from the GOT@.
9454 While this is relatively efficient, it will only work if the GOT
9455 is smaller than about 64k. Anything larger will cause the linker
9456 to report an error such as:
9458 @cindex relocation truncated to fit (MIPS)
9460 relocation truncated to fit: R_MIPS_GOT16 foobar
9463 If this happens, you should recompile your code with @option{-mxgot}.
9464 It should then work with very large GOTs, although it will also be
9465 less efficient, since it will take three instructions to fetch the
9466 value of a global symbol.
9468 Note that some linkers can create multiple GOTs. If you have such a
9469 linker, you should only need to use @option{-mxgot} when a single object
9470 file accesses more than 64k's worth of GOT entries. Very few do.
9472 These options have no effect unless GCC is generating position
9477 Assume that general-purpose registers are 32 bits wide.
9481 Assume that general-purpose registers are 64 bits wide.
9485 Assume that floating-point registers are 32 bits wide.
9489 Assume that floating-point registers are 64 bits wide.
9492 @opindex mhard-float
9493 Use floating-point coprocessor instructions.
9496 @opindex msoft-float
9497 Do not use floating-point coprocessor instructions. Implement
9498 floating-point calculations using library calls instead.
9500 @item -msingle-float
9501 @opindex msingle-float
9502 Assume that the floating-point coprocessor only supports single-precision
9505 @itemx -mdouble-float
9506 @opindex mdouble-float
9507 Assume that the floating-point coprocessor supports double-precision
9508 operations. This is the default.
9510 @itemx -mpaired-single
9511 @itemx -mno-paired-single
9512 @opindex mpaired-single
9513 @opindex mno-paired-single
9514 Use (do not use) paired-single floating-point instructions.
9515 @xref{MIPS Paired-Single Support}. This option can only be used
9516 when generating 64-bit code and requires hardware floating-point
9517 support to be enabled.
9523 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9524 The option @option{-mips3d} implies @option{-mpaired-single}.
9528 Force @code{int} and @code{long} types to be 64 bits wide. See
9529 @option{-mlong32} for an explanation of the default and the way
9530 that the pointer size is determined.
9534 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9535 an explanation of the default and the way that the pointer size is
9540 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9542 The default size of @code{int}s, @code{long}s and pointers depends on
9543 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9544 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9545 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9546 or the same size as integer registers, whichever is smaller.
9550 @cindex smaller data references (MIPS)
9551 @cindex gp-relative references (MIPS)
9552 Put global and static items less than or equal to @var{num} bytes into
9553 the small data or bss section instead of the normal data or bss section.
9554 This allows the data to be accessed using a single instruction.
9556 All modules should be compiled with the same @option{-G @var{num}}
9559 @item -membedded-data
9560 @itemx -mno-embedded-data
9561 @opindex membedded-data
9562 @opindex mno-embedded-data
9563 Allocate variables to the read-only data section first if possible, then
9564 next in the small data section if possible, otherwise in data. This gives
9565 slightly slower code than the default, but reduces the amount of RAM required
9566 when executing, and thus may be preferred for some embedded systems.
9568 @item -muninit-const-in-rodata
9569 @itemx -mno-uninit-const-in-rodata
9570 @opindex muninit-const-in-rodata
9571 @opindex mno-uninit-const-in-rodata
9572 Put uninitialized @code{const} variables in the read-only data section.
9573 This option is only meaningful in conjunction with @option{-membedded-data}.
9575 @item -msplit-addresses
9576 @itemx -mno-split-addresses
9577 @opindex msplit-addresses
9578 @opindex mno-split-addresses
9579 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9580 relocation operators. This option has been superceded by
9581 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9583 @item -mexplicit-relocs
9584 @itemx -mno-explicit-relocs
9585 @opindex mexplicit-relocs
9586 @opindex mno-explicit-relocs
9587 Use (do not use) assembler relocation operators when dealing with symbolic
9588 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9589 is to use assembler macros instead.
9591 @option{-mexplicit-relocs} is the default if GCC was configured
9592 to use an assembler that supports relocation operators.
9594 @item -mcheck-zero-division
9595 @itemx -mno-check-zero-division
9596 @opindex mcheck-zero-division
9597 @opindex mno-check-zero-division
9598 Trap (do not trap) on integer division by zero. The default is
9599 @option{-mcheck-zero-division}.
9601 @item -mdivide-traps
9602 @itemx -mdivide-breaks
9603 @opindex mdivide-traps
9604 @opindex mdivide-breaks
9605 MIPS systems check for division by zero by generating either a
9606 conditional trap or a break instruction. Using traps results in
9607 smaller code, but is only supported on MIPS II and later. Also, some
9608 versions of the Linux kernel have a bug that prevents trap from
9609 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
9610 allow conditional traps on architectures that support them and
9611 @option{-mdivide-breaks} to force the use of breaks.
9613 The default is usually @option{-mdivide-traps}, but this can be
9614 overridden at configure time using @option{--with-divide=breaks}.
9615 Divide-by-zero checks can be completely disabled using
9616 @option{-mno-check-zero-division}.
9622 Force (do not force) the use of @code{memcpy()} for non-trivial block
9623 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9624 most constant-sized copies.
9627 @itemx -mno-long-calls
9628 @opindex mlong-calls
9629 @opindex mno-long-calls
9630 Disable (do not disable) use of the @code{jal} instruction. Calling
9631 functions using @code{jal} is more efficient but requires the caller
9632 and callee to be in the same 256 megabyte segment.
9634 This option has no effect on abicalls code. The default is
9635 @option{-mno-long-calls}.
9641 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9642 instructions, as provided by the R4650 ISA@.
9645 @itemx -mno-fused-madd
9646 @opindex mfused-madd
9647 @opindex mno-fused-madd
9648 Enable (disable) use of the floating point multiply-accumulate
9649 instructions, when they are available. The default is
9650 @option{-mfused-madd}.
9652 When multiply-accumulate instructions are used, the intermediate
9653 product is calculated to infinite precision and is not subject to
9654 the FCSR Flush to Zero bit. This may be undesirable in some
9659 Tell the MIPS assembler to not run its preprocessor over user
9660 assembler files (with a @samp{.s} suffix) when assembling them.
9663 @itemx -mno-fix-r4000
9665 @opindex mno-fix-r4000
9666 Work around certain R4000 CPU errata:
9669 A double-word or a variable shift may give an incorrect result if executed
9670 immediately after starting an integer division.
9672 A double-word or a variable shift may give an incorrect result if executed
9673 while an integer multiplication is in progress.
9675 An integer division may give an incorrect result if started in a delay slot
9676 of a taken branch or a jump.
9680 @itemx -mno-fix-r4400
9682 @opindex mno-fix-r4400
9683 Work around certain R4400 CPU errata:
9686 A double-word or a variable shift may give an incorrect result if executed
9687 immediately after starting an integer division.
9691 @itemx -mno-fix-vr4120
9692 @opindex mfix-vr4120
9693 Work around certain VR4120 errata:
9696 @code{dmultu} does not always produce the correct result.
9698 @code{div} and @code{ddiv} do not always produce the correct result if one
9699 of the operands is negative.
9701 The workarounds for the division errata rely on special functions in
9702 @file{libgcc.a}. At present, these functions are only provided by
9703 the @code{mips64vr*-elf} configurations.
9705 Other VR4120 errata require a nop to be inserted between certain pairs of
9706 instructions. These errata are handled by the assembler, not by GCC itself.
9711 Work around certain SB-1 CPU core errata.
9712 (This flag currently works around the SB-1 revision 2
9713 ``F1'' and ``F2'' floating point errata.)
9715 @item -mflush-func=@var{func}
9716 @itemx -mno-flush-func
9717 @opindex mflush-func
9718 Specifies the function to call to flush the I and D caches, or to not
9719 call any such function. If called, the function must take the same
9720 arguments as the common @code{_flush_func()}, that is, the address of the
9721 memory range for which the cache is being flushed, the size of the
9722 memory range, and the number 3 (to flush both caches). The default
9723 depends on the target GCC was configured for, but commonly is either
9724 @samp{_flush_func} or @samp{__cpu_flush}.
9726 @item -mbranch-likely
9727 @itemx -mno-branch-likely
9728 @opindex mbranch-likely
9729 @opindex mno-branch-likely
9730 Enable or disable use of Branch Likely instructions, regardless of the
9731 default for the selected architecture. By default, Branch Likely
9732 instructions may be generated if they are supported by the selected
9733 architecture. An exception is for the MIPS32 and MIPS64 architectures
9734 and processors which implement those architectures; for those, Branch
9735 Likely instructions will not be generated by default because the MIPS32
9736 and MIPS64 architectures specifically deprecate their use.
9738 @item -mfp-exceptions
9739 @itemx -mno-fp-exceptions
9740 @opindex mfp-exceptions
9741 Specifies whether FP exceptions are enabled. This affects how we schedule
9742 FP instructions for some processors. The default is that FP exceptions are
9745 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
9746 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
9749 @item -mvr4130-align
9750 @itemx -mno-vr4130-align
9751 @opindex mvr4130-align
9752 The VR4130 pipeline is two-way superscalar, but can only issue two
9753 instructions together if the first one is 8-byte aligned. When this
9754 option is enabled, GCC will align pairs of instructions that it
9755 thinks should execute in parallel.
9757 This option only has an effect when optimizing for the VR4130.
9758 It normally makes code faster, but at the expense of making it bigger.
9759 It is enabled by default at optimization level @option{-O3}.
9763 @subsection MMIX Options
9764 @cindex MMIX Options
9766 These options are defined for the MMIX:
9770 @itemx -mno-libfuncs
9772 @opindex mno-libfuncs
9773 Specify that intrinsic library functions are being compiled, passing all
9774 values in registers, no matter the size.
9779 @opindex mno-epsilon
9780 Generate floating-point comparison instructions that compare with respect
9781 to the @code{rE} epsilon register.
9783 @item -mabi=mmixware
9785 @opindex mabi-mmixware
9787 Generate code that passes function parameters and return values that (in
9788 the called function) are seen as registers @code{$0} and up, as opposed to
9789 the GNU ABI which uses global registers @code{$231} and up.
9792 @itemx -mno-zero-extend
9793 @opindex mzero-extend
9794 @opindex mno-zero-extend
9795 When reading data from memory in sizes shorter than 64 bits, use (do not
9796 use) zero-extending load instructions by default, rather than
9797 sign-extending ones.
9800 @itemx -mno-knuthdiv
9802 @opindex mno-knuthdiv
9803 Make the result of a division yielding a remainder have the same sign as
9804 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
9805 remainder follows the sign of the dividend. Both methods are
9806 arithmetically valid, the latter being almost exclusively used.
9808 @item -mtoplevel-symbols
9809 @itemx -mno-toplevel-symbols
9810 @opindex mtoplevel-symbols
9811 @opindex mno-toplevel-symbols
9812 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
9813 code can be used with the @code{PREFIX} assembly directive.
9817 Generate an executable in the ELF format, rather than the default
9818 @samp{mmo} format used by the @command{mmix} simulator.
9820 @item -mbranch-predict
9821 @itemx -mno-branch-predict
9822 @opindex mbranch-predict
9823 @opindex mno-branch-predict
9824 Use (do not use) the probable-branch instructions, when static branch
9825 prediction indicates a probable branch.
9827 @item -mbase-addresses
9828 @itemx -mno-base-addresses
9829 @opindex mbase-addresses
9830 @opindex mno-base-addresses
9831 Generate (do not generate) code that uses @emph{base addresses}. Using a
9832 base address automatically generates a request (handled by the assembler
9833 and the linker) for a constant to be set up in a global register. The
9834 register is used for one or more base address requests within the range 0
9835 to 255 from the value held in the register. The generally leads to short
9836 and fast code, but the number of different data items that can be
9837 addressed is limited. This means that a program that uses lots of static
9838 data may require @option{-mno-base-addresses}.
9841 @itemx -mno-single-exit
9842 @opindex msingle-exit
9843 @opindex mno-single-exit
9844 Force (do not force) generated code to have a single exit point in each
9848 @node MN10300 Options
9849 @subsection MN10300 Options
9850 @cindex MN10300 options
9852 These @option{-m} options are defined for Matsushita MN10300 architectures:
9857 Generate code to avoid bugs in the multiply instructions for the MN10300
9858 processors. This is the default.
9861 @opindex mno-mult-bug
9862 Do not generate code to avoid bugs in the multiply instructions for the
9867 Generate code which uses features specific to the AM33 processor.
9871 Do not generate code which uses features specific to the AM33 processor. This
9876 Do not link in the C run-time initialization object file.
9880 Indicate to the linker that it should perform a relaxation optimization pass
9881 to shorten branches, calls and absolute memory addresses. This option only
9882 has an effect when used on the command line for the final link step.
9884 This option makes symbolic debugging impossible.
9888 @subsection NS32K Options
9889 @cindex NS32K options
9891 These are the @samp{-m} options defined for the 32000 series. The default
9892 values for these options depends on which style of 32000 was selected when
9893 the compiler was configured; the defaults for the most common choices are
9901 Generate output for a 32032. This is the default
9902 when the compiler is configured for 32032 and 32016 based systems.
9908 Generate output for a 32332. This is the default
9909 when the compiler is configured for 32332-based systems.
9915 Generate output for a 32532. This is the default
9916 when the compiler is configured for 32532-based systems.
9920 Generate output containing 32081 instructions for floating point.
9921 This is the default for all systems.
9925 Generate output containing 32381 instructions for floating point. This
9926 also implies @option{-m32081}. The 32381 is only compatible with the 32332
9927 and 32532 cpus. This is the default for the pc532-netbsd configuration.
9931 Try and generate multiply-add floating point instructions @code{polyF}
9932 and @code{dotF}. This option is only available if the @option{-m32381}
9933 option is in effect. Using these instructions requires changes to
9934 register allocation which generally has a negative impact on
9935 performance. This option should only be enabled when compiling code
9936 particularly likely to make heavy use of multiply-add instructions.
9939 @opindex mnomulti-add
9940 Do not try and generate multiply-add floating point instructions
9941 @code{polyF} and @code{dotF}. This is the default on all platforms.
9944 @opindex msoft-float
9945 Generate output containing library calls for floating point.
9946 @strong{Warning:} the requisite libraries may not be available.
9948 @item -mieee-compare
9949 @itemx -mno-ieee-compare
9950 @opindex mieee-compare
9951 @opindex mno-ieee-compare
9952 Control whether or not the compiler uses IEEE floating point
9953 comparisons. These handle correctly the case where the result of a
9954 comparison is unordered.
9955 @strong{Warning:} the requisite kernel support may not be available.
9958 @opindex mnobitfield
9959 Do not use the bit-field instructions. On some machines it is faster to
9960 use shifting and masking operations. This is the default for the pc532.
9964 Do use the bit-field instructions. This is the default for all platforms
9969 Use a different function-calling convention, in which functions
9970 that take a fixed number of arguments return pop their
9971 arguments on return with the @code{ret} instruction.
9973 This calling convention is incompatible with the one normally
9974 used on Unix, so you cannot use it if you need to call libraries
9975 compiled with the Unix compiler.
9977 Also, you must provide function prototypes for all functions that
9978 take variable numbers of arguments (including @code{printf});
9979 otherwise incorrect code will be generated for calls to those
9982 In addition, seriously incorrect code will result if you call a
9983 function with too many arguments. (Normally, extra arguments are
9984 harmlessly ignored.)
9986 This option takes its name from the 680x0 @code{rtd} instruction.
9991 Use a different function-calling convention where the first two arguments
9992 are passed in registers.
9994 This calling convention is incompatible with the one normally
9995 used on Unix, so you cannot use it if you need to call libraries
9996 compiled with the Unix compiler.
9999 @opindex mnoregparam
10000 Do not pass any arguments in registers. This is the default for all
10005 It is OK to use the sb as an index register which is always loaded with
10006 zero. This is the default for the pc532-netbsd target.
10010 The sb register is not available for use or has not been initialized to
10011 zero by the run time system. This is the default for all targets except
10012 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10013 @option{-fpic} is set.
10017 Many ns32000 series addressing modes use displacements of up to 512MB@.
10018 If an address is above 512MB then displacements from zero can not be used.
10019 This option causes code to be generated which can be loaded above 512MB@.
10020 This may be useful for operating systems or ROM code.
10024 Assume code will be loaded in the first 512MB of virtual address space.
10025 This is the default for all platforms.
10029 @node PDP-11 Options
10030 @subsection PDP-11 Options
10031 @cindex PDP-11 Options
10033 These options are defined for the PDP-11:
10038 Use hardware FPP floating point. This is the default. (FIS floating
10039 point on the PDP-11/40 is not supported.)
10042 @opindex msoft-float
10043 Do not use hardware floating point.
10047 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10051 Return floating-point results in memory. This is the default.
10055 Generate code for a PDP-11/40.
10059 Generate code for a PDP-11/45. This is the default.
10063 Generate code for a PDP-11/10.
10065 @item -mbcopy-builtin
10066 @opindex bcopy-builtin
10067 Use inline @code{movmemhi} patterns for copying memory. This is the
10072 Do not use inline @code{movmemhi} patterns for copying memory.
10078 Use 16-bit @code{int}. This is the default.
10084 Use 32-bit @code{int}.
10087 @itemx -mno-float32
10089 @opindex mno-float32
10090 Use 64-bit @code{float}. This is the default.
10093 @itemx -mno-float64
10095 @opindex mno-float64
10096 Use 32-bit @code{float}.
10100 Use @code{abshi2} pattern. This is the default.
10104 Do not use @code{abshi2} pattern.
10106 @item -mbranch-expensive
10107 @opindex mbranch-expensive
10108 Pretend that branches are expensive. This is for experimenting with
10109 code generation only.
10111 @item -mbranch-cheap
10112 @opindex mbranch-cheap
10113 Do not pretend that branches are expensive. This is the default.
10117 Generate code for a system with split I&D@.
10121 Generate code for a system without split I&D@. This is the default.
10125 Use Unix assembler syntax. This is the default when configured for
10126 @samp{pdp11-*-bsd}.
10130 Use DEC assembler syntax. This is the default when configured for any
10131 PDP-11 target other than @samp{pdp11-*-bsd}.
10134 @node PowerPC Options
10135 @subsection PowerPC Options
10136 @cindex PowerPC options
10138 These are listed under @xref{RS/6000 and PowerPC Options}.
10140 @node RS/6000 and PowerPC Options
10141 @subsection IBM RS/6000 and PowerPC Options
10142 @cindex RS/6000 and PowerPC Options
10143 @cindex IBM RS/6000 and PowerPC Options
10145 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10152 @itemx -mno-powerpc
10153 @itemx -mpowerpc-gpopt
10154 @itemx -mno-powerpc-gpopt
10155 @itemx -mpowerpc-gfxopt
10156 @itemx -mno-powerpc-gfxopt
10158 @itemx -mno-powerpc64
10162 @opindex mno-power2
10164 @opindex mno-powerpc
10165 @opindex mpowerpc-gpopt
10166 @opindex mno-powerpc-gpopt
10167 @opindex mpowerpc-gfxopt
10168 @opindex mno-powerpc-gfxopt
10169 @opindex mpowerpc64
10170 @opindex mno-powerpc64
10171 GCC supports two related instruction set architectures for the
10172 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10173 instructions supported by the @samp{rios} chip set used in the original
10174 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10175 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10176 the IBM 4xx microprocessors.
10178 Neither architecture is a subset of the other. However there is a
10179 large common subset of instructions supported by both. An MQ
10180 register is included in processors supporting the POWER architecture.
10182 You use these options to specify which instructions are available on the
10183 processor you are using. The default value of these options is
10184 determined when configuring GCC@. Specifying the
10185 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10186 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10187 rather than the options listed above.
10189 The @option{-mpower} option allows GCC to generate instructions that
10190 are found only in the POWER architecture and to use the MQ register.
10191 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10192 to generate instructions that are present in the POWER2 architecture but
10193 not the original POWER architecture.
10195 The @option{-mpowerpc} option allows GCC to generate instructions that
10196 are found only in the 32-bit subset of the PowerPC architecture.
10197 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10198 GCC to use the optional PowerPC architecture instructions in the
10199 General Purpose group, including floating-point square root. Specifying
10200 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10201 use the optional PowerPC architecture instructions in the Graphics
10202 group, including floating-point select.
10204 The @option{-mpowerpc64} option allows GCC to generate the additional
10205 64-bit instructions that are found in the full PowerPC64 architecture
10206 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10207 @option{-mno-powerpc64}.
10209 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10210 will use only the instructions in the common subset of both
10211 architectures plus some special AIX common-mode calls, and will not use
10212 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10213 permits GCC to use any instruction from either architecture and to
10214 allow use of the MQ register; specify this for the Motorola MPC601.
10216 @item -mnew-mnemonics
10217 @itemx -mold-mnemonics
10218 @opindex mnew-mnemonics
10219 @opindex mold-mnemonics
10220 Select which mnemonics to use in the generated assembler code. With
10221 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10222 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10223 assembler mnemonics defined for the POWER architecture. Instructions
10224 defined in only one architecture have only one mnemonic; GCC uses that
10225 mnemonic irrespective of which of these options is specified.
10227 GCC defaults to the mnemonics appropriate for the architecture in
10228 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10229 value of these option. Unless you are building a cross-compiler, you
10230 should normally not specify either @option{-mnew-mnemonics} or
10231 @option{-mold-mnemonics}, but should instead accept the default.
10233 @item -mcpu=@var{cpu_type}
10235 Set architecture type, register usage, choice of mnemonics, and
10236 instruction scheduling parameters for machine type @var{cpu_type}.
10237 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10238 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10239 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10240 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10241 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10242 @samp{860}, @samp{970}, @samp{common}, @samp{ec603e}, @samp{G3},
10243 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10244 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10245 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
10247 @option{-mcpu=common} selects a completely generic processor. Code
10248 generated under this option will run on any POWER or PowerPC processor.
10249 GCC will use only the instructions in the common subset of both
10250 architectures, and will not use the MQ register. GCC assumes a generic
10251 processor model for scheduling purposes.
10253 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10254 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10255 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10256 types, with an appropriate, generic processor model assumed for
10257 scheduling purposes.
10259 The other options specify a specific processor. Code generated under
10260 those options will run best on that processor, and may not run at all on
10263 The @option{-mcpu} options automatically enable or disable the
10264 following options: @option{-maltivec}, @option{-mhard-float},
10265 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10266 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10267 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10268 @option{-mstring}. The particular options set for any particular CPU
10269 will vary between compiler versions, depending on what setting seems
10270 to produce optimal code for that CPU; it doesn't necessarily reflect
10271 the actual hardware's capabilities. If you wish to set an individual
10272 option to a particular value, you may specify it after the
10273 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10275 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10276 not enabled or disabled by the @option{-mcpu} option at present, since
10277 AIX does not have full support for these options. You may still
10278 enable or disable them individually if you're sure it'll work in your
10281 @item -mtune=@var{cpu_type}
10283 Set the instruction scheduling parameters for machine type
10284 @var{cpu_type}, but do not set the architecture type, register usage, or
10285 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10286 values for @var{cpu_type} are used for @option{-mtune} as for
10287 @option{-mcpu}. If both are specified, the code generated will use the
10288 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10289 scheduling parameters set by @option{-mtune}.
10292 @itemx -mno-altivec
10294 @opindex mno-altivec
10295 Generate code that uses (does not use) AltiVec instructions, and also
10296 enable the use of built-in functions that allow more direct access to
10297 the AltiVec instruction set. You may also need to set
10298 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10303 Extend the current ABI with SPE ABI extensions. This does not change
10304 the default ABI, instead it adds the SPE ABI extensions to the current
10308 @opindex mabi=no-spe
10309 Disable Booke SPE ABI extensions for the current ABI@.
10311 @item -misel=@var{yes/no}
10314 This switch enables or disables the generation of ISEL instructions.
10316 @item -mspe=@var{yes/no}
10319 This switch enables or disables the generation of SPE simd
10322 @item -mfloat-gprs=@var{yes/single/double/no}
10323 @itemx -mfloat-gprs
10324 @opindex mfloat-gprs
10325 This switch enables or disables the generation of floating point
10326 operations on the general purpose registers for architectures that
10329 The argument @var{yes} or @var{single} enables the use of
10330 single-precision floating point operations.
10332 The argument @var{double} enables the use of single and
10333 double-precision floating point operations.
10335 The argument @var{no} disables floating point operations on the
10336 general purpose registers.
10338 This option is currently only available on the MPC854x.
10344 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10345 targets (including GNU/Linux). The 32-bit environment sets int, long
10346 and pointer to 32 bits and generates code that runs on any PowerPC
10347 variant. The 64-bit environment sets int to 32 bits and long and
10348 pointer to 64 bits, and generates code for PowerPC64, as for
10349 @option{-mpowerpc64}.
10352 @itemx -mno-fp-in-toc
10353 @itemx -mno-sum-in-toc
10354 @itemx -mminimal-toc
10356 @opindex mno-fp-in-toc
10357 @opindex mno-sum-in-toc
10358 @opindex mminimal-toc
10359 Modify generation of the TOC (Table Of Contents), which is created for
10360 every executable file. The @option{-mfull-toc} option is selected by
10361 default. In that case, GCC will allocate at least one TOC entry for
10362 each unique non-automatic variable reference in your program. GCC
10363 will also place floating-point constants in the TOC@. However, only
10364 16,384 entries are available in the TOC@.
10366 If you receive a linker error message that saying you have overflowed
10367 the available TOC space, you can reduce the amount of TOC space used
10368 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10369 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10370 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10371 generate code to calculate the sum of an address and a constant at
10372 run-time instead of putting that sum into the TOC@. You may specify one
10373 or both of these options. Each causes GCC to produce very slightly
10374 slower and larger code at the expense of conserving TOC space.
10376 If you still run out of space in the TOC even when you specify both of
10377 these options, specify @option{-mminimal-toc} instead. This option causes
10378 GCC to make only one TOC entry for every file. When you specify this
10379 option, GCC will produce code that is slower and larger but which
10380 uses extremely little TOC space. You may wish to use this option
10381 only on files that contain less frequently executed code.
10387 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10388 @code{long} type, and the infrastructure needed to support them.
10389 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10390 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10391 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10394 @itemx -mno-xl-call
10396 @opindex mno-xl-call
10397 On AIX, pass floating-point arguments to prototyped functions beyond the
10398 register save area (RSA) on the stack in addition to argument FPRs. The
10399 AIX calling convention was extended but not initially documented to
10400 handle an obscure K&R C case of calling a function that takes the
10401 address of its arguments with fewer arguments than declared. AIX XL
10402 compilers access floating point arguments which do not fit in the
10403 RSA from the stack when a subroutine is compiled without
10404 optimization. Because always storing floating-point arguments on the
10405 stack is inefficient and rarely needed, this option is not enabled by
10406 default and only is necessary when calling subroutines compiled by AIX
10407 XL compilers without optimization.
10411 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10412 application written to use message passing with special startup code to
10413 enable the application to run. The system must have PE installed in the
10414 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10415 must be overridden with the @option{-specs=} option to specify the
10416 appropriate directory location. The Parallel Environment does not
10417 support threads, so the @option{-mpe} option and the @option{-pthread}
10418 option are incompatible.
10420 @item -malign-natural
10421 @itemx -malign-power
10422 @opindex malign-natural
10423 @opindex malign-power
10424 On AIX, Darwin, and 64-bit PowerPC GNU/Linux, the option
10425 @option{-malign-natural} overrides the ABI-defined alignment of larger
10426 types, such as floating-point doubles, on their natural size-based boundary.
10427 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10428 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10431 @itemx -mhard-float
10432 @opindex msoft-float
10433 @opindex mhard-float
10434 Generate code that does not use (uses) the floating-point register set.
10435 Software floating point emulation is provided if you use the
10436 @option{-msoft-float} option, and pass the option to GCC when linking.
10439 @itemx -mno-multiple
10441 @opindex mno-multiple
10442 Generate code that uses (does not use) the load multiple word
10443 instructions and the store multiple word instructions. These
10444 instructions are generated by default on POWER systems, and not
10445 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10446 endian PowerPC systems, since those instructions do not work when the
10447 processor is in little endian mode. The exceptions are PPC740 and
10448 PPC750 which permit the instructions usage in little endian mode.
10453 @opindex mno-string
10454 Generate code that uses (does not use) the load string instructions
10455 and the store string word instructions to save multiple registers and
10456 do small block moves. These instructions are generated by default on
10457 POWER systems, and not generated on PowerPC systems. Do not use
10458 @option{-mstring} on little endian PowerPC systems, since those
10459 instructions do not work when the processor is in little endian mode.
10460 The exceptions are PPC740 and PPC750 which permit the instructions
10461 usage in little endian mode.
10466 @opindex mno-update
10467 Generate code that uses (does not use) the load or store instructions
10468 that update the base register to the address of the calculated memory
10469 location. These instructions are generated by default. If you use
10470 @option{-mno-update}, there is a small window between the time that the
10471 stack pointer is updated and the address of the previous frame is
10472 stored, which means code that walks the stack frame across interrupts or
10473 signals may get corrupted data.
10476 @itemx -mno-fused-madd
10477 @opindex mfused-madd
10478 @opindex mno-fused-madd
10479 Generate code that uses (does not use) the floating point multiply and
10480 accumulate instructions. These instructions are generated by default if
10481 hardware floating is used.
10483 @item -mno-bit-align
10485 @opindex mno-bit-align
10486 @opindex mbit-align
10487 On System V.4 and embedded PowerPC systems do not (do) force structures
10488 and unions that contain bit-fields to be aligned to the base type of the
10491 For example, by default a structure containing nothing but 8
10492 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10493 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10494 the structure would be aligned to a 1 byte boundary and be one byte in
10497 @item -mno-strict-align
10498 @itemx -mstrict-align
10499 @opindex mno-strict-align
10500 @opindex mstrict-align
10501 On System V.4 and embedded PowerPC systems do not (do) assume that
10502 unaligned memory references will be handled by the system.
10504 @item -mrelocatable
10505 @itemx -mno-relocatable
10506 @opindex mrelocatable
10507 @opindex mno-relocatable
10508 On embedded PowerPC systems generate code that allows (does not allow)
10509 the program to be relocated to a different address at runtime. If you
10510 use @option{-mrelocatable} on any module, all objects linked together must
10511 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10513 @item -mrelocatable-lib
10514 @itemx -mno-relocatable-lib
10515 @opindex mrelocatable-lib
10516 @opindex mno-relocatable-lib
10517 On embedded PowerPC systems generate code that allows (does not allow)
10518 the program to be relocated to a different address at runtime. Modules
10519 compiled with @option{-mrelocatable-lib} can be linked with either modules
10520 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10521 with modules compiled with the @option{-mrelocatable} options.
10527 On System V.4 and embedded PowerPC systems do not (do) assume that
10528 register 2 contains a pointer to a global area pointing to the addresses
10529 used in the program.
10532 @itemx -mlittle-endian
10534 @opindex mlittle-endian
10535 On System V.4 and embedded PowerPC systems compile code for the
10536 processor in little endian mode. The @option{-mlittle-endian} option is
10537 the same as @option{-mlittle}.
10540 @itemx -mbig-endian
10542 @opindex mbig-endian
10543 On System V.4 and embedded PowerPC systems compile code for the
10544 processor in big endian mode. The @option{-mbig-endian} option is
10545 the same as @option{-mbig}.
10547 @item -mdynamic-no-pic
10548 @opindex mdynamic-no-pic
10549 On Darwin and Mac OS X systems, compile code so that it is not
10550 relocatable, but that its external references are relocatable. The
10551 resulting code is suitable for applications, but not shared
10554 @item -mprioritize-restricted-insns=@var{priority}
10555 @opindex mprioritize-restricted-insns
10556 This option controls the priority that is assigned to
10557 dispatch-slot restricted instructions during the second scheduling
10558 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10559 @var{no/highest/second-highest} priority to dispatch slot restricted
10562 @item -msched-costly-dep=@var{dependence_type}
10563 @opindex msched-costly-dep
10564 This option controls which dependences are considered costly
10565 by the target during instruction scheduling. The argument
10566 @var{dependence_type} takes one of the following values:
10567 @var{no}: no dependence is costly,
10568 @var{all}: all dependences are costly,
10569 @var{true_store_to_load}: a true dependence from store to load is costly,
10570 @var{store_to_load}: any dependence from store to load is costly,
10571 @var{number}: any dependence which latency >= @var{number} is costly.
10573 @item -minsert-sched-nops=@var{scheme}
10574 @opindex minsert-sched-nops
10575 This option controls which nop insertion scheme will be used during
10576 the second scheduling pass. The argument @var{scheme} takes one of the
10578 @var{no}: Don't insert nops.
10579 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10580 according to the scheduler's grouping.
10581 @var{regroup_exact}: Insert nops to force costly dependent insns into
10582 separate groups. Insert exactly as many nops as needed to force an insn
10583 to a new group, according to the estimated processor grouping.
10584 @var{number}: Insert nops to force costly dependent insns into
10585 separate groups. Insert @var{number} nops to force an insn to a new group.
10588 @opindex mcall-sysv
10589 On System V.4 and embedded PowerPC systems compile code using calling
10590 conventions that adheres to the March 1995 draft of the System V
10591 Application Binary Interface, PowerPC processor supplement. This is the
10592 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10594 @item -mcall-sysv-eabi
10595 @opindex mcall-sysv-eabi
10596 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10598 @item -mcall-sysv-noeabi
10599 @opindex mcall-sysv-noeabi
10600 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10602 @item -mcall-solaris
10603 @opindex mcall-solaris
10604 On System V.4 and embedded PowerPC systems compile code for the Solaris
10608 @opindex mcall-linux
10609 On System V.4 and embedded PowerPC systems compile code for the
10610 Linux-based GNU system.
10614 On System V.4 and embedded PowerPC systems compile code for the
10615 Hurd-based GNU system.
10617 @item -mcall-netbsd
10618 @opindex mcall-netbsd
10619 On System V.4 and embedded PowerPC systems compile code for the
10620 NetBSD operating system.
10622 @item -maix-struct-return
10623 @opindex maix-struct-return
10624 Return all structures in memory (as specified by the AIX ABI)@.
10626 @item -msvr4-struct-return
10627 @opindex msvr4-struct-return
10628 Return structures smaller than 8 bytes in registers (as specified by the
10631 @item -mabi=altivec
10632 @opindex mabi=altivec
10633 Extend the current ABI with AltiVec ABI extensions. This does not
10634 change the default ABI, instead it adds the AltiVec ABI extensions to
10637 @item -mabi=no-altivec
10638 @opindex mabi=no-altivec
10639 Disable AltiVec ABI extensions for the current ABI@.
10642 @itemx -mno-prototype
10643 @opindex mprototype
10644 @opindex mno-prototype
10645 On System V.4 and embedded PowerPC systems assume that all calls to
10646 variable argument functions are properly prototyped. Otherwise, the
10647 compiler must insert an instruction before every non prototyped call to
10648 set or clear bit 6 of the condition code register (@var{CR}) to
10649 indicate whether floating point values were passed in the floating point
10650 registers in case the function takes a variable arguments. With
10651 @option{-mprototype}, only calls to prototyped variable argument functions
10652 will set or clear the bit.
10656 On embedded PowerPC systems, assume that the startup module is called
10657 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10658 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10663 On embedded PowerPC systems, assume that the startup module is called
10664 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10669 On embedded PowerPC systems, assume that the startup module is called
10670 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10673 @item -myellowknife
10674 @opindex myellowknife
10675 On embedded PowerPC systems, assume that the startup module is called
10676 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10681 On System V.4 and embedded PowerPC systems, specify that you are
10682 compiling for a VxWorks system.
10686 Specify that you are compiling for the WindISS simulation environment.
10690 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10691 header to indicate that @samp{eabi} extended relocations are used.
10697 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10698 Embedded Applications Binary Interface (eabi) which is a set of
10699 modifications to the System V.4 specifications. Selecting @option{-meabi}
10700 means that the stack is aligned to an 8 byte boundary, a function
10701 @code{__eabi} is called to from @code{main} to set up the eabi
10702 environment, and the @option{-msdata} option can use both @code{r2} and
10703 @code{r13} to point to two separate small data areas. Selecting
10704 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10705 do not call an initialization function from @code{main}, and the
10706 @option{-msdata} option will only use @code{r13} to point to a single
10707 small data area. The @option{-meabi} option is on by default if you
10708 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10711 @opindex msdata=eabi
10712 On System V.4 and embedded PowerPC systems, put small initialized
10713 @code{const} global and static data in the @samp{.sdata2} section, which
10714 is pointed to by register @code{r2}. Put small initialized
10715 non-@code{const} global and static data in the @samp{.sdata} section,
10716 which is pointed to by register @code{r13}. Put small uninitialized
10717 global and static data in the @samp{.sbss} section, which is adjacent to
10718 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10719 incompatible with the @option{-mrelocatable} option. The
10720 @option{-msdata=eabi} option also sets the @option{-memb} option.
10723 @opindex msdata=sysv
10724 On System V.4 and embedded PowerPC systems, put small global and static
10725 data in the @samp{.sdata} section, which is pointed to by register
10726 @code{r13}. Put small uninitialized global and static data in the
10727 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
10728 The @option{-msdata=sysv} option is incompatible with the
10729 @option{-mrelocatable} option.
10731 @item -msdata=default
10733 @opindex msdata=default
10735 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
10736 compile code the same as @option{-msdata=eabi}, otherwise compile code the
10737 same as @option{-msdata=sysv}.
10740 @opindex msdata-data
10741 On System V.4 and embedded PowerPC systems, put small global and static
10742 data in the @samp{.sdata} section. Put small uninitialized global and
10743 static data in the @samp{.sbss} section. Do not use register @code{r13}
10744 to address small data however. This is the default behavior unless
10745 other @option{-msdata} options are used.
10749 @opindex msdata=none
10751 On embedded PowerPC systems, put all initialized global and static data
10752 in the @samp{.data} section, and all uninitialized data in the
10753 @samp{.bss} section.
10757 @cindex smaller data references (PowerPC)
10758 @cindex .sdata/.sdata2 references (PowerPC)
10759 On embedded PowerPC systems, put global and static items less than or
10760 equal to @var{num} bytes into the small data or bss sections instead of
10761 the normal data or bss section. By default, @var{num} is 8. The
10762 @option{-G @var{num}} switch is also passed to the linker.
10763 All modules should be compiled with the same @option{-G @var{num}} value.
10766 @itemx -mno-regnames
10768 @opindex mno-regnames
10769 On System V.4 and embedded PowerPC systems do (do not) emit register
10770 names in the assembly language output using symbolic forms.
10773 @itemx -mno-longcall
10775 @opindex mno-longcall
10776 Default to making all function calls indirectly, using a register, so
10777 that functions which reside further than 32 megabytes (33,554,432
10778 bytes) from the current location can be called. This setting can be
10779 overridden by the @code{shortcall} function attribute, or by
10780 @code{#pragma longcall(0)}.
10782 Some linkers are capable of detecting out-of-range calls and generating
10783 glue code on the fly. On these systems, long calls are unnecessary and
10784 generate slower code. As of this writing, the AIX linker can do this,
10785 as can the GNU linker for PowerPC/64. It is planned to add this feature
10786 to the GNU linker for 32-bit PowerPC systems as well.
10788 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
10789 callee, L42'', plus a ``branch island'' (glue code). The two target
10790 addresses represent the callee and the ``branch island''. The
10791 Darwin/PPC linker will prefer the first address and generate a ``bl
10792 callee'' if the PPC ``bl'' instruction will reach the callee directly;
10793 otherwise, the linker will generate ``bl L42'' to call the ``branch
10794 island''. The ``branch island'' is appended to the body of the
10795 calling function; it computes the full 32-bit address of the callee
10798 On Mach-O (Darwin) systems, this option directs the compiler emit to
10799 the glue for every direct call, and the Darwin linker decides whether
10800 to use or discard it.
10802 In the future, we may cause GCC to ignore all longcall specifications
10803 when the linker is known to generate glue.
10807 Adds support for multithreading with the @dfn{pthreads} library.
10808 This option sets flags for both the preprocessor and linker.
10812 @node S/390 and zSeries Options
10813 @subsection S/390 and zSeries Options
10814 @cindex S/390 and zSeries Options
10816 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
10820 @itemx -msoft-float
10821 @opindex mhard-float
10822 @opindex msoft-float
10823 Use (do not use) the hardware floating-point instructions and registers
10824 for floating-point operations. When @option{-msoft-float} is specified,
10825 functions in @file{libgcc.a} will be used to perform floating-point
10826 operations. When @option{-mhard-float} is specified, the compiler
10827 generates IEEE floating-point instructions. This is the default.
10830 @itemx -mno-backchain
10831 @opindex mbackchain
10832 @opindex mno-backchain
10833 Store (do not store) the address of the caller's frame as backchain pointer
10834 into the callee's stack frame.
10835 A backchain may be needed to allow debugging using tools that do not understand
10836 DWARF-2 call frame information.
10837 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
10838 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
10839 the backchain is placed into the topmost word of the 96/160 byte register
10842 In general, code compiled with @option{-mbackchain} is call-compatible with
10843 code compiled with @option{-mmo-backchain}; however, use of the backchain
10844 for debugging purposes usually requires that the whole binary is built with
10845 @option{-mbackchain}. Note that the combination of @option{-mbackchain} and
10846 @option{-mpacked-stack} generates code that is not ABI-compatible.
10848 The default is to not maintain the backchain.
10850 @item -mpacked-stack
10851 @item -mno-packed-stack
10852 @opindex mpacked-stack
10853 @opindex mno-packed-stack
10854 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
10855 specified, the compiler uses the all fields of the 96/160 byte register save
10856 area only for their default purpose; unused fields still take up stack space.
10857 When @option{-mpacked-stack} is specified, register save slots are densely
10858 packed at the top of the register save area; unused space is reused for other
10859 purposes, allowing for more efficient use of the available stack space.
10860 However, when @option{-mbackchain} is also in effect, the topmost word of
10861 the save area is always used to store the backchain, and the return address
10862 register is always saved two words below the backchain.
10864 As long as the stack frame backchain is not used, code generated with
10865 @option{-mpacked-stack} is call-compatible with code generated with
10866 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
10867 S/390 or zSeries generated code that uses the stack frame backchain at run
10868 time, not just for debugging purposes. Such code is not call-compatible
10869 with code compiled with @option{-mpacked-stack}. Also, note that the
10870 combination of @option{-mbackchain} and @option{-mpacked-stack} generates code
10871 that is not ABI-compatible.
10873 The default is to not use the packed stack layout.
10876 @itemx -mno-small-exec
10877 @opindex msmall-exec
10878 @opindex mno-small-exec
10879 Generate (or do not generate) code using the @code{bras} instruction
10880 to do subroutine calls.
10881 This only works reliably if the total executable size does not
10882 exceed 64k. The default is to use the @code{basr} instruction instead,
10883 which does not have this limitation.
10889 When @option{-m31} is specified, generate code compliant to the
10890 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
10891 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
10892 particular to generate 64-bit instructions. For the @samp{s390}
10893 targets, the default is @option{-m31}, while the @samp{s390x}
10894 targets default to @option{-m64}.
10900 When @option{-mzarch} is specified, generate code using the
10901 instructions available on z/Architecture.
10902 When @option{-mesa} is specified, generate code using the
10903 instructions available on ESA/390. Note that @option{-mesa} is
10904 not possible with @option{-m64}.
10905 When generating code compliant to the GNU/Linux for S/390 ABI,
10906 the default is @option{-mesa}. When generating code compliant
10907 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
10913 Generate (or do not generate) code using the @code{mvcle} instruction
10914 to perform block moves. When @option{-mno-mvcle} is specified,
10915 use a @code{mvc} loop instead. This is the default.
10921 Print (or do not print) additional debug information when compiling.
10922 The default is to not print debug information.
10924 @item -march=@var{cpu-type}
10926 Generate code that will run on @var{cpu-type}, which is the name of a system
10927 representing a certain processor type. Possible values for
10928 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
10929 When generating code using the instructions available on z/Architecture,
10930 the default is @option{-march=z900}. Otherwise, the default is
10931 @option{-march=g5}.
10933 @item -mtune=@var{cpu-type}
10935 Tune to @var{cpu-type} everything applicable about the generated code,
10936 except for the ABI and the set of available instructions.
10937 The list of @var{cpu-type} values is the same as for @option{-march}.
10938 The default is the value used for @option{-march}.
10941 @itemx -mno-tpf-trace
10942 @opindex mtpf-trace
10943 @opindex mno-tpf-trace
10944 Generate code that adds (does not add) in TPF OS specific branches to trace
10945 routines in the operating system. This option is off by default, even
10946 when compiling for the TPF OS@.
10949 @itemx -mno-fused-madd
10950 @opindex mfused-madd
10951 @opindex mno-fused-madd
10952 Generate code that uses (does not use) the floating point multiply and
10953 accumulate instructions. These instructions are generated by default if
10954 hardware floating point is used.
10956 @item -mwarn-framesize=@var{framesize}
10957 @opindex mwarn-framesize
10958 Emit a warning if the current function exceeds the given frame size. Because
10959 this is a compile time check it doesn't need to be a real problem when the program
10960 runs. It is intended to identify functions which most probably cause
10961 a stack overflow. It is useful to be used in an environment with limited stack
10962 size e.g.@: the linux kernel.
10964 @item -mwarn-dynamicstack
10965 @opindex mwarn-dynamicstack
10966 Emit a warning if the function calls alloca or uses dynamically
10967 sized arrays. This is generally a bad idea with a limited stack size.
10969 @item -mstack-guard=@var{stack-guard}
10970 @item -mstack-size=@var{stack-size}
10971 @opindex mstack-guard
10972 @opindex mstack-size
10973 These arguments always have to be used in conjunction. If they are present the s390
10974 back end emits additional instructions in the function prologue which trigger a trap
10975 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
10976 (remember that the stack on s390 grows downward). These options are intended to
10977 be used to help debugging stack overflow problems. The additionally emitted code
10978 cause only little overhead and hence can also be used in production like systems
10979 without greater performance degradation. The given values have to be exact
10980 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
10981 In order to be efficient the extra code makes the assumption that the stack starts
10982 at an address aligned to the value given by @var{stack-size}.
10986 @subsection SH Options
10988 These @samp{-m} options are defined for the SH implementations:
10993 Generate code for the SH1.
10997 Generate code for the SH2.
11000 Generate code for the SH2e.
11004 Generate code for the SH3.
11008 Generate code for the SH3e.
11012 Generate code for the SH4 without a floating-point unit.
11014 @item -m4-single-only
11015 @opindex m4-single-only
11016 Generate code for the SH4 with a floating-point unit that only
11017 supports single-precision arithmetic.
11021 Generate code for the SH4 assuming the floating-point unit is in
11022 single-precision mode by default.
11026 Generate code for the SH4.
11030 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11031 floating-point unit is not used.
11033 @item -m4a-single-only
11034 @opindex m4a-single-only
11035 Generate code for the SH4a, in such a way that no double-precision
11036 floating point operations are used.
11039 @opindex m4a-single
11040 Generate code for the SH4a assuming the floating-point unit is in
11041 single-precision mode by default.
11045 Generate code for the SH4a.
11049 Same as @option{-m4a-nofpu}, except that it implicitly passes
11050 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11051 instructions at the moment.
11055 Compile code for the processor in big endian mode.
11059 Compile code for the processor in little endian mode.
11063 Align doubles at 64-bit boundaries. Note that this changes the calling
11064 conventions, and thus some functions from the standard C library will
11065 not work unless you recompile it first with @option{-mdalign}.
11069 Shorten some address references at link time, when possible; uses the
11070 linker option @option{-relax}.
11074 Use 32-bit offsets in @code{switch} tables. The default is to use
11079 Enable the use of the instruction @code{fmovd}.
11083 Comply with the calling conventions defined by Renesas.
11087 Comply with the calling conventions defined by Renesas.
11091 Comply with the calling conventions defined for GCC before the Renesas
11092 conventions were available. This option is the default for all
11093 targets of the SH toolchain except for @samp{sh-symbianelf}.
11096 @opindex mnomacsave
11097 Mark the @code{MAC} register as call-clobbered, even if
11098 @option{-mhitachi} is given.
11102 Increase IEEE-compliance of floating-point code.
11106 Dump instruction size and location in the assembly code.
11109 @opindex mpadstruct
11110 This option is deprecated. It pads structures to multiple of 4 bytes,
11111 which is incompatible with the SH ABI@.
11115 Optimize for space instead of speed. Implied by @option{-Os}.
11118 @opindex mprefergot
11119 When generating position-independent code, emit function calls using
11120 the Global Offset Table instead of the Procedure Linkage Table.
11124 Generate a library function call to invalidate instruction cache
11125 entries, after fixing up a trampoline. This library function call
11126 doesn't assume it can write to the whole memory address space. This
11127 is the default when the target is @code{sh-*-linux*}.
11130 @node SPARC Options
11131 @subsection SPARC Options
11132 @cindex SPARC options
11134 These @samp{-m} options are supported on the SPARC:
11137 @item -mno-app-regs
11139 @opindex mno-app-regs
11141 Specify @option{-mapp-regs} to generate output using the global registers
11142 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11145 To be fully SVR4 ABI compliant at the cost of some performance loss,
11146 specify @option{-mno-app-regs}. You should compile libraries and system
11147 software with this option.
11150 @itemx -mhard-float
11152 @opindex mhard-float
11153 Generate output containing floating point instructions. This is the
11157 @itemx -msoft-float
11159 @opindex msoft-float
11160 Generate output containing library calls for floating point.
11161 @strong{Warning:} the requisite libraries are not available for all SPARC
11162 targets. Normally the facilities of the machine's usual C compiler are
11163 used, but this cannot be done directly in cross-compilation. You must make
11164 your own arrangements to provide suitable library functions for
11165 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11166 @samp{sparclite-*-*} do provide software floating point support.
11168 @option{-msoft-float} changes the calling convention in the output file;
11169 therefore, it is only useful if you compile @emph{all} of a program with
11170 this option. In particular, you need to compile @file{libgcc.a}, the
11171 library that comes with GCC, with @option{-msoft-float} in order for
11174 @item -mhard-quad-float
11175 @opindex mhard-quad-float
11176 Generate output containing quad-word (long double) floating point
11179 @item -msoft-quad-float
11180 @opindex msoft-quad-float
11181 Generate output containing library calls for quad-word (long double)
11182 floating point instructions. The functions called are those specified
11183 in the SPARC ABI@. This is the default.
11185 As of this writing, there are no SPARC implementations that have hardware
11186 support for the quad-word floating point instructions. They all invoke
11187 a trap handler for one of these instructions, and then the trap handler
11188 emulates the effect of the instruction. Because of the trap handler overhead,
11189 this is much slower than calling the ABI library routines. Thus the
11190 @option{-msoft-quad-float} option is the default.
11192 @item -mno-unaligned-doubles
11193 @itemx -munaligned-doubles
11194 @opindex mno-unaligned-doubles
11195 @opindex munaligned-doubles
11196 Assume that doubles have 8 byte alignment. This is the default.
11198 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11199 alignment only if they are contained in another type, or if they have an
11200 absolute address. Otherwise, it assumes they have 4 byte alignment.
11201 Specifying this option avoids some rare compatibility problems with code
11202 generated by other compilers. It is not the default because it results
11203 in a performance loss, especially for floating point code.
11205 @item -mno-faster-structs
11206 @itemx -mfaster-structs
11207 @opindex mno-faster-structs
11208 @opindex mfaster-structs
11209 With @option{-mfaster-structs}, the compiler assumes that structures
11210 should have 8 byte alignment. This enables the use of pairs of
11211 @code{ldd} and @code{std} instructions for copies in structure
11212 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11213 However, the use of this changed alignment directly violates the SPARC
11214 ABI@. Thus, it's intended only for use on targets where the developer
11215 acknowledges that their resulting code will not be directly in line with
11216 the rules of the ABI@.
11218 @item -mimpure-text
11219 @opindex mimpure-text
11220 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11221 the compiler to not pass @option{-z text} to the linker when linking a
11222 shared object. Using this option, you can link position-dependent
11223 code into a shared object.
11225 @option{-mimpure-text} suppresses the ``relocations remain against
11226 allocatable but non-writable sections'' linker error message.
11227 However, the necessary relocations will trigger copy-on-write, and the
11228 shared object is not actually shared across processes. Instead of
11229 using @option{-mimpure-text}, you should compile all source code with
11230 @option{-fpic} or @option{-fPIC}.
11232 This option is only available on SunOS and Solaris.
11234 @item -mcpu=@var{cpu_type}
11236 Set the instruction set, register set, and instruction scheduling parameters
11237 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11238 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11239 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11240 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11241 @samp{ultrasparc3}.
11243 Default instruction scheduling parameters are used for values that select
11244 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11245 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11247 Here is a list of each supported architecture and their supported
11252 v8: supersparc, hypersparc
11253 sparclite: f930, f934, sparclite86x
11255 v9: ultrasparc, ultrasparc3
11258 By default (unless configured otherwise), GCC generates code for the V7
11259 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11260 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11261 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11262 SPARCStation 1, 2, IPX etc.
11264 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11265 architecture. The only difference from V7 code is that the compiler emits
11266 the integer multiply and integer divide instructions which exist in SPARC-V8
11267 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11268 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11271 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11272 the SPARC architecture. This adds the integer multiply, integer divide step
11273 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11274 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11275 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11276 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11277 MB86934 chip, which is the more recent SPARClite with FPU@.
11279 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11280 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11281 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11282 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11283 optimizes it for the TEMIC SPARClet chip.
11285 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11286 architecture. This adds 64-bit integer and floating-point move instructions,
11287 3 additional floating-point condition code registers and conditional move
11288 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11289 optimizes it for the Sun UltraSPARC I/II chips. With
11290 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11291 Sun UltraSPARC III chip.
11293 @item -mtune=@var{cpu_type}
11295 Set the instruction scheduling parameters for machine type
11296 @var{cpu_type}, but do not set the instruction set or register set that the
11297 option @option{-mcpu=@var{cpu_type}} would.
11299 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11300 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11301 that select a particular cpu implementation. Those are @samp{cypress},
11302 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11303 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11304 @samp{ultrasparc3}.
11309 @opindex mno-v8plus
11310 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11311 difference from the V8 ABI is that the global and out registers are
11312 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11313 mode for all SPARC-V9 processors.
11319 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11320 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11323 These @samp{-m} options are supported in addition to the above
11324 on SPARC-V9 processors in 64-bit environments:
11327 @item -mlittle-endian
11328 @opindex mlittle-endian
11329 Generate code for a processor running in little-endian mode. It is only
11330 available for a few configurations and most notably not on Solaris.
11336 Generate code for a 32-bit or 64-bit environment.
11337 The 32-bit environment sets int, long and pointer to 32 bits.
11338 The 64-bit environment sets int to 32 bits and long and pointer
11341 @item -mcmodel=medlow
11342 @opindex mcmodel=medlow
11343 Generate code for the Medium/Low code model: 64-bit addresses, programs
11344 must be linked in the low 32 bits of memory. Programs can be statically
11345 or dynamically linked.
11347 @item -mcmodel=medmid
11348 @opindex mcmodel=medmid
11349 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11350 must be linked in the low 44 bits of memory, the text and data segments must
11351 be less than 2GB in size and the data segment must be located within 2GB of
11354 @item -mcmodel=medany
11355 @opindex mcmodel=medany
11356 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11357 may be linked anywhere in memory, the text and data segments must be less
11358 than 2GB in size and the data segment must be located within 2GB of the
11361 @item -mcmodel=embmedany
11362 @opindex mcmodel=embmedany
11363 Generate code for the Medium/Anywhere code model for embedded systems:
11364 64-bit addresses, the text and data segments must be less than 2GB in
11365 size, both starting anywhere in memory (determined at link time). The
11366 global register %g4 points to the base of the data segment. Programs
11367 are statically linked and PIC is not supported.
11370 @itemx -mno-stack-bias
11371 @opindex mstack-bias
11372 @opindex mno-stack-bias
11373 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11374 frame pointer if present, are offset by @minus{}2047 which must be added back
11375 when making stack frame references. This is the default in 64-bit mode.
11376 Otherwise, assume no such offset is present.
11379 These switches are supported in addition to the above on Solaris:
11384 Add support for multithreading using the Solaris threads library. This
11385 option sets flags for both the preprocessor and linker. This option does
11386 not affect the thread safety of object code produced by the compiler or
11387 that of libraries supplied with it.
11391 Add support for multithreading using the POSIX threads library. This
11392 option sets flags for both the preprocessor and linker. This option does
11393 not affect the thread safety of object code produced by the compiler or
11394 that of libraries supplied with it.
11397 @node System V Options
11398 @subsection Options for System V
11400 These additional options are available on System V Release 4 for
11401 compatibility with other compilers on those systems:
11406 Create a shared object.
11407 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11411 Identify the versions of each tool used by the compiler, in a
11412 @code{.ident} assembler directive in the output.
11416 Refrain from adding @code{.ident} directives to the output file (this is
11419 @item -YP,@var{dirs}
11421 Search the directories @var{dirs}, and no others, for libraries
11422 specified with @option{-l}.
11424 @item -Ym,@var{dir}
11426 Look in the directory @var{dir} to find the M4 preprocessor.
11427 The assembler uses this option.
11428 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11429 @c the generic assembler that comes with Solaris takes just -Ym.
11432 @node TMS320C3x/C4x Options
11433 @subsection TMS320C3x/C4x Options
11434 @cindex TMS320C3x/C4x Options
11436 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11440 @item -mcpu=@var{cpu_type}
11442 Set the instruction set, register set, and instruction scheduling
11443 parameters for machine type @var{cpu_type}. Supported values for
11444 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11445 @samp{c44}. The default is @samp{c40} to generate code for the
11450 @itemx -msmall-memory
11452 @opindex mbig-memory
11454 @opindex msmall-memory
11456 Generates code for the big or small memory model. The small memory
11457 model assumed that all data fits into one 64K word page. At run-time
11458 the data page (DP) register must be set to point to the 64K page
11459 containing the .bss and .data program sections. The big memory model is
11460 the default and requires reloading of the DP register for every direct
11467 Allow (disallow) allocation of general integer operands into the block
11468 count register BK@.
11474 Enable (disable) generation of code using decrement and branch,
11475 DBcond(D), instructions. This is enabled by default for the C4x. To be
11476 on the safe side, this is disabled for the C3x, since the maximum
11477 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11478 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11479 that it can utilize the decrement and branch instruction, but will give
11480 up if there is more than one memory reference in the loop. Thus a loop
11481 where the loop counter is decremented can generate slightly more
11482 efficient code, in cases where the RPTB instruction cannot be utilized.
11484 @item -mdp-isr-reload
11486 @opindex mdp-isr-reload
11488 Force the DP register to be saved on entry to an interrupt service
11489 routine (ISR), reloaded to point to the data section, and restored on
11490 exit from the ISR@. This should not be required unless someone has
11491 violated the small memory model by modifying the DP register, say within
11498 For the C3x use the 24-bit MPYI instruction for integer multiplies
11499 instead of a library call to guarantee 32-bit results. Note that if one
11500 of the operands is a constant, then the multiplication will be performed
11501 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11502 then squaring operations are performed inline instead of a library call.
11505 @itemx -mno-fast-fix
11507 @opindex mno-fast-fix
11508 The C3x/C4x FIX instruction to convert a floating point value to an
11509 integer value chooses the nearest integer less than or equal to the
11510 floating point value rather than to the nearest integer. Thus if the
11511 floating point number is negative, the result will be incorrectly
11512 truncated an additional code is necessary to detect and correct this
11513 case. This option can be used to disable generation of the additional
11514 code required to correct the result.
11520 Enable (disable) generation of repeat block sequences using the RPTB
11521 instruction for zero overhead looping. The RPTB construct is only used
11522 for innermost loops that do not call functions or jump across the loop
11523 boundaries. There is no advantage having nested RPTB loops due to the
11524 overhead required to save and restore the RC, RS, and RE registers.
11525 This is enabled by default with @option{-O2}.
11527 @item -mrpts=@var{count}
11531 Enable (disable) the use of the single instruction repeat instruction
11532 RPTS@. If a repeat block contains a single instruction, and the loop
11533 count can be guaranteed to be less than the value @var{count}, GCC will
11534 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11535 then a RPTS will be emitted even if the loop count cannot be determined
11536 at compile time. Note that the repeated instruction following RPTS does
11537 not have to be reloaded from memory each iteration, thus freeing up the
11538 CPU buses for operands. However, since interrupts are blocked by this
11539 instruction, it is disabled by default.
11541 @item -mloop-unsigned
11542 @itemx -mno-loop-unsigned
11543 @opindex mloop-unsigned
11544 @opindex mno-loop-unsigned
11545 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11546 is @math{2^{31} + 1} since these instructions test if the iteration count is
11547 negative to terminate the loop. If the iteration count is unsigned
11548 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11549 exceeded. This switch allows an unsigned iteration count.
11553 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11554 with. This also enforces compatibility with the API employed by the TI
11555 C3x C compiler. For example, long doubles are passed as structures
11556 rather than in floating point registers.
11562 Generate code that uses registers (stack) for passing arguments to functions.
11563 By default, arguments are passed in registers where possible rather
11564 than by pushing arguments on to the stack.
11566 @item -mparallel-insns
11567 @itemx -mno-parallel-insns
11568 @opindex mparallel-insns
11569 @opindex mno-parallel-insns
11570 Allow the generation of parallel instructions. This is enabled by
11571 default with @option{-O2}.
11573 @item -mparallel-mpy
11574 @itemx -mno-parallel-mpy
11575 @opindex mparallel-mpy
11576 @opindex mno-parallel-mpy
11577 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11578 provided @option{-mparallel-insns} is also specified. These instructions have
11579 tight register constraints which can pessimize the code generation
11580 of large functions.
11585 @subsection V850 Options
11586 @cindex V850 Options
11588 These @samp{-m} options are defined for V850 implementations:
11592 @itemx -mno-long-calls
11593 @opindex mlong-calls
11594 @opindex mno-long-calls
11595 Treat all calls as being far away (near). If calls are assumed to be
11596 far away, the compiler will always load the functions address up into a
11597 register, and call indirect through the pointer.
11603 Do not optimize (do optimize) basic blocks that use the same index
11604 pointer 4 or more times to copy pointer into the @code{ep} register, and
11605 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11606 option is on by default if you optimize.
11608 @item -mno-prolog-function
11609 @itemx -mprolog-function
11610 @opindex mno-prolog-function
11611 @opindex mprolog-function
11612 Do not use (do use) external functions to save and restore registers
11613 at the prologue and epilogue of a function. The external functions
11614 are slower, but use less code space if more than one function saves
11615 the same number of registers. The @option{-mprolog-function} option
11616 is on by default if you optimize.
11620 Try to make the code as small as possible. At present, this just turns
11621 on the @option{-mep} and @option{-mprolog-function} options.
11623 @item -mtda=@var{n}
11625 Put static or global variables whose size is @var{n} bytes or less into
11626 the tiny data area that register @code{ep} points to. The tiny data
11627 area can hold up to 256 bytes in total (128 bytes for byte references).
11629 @item -msda=@var{n}
11631 Put static or global variables whose size is @var{n} bytes or less into
11632 the small data area that register @code{gp} points to. The small data
11633 area can hold up to 64 kilobytes.
11635 @item -mzda=@var{n}
11637 Put static or global variables whose size is @var{n} bytes or less into
11638 the first 32 kilobytes of memory.
11642 Specify that the target processor is the V850.
11645 @opindex mbig-switch
11646 Generate code suitable for big switch tables. Use this option only if
11647 the assembler/linker complain about out of range branches within a switch
11652 This option will cause r2 and r5 to be used in the code generated by
11653 the compiler. This setting is the default.
11655 @item -mno-app-regs
11656 @opindex mno-app-regs
11657 This option will cause r2 and r5 to be treated as fixed registers.
11661 Specify that the target processor is the V850E1. The preprocessor
11662 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11663 this option is used.
11667 Specify that the target processor is the V850E@. The preprocessor
11668 constant @samp{__v850e__} will be defined if this option is used.
11670 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11671 are defined then a default target processor will be chosen and the
11672 relevant @samp{__v850*__} preprocessor constant will be defined.
11674 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11675 defined, regardless of which processor variant is the target.
11677 @item -mdisable-callt
11678 @opindex mdisable-callt
11679 This option will suppress generation of the CALLT instruction for the
11680 v850e and v850e1 flavors of the v850 architecture. The default is
11681 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11686 @subsection VAX Options
11687 @cindex VAX options
11689 These @samp{-m} options are defined for the VAX:
11694 Do not output certain jump instructions (@code{aobleq} and so on)
11695 that the Unix assembler for the VAX cannot handle across long
11700 Do output those jump instructions, on the assumption that you
11701 will assemble with the GNU assembler.
11705 Output code for g-format floating point numbers instead of d-format.
11708 @node x86-64 Options
11709 @subsection x86-64 Options
11710 @cindex x86-64 options
11712 These are listed under @xref{i386 and x86-64 Options}.
11714 @node Xstormy16 Options
11715 @subsection Xstormy16 Options
11716 @cindex Xstormy16 Options
11718 These options are defined for Xstormy16:
11723 Choose startup files and linker script suitable for the simulator.
11726 @node Xtensa Options
11727 @subsection Xtensa Options
11728 @cindex Xtensa Options
11730 These options are supported for Xtensa targets:
11734 @itemx -mno-const16
11736 @opindex mno-const16
11737 Enable or disable use of @code{CONST16} instructions for loading
11738 constant values. The @code{CONST16} instruction is currently not a
11739 standard option from Tensilica. When enabled, @code{CONST16}
11740 instructions are always used in place of the standard @code{L32R}
11741 instructions. The use of @code{CONST16} is enabled by default only if
11742 the @code{L32R} instruction is not available.
11745 @itemx -mno-fused-madd
11746 @opindex mfused-madd
11747 @opindex mno-fused-madd
11748 Enable or disable use of fused multiply/add and multiply/subtract
11749 instructions in the floating-point option. This has no effect if the
11750 floating-point option is not also enabled. Disabling fused multiply/add
11751 and multiply/subtract instructions forces the compiler to use separate
11752 instructions for the multiply and add/subtract operations. This may be
11753 desirable in some cases where strict IEEE 754-compliant results are
11754 required: the fused multiply add/subtract instructions do not round the
11755 intermediate result, thereby producing results with @emph{more} bits of
11756 precision than specified by the IEEE standard. Disabling fused multiply
11757 add/subtract instructions also ensures that the program output is not
11758 sensitive to the compiler's ability to combine multiply and add/subtract
11761 @item -mtext-section-literals
11762 @itemx -mno-text-section-literals
11763 @opindex mtext-section-literals
11764 @opindex mno-text-section-literals
11765 Control the treatment of literal pools. The default is
11766 @option{-mno-text-section-literals}, which places literals in a separate
11767 section in the output file. This allows the literal pool to be placed
11768 in a data RAM/ROM, and it also allows the linker to combine literal
11769 pools from separate object files to remove redundant literals and
11770 improve code size. With @option{-mtext-section-literals}, the literals
11771 are interspersed in the text section in order to keep them as close as
11772 possible to their references. This may be necessary for large assembly
11775 @item -mtarget-align
11776 @itemx -mno-target-align
11777 @opindex mtarget-align
11778 @opindex mno-target-align
11779 When this option is enabled, GCC instructs the assembler to
11780 automatically align instructions to reduce branch penalties at the
11781 expense of some code density. The assembler attempts to widen density
11782 instructions to align branch targets and the instructions following call
11783 instructions. If there are not enough preceding safe density
11784 instructions to align a target, no widening will be performed. The
11785 default is @option{-mtarget-align}. These options do not affect the
11786 treatment of auto-aligned instructions like @code{LOOP}, which the
11787 assembler will always align, either by widening density instructions or
11788 by inserting no-op instructions.
11791 @itemx -mno-longcalls
11792 @opindex mlongcalls
11793 @opindex mno-longcalls
11794 When this option is enabled, GCC instructs the assembler to translate
11795 direct calls to indirect calls unless it can determine that the target
11796 of a direct call is in the range allowed by the call instruction. This
11797 translation typically occurs for calls to functions in other source
11798 files. Specifically, the assembler translates a direct @code{CALL}
11799 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
11800 The default is @option{-mno-longcalls}. This option should be used in
11801 programs where the call target can potentially be out of range. This
11802 option is implemented in the assembler, not the compiler, so the
11803 assembly code generated by GCC will still show direct call
11804 instructions---look at the disassembled object code to see the actual
11805 instructions. Note that the assembler will use an indirect call for
11806 every cross-file call, not just those that really will be out of range.
11809 @node zSeries Options
11810 @subsection zSeries Options
11811 @cindex zSeries options
11813 These are listed under @xref{S/390 and zSeries Options}.
11815 @node Code Gen Options
11816 @section Options for Code Generation Conventions
11817 @cindex code generation conventions
11818 @cindex options, code generation
11819 @cindex run-time options
11821 These machine-independent options control the interface conventions
11822 used in code generation.
11824 Most of them have both positive and negative forms; the negative form
11825 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
11826 one of the forms is listed---the one which is not the default. You
11827 can figure out the other form by either removing @samp{no-} or adding
11831 @item -fbounds-check
11832 @opindex fbounds-check
11833 For front-ends that support it, generate additional code to check that
11834 indices used to access arrays are within the declared range. This is
11835 currently only supported by the Java and Fortran 77 front-ends, where
11836 this option defaults to true and false respectively.
11840 This option generates traps for signed overflow on addition, subtraction,
11841 multiplication operations.
11845 This option instructs the compiler to assume that signed arithmetic
11846 overflow of addition, subtraction and multiplication wraps around
11847 using twos-complement representation. This flag enables some optimizations
11848 and disables other. This option is enabled by default for the Java
11849 front-end, as required by the Java language specification.
11852 @opindex fexceptions
11853 Enable exception handling. Generates extra code needed to propagate
11854 exceptions. For some targets, this implies GCC will generate frame
11855 unwind information for all functions, which can produce significant data
11856 size overhead, although it does not affect execution. If you do not
11857 specify this option, GCC will enable it by default for languages like
11858 C++ which normally require exception handling, and disable it for
11859 languages like C that do not normally require it. However, you may need
11860 to enable this option when compiling C code that needs to interoperate
11861 properly with exception handlers written in C++. You may also wish to
11862 disable this option if you are compiling older C++ programs that don't
11863 use exception handling.
11865 @item -fnon-call-exceptions
11866 @opindex fnon-call-exceptions
11867 Generate code that allows trapping instructions to throw exceptions.
11868 Note that this requires platform-specific runtime support that does
11869 not exist everywhere. Moreover, it only allows @emph{trapping}
11870 instructions to throw exceptions, i.e.@: memory references or floating
11871 point instructions. It does not allow exceptions to be thrown from
11872 arbitrary signal handlers such as @code{SIGALRM}.
11874 @item -funwind-tables
11875 @opindex funwind-tables
11876 Similar to @option{-fexceptions}, except that it will just generate any needed
11877 static data, but will not affect the generated code in any other way.
11878 You will normally not enable this option; instead, a language processor
11879 that needs this handling would enable it on your behalf.
11881 @item -fasynchronous-unwind-tables
11882 @opindex fasynchronous-unwind-tables
11883 Generate unwind table in dwarf2 format, if supported by target machine. The
11884 table is exact at each instruction boundary, so it can be used for stack
11885 unwinding from asynchronous events (such as debugger or garbage collector).
11887 @item -fpcc-struct-return
11888 @opindex fpcc-struct-return
11889 Return ``short'' @code{struct} and @code{union} values in memory like
11890 longer ones, rather than in registers. This convention is less
11891 efficient, but it has the advantage of allowing intercallability between
11892 GCC-compiled files and files compiled with other compilers, particularly
11893 the Portable C Compiler (pcc).
11895 The precise convention for returning structures in memory depends
11896 on the target configuration macros.
11898 Short structures and unions are those whose size and alignment match
11899 that of some integer type.
11901 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11902 switch is not binary compatible with code compiled with the
11903 @option{-freg-struct-return} switch.
11904 Use it to conform to a non-default application binary interface.
11906 @item -freg-struct-return
11907 @opindex freg-struct-return
11908 Return @code{struct} and @code{union} values in registers when possible.
11909 This is more efficient for small structures than
11910 @option{-fpcc-struct-return}.
11912 If you specify neither @option{-fpcc-struct-return} nor
11913 @option{-freg-struct-return}, GCC defaults to whichever convention is
11914 standard for the target. If there is no standard convention, GCC
11915 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11916 the principal compiler. In those cases, we can choose the standard, and
11917 we chose the more efficient register return alternative.
11919 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11920 switch is not binary compatible with code compiled with the
11921 @option{-fpcc-struct-return} switch.
11922 Use it to conform to a non-default application binary interface.
11924 @item -fshort-enums
11925 @opindex fshort-enums
11926 Allocate to an @code{enum} type only as many bytes as it needs for the
11927 declared range of possible values. Specifically, the @code{enum} type
11928 will be equivalent to the smallest integer type which has enough room.
11930 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11931 code that is not binary compatible with code generated without that switch.
11932 Use it to conform to a non-default application binary interface.
11934 @item -fshort-double
11935 @opindex fshort-double
11936 Use the same size for @code{double} as for @code{float}.
11938 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
11939 code that is not binary compatible with code generated without that switch.
11940 Use it to conform to a non-default application binary interface.
11942 @item -fshort-wchar
11943 @opindex fshort-wchar
11944 Override the underlying type for @samp{wchar_t} to be @samp{short
11945 unsigned int} instead of the default for the target. This option is
11946 useful for building programs to run under WINE@.
11948 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11949 code that is not binary compatible with code generated without that switch.
11950 Use it to conform to a non-default application binary interface.
11952 @item -fshared-data
11953 @opindex fshared-data
11954 Requests that the data and non-@code{const} variables of this
11955 compilation be shared data rather than private data. The distinction
11956 makes sense only on certain operating systems, where shared data is
11957 shared between processes running the same program, while private data
11958 exists in one copy per process.
11961 @opindex fno-common
11962 In C, allocate even uninitialized global variables in the data section of the
11963 object file, rather than generating them as common blocks. This has the
11964 effect that if the same variable is declared (without @code{extern}) in
11965 two different compilations, you will get an error when you link them.
11966 The only reason this might be useful is if you wish to verify that the
11967 program will work on other systems which always work this way.
11971 Ignore the @samp{#ident} directive.
11973 @item -finhibit-size-directive
11974 @opindex finhibit-size-directive
11975 Don't output a @code{.size} assembler directive, or anything else that
11976 would cause trouble if the function is split in the middle, and the
11977 two halves are placed at locations far apart in memory. This option is
11978 used when compiling @file{crtstuff.c}; you should not need to use it
11981 @item -fverbose-asm
11982 @opindex fverbose-asm
11983 Put extra commentary information in the generated assembly code to
11984 make it more readable. This option is generally only of use to those
11985 who actually need to read the generated assembly code (perhaps while
11986 debugging the compiler itself).
11988 @option{-fno-verbose-asm}, the default, causes the
11989 extra information to be omitted and is useful when comparing two assembler
11994 @cindex global offset table
11996 Generate position-independent code (PIC) suitable for use in a shared
11997 library, if supported for the target machine. Such code accesses all
11998 constant addresses through a global offset table (GOT)@. The dynamic
11999 loader resolves the GOT entries when the program starts (the dynamic
12000 loader is not part of GCC; it is part of the operating system). If
12001 the GOT size for the linked executable exceeds a machine-specific
12002 maximum size, you get an error message from the linker indicating that
12003 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12004 instead. (These maximums are 8k on the SPARC and 32k
12005 on the m68k and RS/6000. The 386 has no such limit.)
12007 Position-independent code requires special support, and therefore works
12008 only on certain machines. For the 386, GCC supports PIC for System V
12009 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12010 position-independent.
12014 If supported for the target machine, emit position-independent code,
12015 suitable for dynamic linking and avoiding any limit on the size of the
12016 global offset table. This option makes a difference on the m68k
12019 Position-independent code requires special support, and therefore works
12020 only on certain machines.
12026 These options are similar to @option{-fpic} and @option{-fPIC}, but
12027 generated position independent code can be only linked into executables.
12028 Usually these options are used when @option{-pie} GCC option will be
12029 used during linking.
12031 @item -ffixed-@var{reg}
12033 Treat the register named @var{reg} as a fixed register; generated code
12034 should never refer to it (except perhaps as a stack pointer, frame
12035 pointer or in some other fixed role).
12037 @var{reg} must be the name of a register. The register names accepted
12038 are machine-specific and are defined in the @code{REGISTER_NAMES}
12039 macro in the machine description macro file.
12041 This flag does not have a negative form, because it specifies a
12044 @item -fcall-used-@var{reg}
12045 @opindex fcall-used
12046 Treat the register named @var{reg} as an allocable register that is
12047 clobbered by function calls. It may be allocated for temporaries or
12048 variables that do not live across a call. Functions compiled this way
12049 will not save and restore the register @var{reg}.
12051 It is an error to used this flag with the frame pointer or stack pointer.
12052 Use of this flag for other registers that have fixed pervasive roles in
12053 the machine's execution model will produce disastrous results.
12055 This flag does not have a negative form, because it specifies a
12058 @item -fcall-saved-@var{reg}
12059 @opindex fcall-saved
12060 Treat the register named @var{reg} as an allocable register saved by
12061 functions. It may be allocated even for temporaries or variables that
12062 live across a call. Functions compiled this way will save and restore
12063 the register @var{reg} if they use it.
12065 It is an error to used this flag with the frame pointer or stack pointer.
12066 Use of this flag for other registers that have fixed pervasive roles in
12067 the machine's execution model will produce disastrous results.
12069 A different sort of disaster will result from the use of this flag for
12070 a register in which function values may be returned.
12072 This flag does not have a negative form, because it specifies a
12075 @item -fpack-struct[=@var{n}]
12076 @opindex fpack-struct
12077 Without a value specified, pack all structure members together without
12078 holes. When a value is specified (which must be a small power of two), pack
12079 structure members according to this value, representing the maximum
12080 alignment (that is, objects with default alignment requirements larger than
12081 this will be output potentially unaligned at the next fitting location.
12083 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12084 code that is not binary compatible with code generated without that switch.
12085 Additionally, it makes the code suboptimal.
12086 Use it to conform to a non-default application binary interface.
12088 @item -finstrument-functions
12089 @opindex finstrument-functions
12090 Generate instrumentation calls for entry and exit to functions. Just
12091 after function entry and just before function exit, the following
12092 profiling functions will be called with the address of the current
12093 function and its call site. (On some platforms,
12094 @code{__builtin_return_address} does not work beyond the current
12095 function, so the call site information may not be available to the
12096 profiling functions otherwise.)
12099 void __cyg_profile_func_enter (void *this_fn,
12101 void __cyg_profile_func_exit (void *this_fn,
12105 The first argument is the address of the start of the current function,
12106 which may be looked up exactly in the symbol table.
12108 This instrumentation is also done for functions expanded inline in other
12109 functions. The profiling calls will indicate where, conceptually, the
12110 inline function is entered and exited. This means that addressable
12111 versions of such functions must be available. If all your uses of a
12112 function are expanded inline, this may mean an additional expansion of
12113 code size. If you use @samp{extern inline} in your C code, an
12114 addressable version of such functions must be provided. (This is
12115 normally the case anyways, but if you get lucky and the optimizer always
12116 expands the functions inline, you might have gotten away without
12117 providing static copies.)
12119 A function may be given the attribute @code{no_instrument_function}, in
12120 which case this instrumentation will not be done. This can be used, for
12121 example, for the profiling functions listed above, high-priority
12122 interrupt routines, and any functions from which the profiling functions
12123 cannot safely be called (perhaps signal handlers, if the profiling
12124 routines generate output or allocate memory).
12126 @item -fstack-check
12127 @opindex fstack-check
12128 Generate code to verify that you do not go beyond the boundary of the
12129 stack. You should specify this flag if you are running in an
12130 environment with multiple threads, but only rarely need to specify it in
12131 a single-threaded environment since stack overflow is automatically
12132 detected on nearly all systems if there is only one stack.
12134 Note that this switch does not actually cause checking to be done; the
12135 operating system must do that. The switch causes generation of code
12136 to ensure that the operating system sees the stack being extended.
12138 @item -fstack-limit-register=@var{reg}
12139 @itemx -fstack-limit-symbol=@var{sym}
12140 @itemx -fno-stack-limit
12141 @opindex fstack-limit-register
12142 @opindex fstack-limit-symbol
12143 @opindex fno-stack-limit
12144 Generate code to ensure that the stack does not grow beyond a certain value,
12145 either the value of a register or the address of a symbol. If the stack
12146 would grow beyond the value, a signal is raised. For most targets,
12147 the signal is raised before the stack overruns the boundary, so
12148 it is possible to catch the signal without taking special precautions.
12150 For instance, if the stack starts at absolute address @samp{0x80000000}
12151 and grows downwards, you can use the flags
12152 @option{-fstack-limit-symbol=__stack_limit} and
12153 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12154 of 128KB@. Note that this may only work with the GNU linker.
12156 @cindex aliasing of parameters
12157 @cindex parameters, aliased
12158 @item -fargument-alias
12159 @itemx -fargument-noalias
12160 @itemx -fargument-noalias-global
12161 @opindex fargument-alias
12162 @opindex fargument-noalias
12163 @opindex fargument-noalias-global
12164 Specify the possible relationships among parameters and between
12165 parameters and global data.
12167 @option{-fargument-alias} specifies that arguments (parameters) may
12168 alias each other and may alias global storage.@*
12169 @option{-fargument-noalias} specifies that arguments do not alias
12170 each other, but may alias global storage.@*
12171 @option{-fargument-noalias-global} specifies that arguments do not
12172 alias each other and do not alias global storage.
12174 Each language will automatically use whatever option is required by
12175 the language standard. You should not need to use these options yourself.
12177 @item -fleading-underscore
12178 @opindex fleading-underscore
12179 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12180 change the way C symbols are represented in the object file. One use
12181 is to help link with legacy assembly code.
12183 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12184 generate code that is not binary compatible with code generated without that
12185 switch. Use it to conform to a non-default application binary interface.
12186 Not all targets provide complete support for this switch.
12188 @item -ftls-model=@var{model}
12189 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12190 The @var{model} argument should be one of @code{global-dynamic},
12191 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12193 The default without @option{-fpic} is @code{initial-exec}; with
12194 @option{-fpic} the default is @code{global-dynamic}.
12196 @item -fvisibility=@var{default|internal|hidden|protected}
12197 @opindex fvisibility
12198 Set the default ELF image symbol visibility to the specified option---all
12199 symbols will be marked with this unless overridden within the code.
12200 Using this feature can very substantially improve linking and
12201 load times of shared object libraries, produce more optimized
12202 code, provide near-perfect API export and prevent symbol clashes.
12203 It is @strong{strongly} recommended that you use this in any shared objects
12206 Despite the nomenclature, @code{default} always means public ie;
12207 available to be linked against from outside the shared object.
12208 @code{protected} and @code{internal} are pretty useless in real-world
12209 usage so the only other commonly used option will be @code{hidden}.
12210 The default if @option{-fvisibility} isn't specified is
12211 @code{default}, i.e., make every
12212 symbol public---this causes the same behavior as previous versions of
12215 A good explanation of the benefits offered by ensuring ELF
12216 symbols have the correct visibility is given by ``How To Write
12217 Shared Libraries'' by Ulrich Drepper (which can be found at
12218 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12219 solution made possible by this option to marking things hidden when
12220 the default is public is to make the default hidden and mark things
12221 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12222 and @code{__attribute__ ((visibility("default")))} instead of
12223 @code{__declspec(dllexport)} you get almost identical semantics with
12224 identical syntax. This is a great boon to those working with
12225 cross-platform projects.
12227 For those adding visibility support to existing code, you may find
12228 @samp{#pragma GCC visibility} of use. This works by you enclosing
12229 the declarations you wish to set visibility for with (for example)
12230 @samp{#pragma GCC visibility push(hidden)} and
12231 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12232 times. Bear in mind that symbol visibility should be viewed @strong{as
12233 part of the API interface contract} and thus all new code should
12234 always specify visibility when it is not the default ie; declarations
12235 only for use within the local DSO should @strong{always} be marked explicitly
12236 as hidden as so to avoid PLT indirection overheads---making this
12237 abundantly clear also aids readability and self-documentation of the code.
12238 Note that due to ISO C++ specification requirements, operator new and
12239 operator delete must always be of default visibility.
12241 An overview of these techniques, their benefits and how to use them
12242 is at @w{@uref{http://www.nedprod.com/programs/gccvisibility.html}}.
12248 @node Environment Variables
12249 @section Environment Variables Affecting GCC
12250 @cindex environment variables
12252 @c man begin ENVIRONMENT
12253 This section describes several environment variables that affect how GCC
12254 operates. Some of them work by specifying directories or prefixes to use
12255 when searching for various kinds of files. Some are used to specify other
12256 aspects of the compilation environment.
12258 Note that you can also specify places to search using options such as
12259 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12260 take precedence over places specified using environment variables, which
12261 in turn take precedence over those specified by the configuration of GCC@.
12262 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12263 GNU Compiler Collection (GCC) Internals}.
12268 @c @itemx LC_COLLATE
12270 @c @itemx LC_MONETARY
12271 @c @itemx LC_NUMERIC
12276 @c @findex LC_COLLATE
12277 @findex LC_MESSAGES
12278 @c @findex LC_MONETARY
12279 @c @findex LC_NUMERIC
12283 These environment variables control the way that GCC uses
12284 localization information that allow GCC to work with different
12285 national conventions. GCC inspects the locale categories
12286 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12287 so. These locale categories can be set to any value supported by your
12288 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12289 Kingdom encoded in UTF-8.
12291 The @env{LC_CTYPE} environment variable specifies character
12292 classification. GCC uses it to determine the character boundaries in
12293 a string; this is needed for some multibyte encodings that contain quote
12294 and escape characters that would otherwise be interpreted as a string
12297 The @env{LC_MESSAGES} environment variable specifies the language to
12298 use in diagnostic messages.
12300 If the @env{LC_ALL} environment variable is set, it overrides the value
12301 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12302 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12303 environment variable. If none of these variables are set, GCC
12304 defaults to traditional C English behavior.
12308 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12309 files. GCC uses temporary files to hold the output of one stage of
12310 compilation which is to be used as input to the next stage: for example,
12311 the output of the preprocessor, which is the input to the compiler
12314 @item GCC_EXEC_PREFIX
12315 @findex GCC_EXEC_PREFIX
12316 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12317 names of the subprograms executed by the compiler. No slash is added
12318 when this prefix is combined with the name of a subprogram, but you can
12319 specify a prefix that ends with a slash if you wish.
12321 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12322 an appropriate prefix to use based on the pathname it was invoked with.
12324 If GCC cannot find the subprogram using the specified prefix, it
12325 tries looking in the usual places for the subprogram.
12327 The default value of @env{GCC_EXEC_PREFIX} is
12328 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12329 of @code{prefix} when you ran the @file{configure} script.
12331 Other prefixes specified with @option{-B} take precedence over this prefix.
12333 This prefix is also used for finding files such as @file{crt0.o} that are
12336 In addition, the prefix is used in an unusual way in finding the
12337 directories to search for header files. For each of the standard
12338 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12339 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12340 replacing that beginning with the specified prefix to produce an
12341 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12342 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12343 These alternate directories are searched first; the standard directories
12346 @item COMPILER_PATH
12347 @findex COMPILER_PATH
12348 The value of @env{COMPILER_PATH} is a colon-separated list of
12349 directories, much like @env{PATH}. GCC tries the directories thus
12350 specified when searching for subprograms, if it can't find the
12351 subprograms using @env{GCC_EXEC_PREFIX}.
12354 @findex LIBRARY_PATH
12355 The value of @env{LIBRARY_PATH} is a colon-separated list of
12356 directories, much like @env{PATH}. When configured as a native compiler,
12357 GCC tries the directories thus specified when searching for special
12358 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12359 using GCC also uses these directories when searching for ordinary
12360 libraries for the @option{-l} option (but directories specified with
12361 @option{-L} come first).
12365 @cindex locale definition
12366 This variable is used to pass locale information to the compiler. One way in
12367 which this information is used is to determine the character set to be used
12368 when character literals, string literals and comments are parsed in C and C++.
12369 When the compiler is configured to allow multibyte characters,
12370 the following values for @env{LANG} are recognized:
12374 Recognize JIS characters.
12376 Recognize SJIS characters.
12378 Recognize EUCJP characters.
12381 If @env{LANG} is not defined, or if it has some other value, then the
12382 compiler will use mblen and mbtowc as defined by the default locale to
12383 recognize and translate multibyte characters.
12387 Some additional environments variables affect the behavior of the
12390 @include cppenv.texi
12394 @node Precompiled Headers
12395 @section Using Precompiled Headers
12396 @cindex precompiled headers
12397 @cindex speed of compilation
12399 Often large projects have many header files that are included in every
12400 source file. The time the compiler takes to process these header files
12401 over and over again can account for nearly all of the time required to
12402 build the project. To make builds faster, GCC allows users to
12403 `precompile' a header file; then, if builds can use the precompiled
12404 header file they will be much faster.
12406 @strong{Caution:} There are a few known situations where GCC will
12407 crash when trying to use a precompiled header. If you have trouble
12408 with a precompiled header, you should remove the precompiled header
12409 and compile without it. In addition, please use GCC's on-line
12410 defect-tracking system to report any problems you encounter with
12411 precompiled headers. @xref{Bugs}.
12413 To create a precompiled header file, simply compile it as you would any
12414 other file, if necessary using the @option{-x} option to make the driver
12415 treat it as a C or C++ header file. You will probably want to use a
12416 tool like @command{make} to keep the precompiled header up-to-date when
12417 the headers it contains change.
12419 A precompiled header file will be searched for when @code{#include} is
12420 seen in the compilation. As it searches for the included file
12421 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12422 compiler looks for a precompiled header in each directory just before it
12423 looks for the include file in that directory. The name searched for is
12424 the name specified in the @code{#include} with @samp{.gch} appended. If
12425 the precompiled header file can't be used, it is ignored.
12427 For instance, if you have @code{#include "all.h"}, and you have
12428 @file{all.h.gch} in the same directory as @file{all.h}, then the
12429 precompiled header file will be used if possible, and the original
12430 header will be used otherwise.
12432 Alternatively, you might decide to put the precompiled header file in a
12433 directory and use @option{-I} to ensure that directory is searched
12434 before (or instead of) the directory containing the original header.
12435 Then, if you want to check that the precompiled header file is always
12436 used, you can put a file of the same name as the original header in this
12437 directory containing an @code{#error} command.
12439 This also works with @option{-include}. So yet another way to use
12440 precompiled headers, good for projects not designed with precompiled
12441 header files in mind, is to simply take most of the header files used by
12442 a project, include them from another header file, precompile that header
12443 file, and @option{-include} the precompiled header. If the header files
12444 have guards against multiple inclusion, they will be skipped because
12445 they've already been included (in the precompiled header).
12447 If you need to precompile the same header file for different
12448 languages, targets, or compiler options, you can instead make a
12449 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12450 header in the directory, perhaps using @option{-o}. It doesn't matter
12451 what you call the files in the directory, every precompiled header in
12452 the directory will be considered. The first precompiled header
12453 encountered in the directory that is valid for this compilation will
12454 be used; they're searched in no particular order.
12456 There are many other possibilities, limited only by your imagination,
12457 good sense, and the constraints of your build system.
12459 A precompiled header file can be used only when these conditions apply:
12463 Only one precompiled header can be used in a particular compilation.
12466 A precompiled header can't be used once the first C token is seen. You
12467 can have preprocessor directives before a precompiled header; you can
12468 even include a precompiled header from inside another header, so long as
12469 there are no C tokens before the @code{#include}.
12472 The precompiled header file must be produced for the same language as
12473 the current compilation. You can't use a C precompiled header for a C++
12477 The precompiled header file must be produced by the same compiler
12478 version and configuration as the current compilation is using.
12479 The easiest way to guarantee this is to use the same compiler binary
12480 for creating and using precompiled headers.
12483 Any macros defined before the precompiled header is included must
12484 either be defined in the same way as when the precompiled header was
12485 generated, or must not affect the precompiled header, which usually
12486 means that the they don't appear in the precompiled header at all.
12488 The @option{-D} option is one way to define a macro before a
12489 precompiled header is included; using a @code{#define} can also do it.
12490 There are also some options that define macros implicitly, like
12491 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12494 @item If debugging information is output when using the precompiled
12495 header, using @option{-g} or similar, the same kind of debugging information
12496 must have been output when building the precompiled header. However,
12497 a precompiled header built using @option{-g} can be used in a compilation
12498 when no debugging information is being output.
12500 @item The same @option{-m} options must generally be used when building
12501 and using the precompiled header. @xref{Submodel Options},
12502 for any cases where this rule is relaxed.
12504 @item Each of the following options must be the same when building and using
12505 the precompiled header:
12507 @gccoptlist{-fexceptions -funit-at-a-time}
12510 Some other command-line options starting with @option{-f},
12511 @option{-p}, or @option{-O} must be defined in the same way as when
12512 the precompiled header was generated. At present, it's not clear
12513 which options are safe to change and which are not; the safest choice
12514 is to use exactly the same options when generating and using the
12515 precompiled header. The following are known to be safe:
12517 @gccoptlist{-fpreprocessed -pedantic-errors}
12521 For all of these except the last, the compiler will automatically
12522 ignore the precompiled header if the conditions aren't met. If you
12523 find an option combination that doesn't work and doesn't cause the
12524 precompiled header to be ignored, please consider filing a bug report,
12527 If you do use differing options when generating and using the
12528 precompiled header, the actual behavior will be a mixture of the
12529 behavior for the options. For instance, if you use @option{-g} to
12530 generate the precompiled header but not when using it, you may or may
12531 not get debugging information for routines in the precompiled header.
12533 @node Running Protoize
12534 @section Running Protoize
12536 The program @code{protoize} is an optional part of GCC@. You can use
12537 it to add prototypes to a program, thus converting the program to ISO
12538 C in one respect. The companion program @code{unprotoize} does the
12539 reverse: it removes argument types from any prototypes that are found.
12541 When you run these programs, you must specify a set of source files as
12542 command line arguments. The conversion programs start out by compiling
12543 these files to see what functions they define. The information gathered
12544 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12546 After scanning comes actual conversion. The specified files are all
12547 eligible to be converted; any files they include (whether sources or
12548 just headers) are eligible as well.
12550 But not all the eligible files are converted. By default,
12551 @code{protoize} and @code{unprotoize} convert only source and header
12552 files in the current directory. You can specify additional directories
12553 whose files should be converted with the @option{-d @var{directory}}
12554 option. You can also specify particular files to exclude with the
12555 @option{-x @var{file}} option. A file is converted if it is eligible, its
12556 directory name matches one of the specified directory names, and its
12557 name within the directory has not been excluded.
12559 Basic conversion with @code{protoize} consists of rewriting most
12560 function definitions and function declarations to specify the types of
12561 the arguments. The only ones not rewritten are those for varargs
12564 @code{protoize} optionally inserts prototype declarations at the
12565 beginning of the source file, to make them available for any calls that
12566 precede the function's definition. Or it can insert prototype
12567 declarations with block scope in the blocks where undeclared functions
12570 Basic conversion with @code{unprotoize} consists of rewriting most
12571 function declarations to remove any argument types, and rewriting
12572 function definitions to the old-style pre-ISO form.
12574 Both conversion programs print a warning for any function declaration or
12575 definition that they can't convert. You can suppress these warnings
12578 The output from @code{protoize} or @code{unprotoize} replaces the
12579 original source file. The original file is renamed to a name ending
12580 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12581 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12582 for DOS) file already exists, then the source file is simply discarded.
12584 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12585 scan the program and collect information about the functions it uses.
12586 So neither of these programs will work until GCC is installed.
12588 Here is a table of the options you can use with @code{protoize} and
12589 @code{unprotoize}. Each option works with both programs unless
12593 @item -B @var{directory}
12594 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12595 usual directory (normally @file{/usr/local/lib}). This file contains
12596 prototype information about standard system functions. This option
12597 applies only to @code{protoize}.
12599 @item -c @var{compilation-options}
12600 Use @var{compilation-options} as the options when running @command{gcc} to
12601 produce the @samp{.X} files. The special option @option{-aux-info} is
12602 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12604 Note that the compilation options must be given as a single argument to
12605 @code{protoize} or @code{unprotoize}. If you want to specify several
12606 @command{gcc} options, you must quote the entire set of compilation options
12607 to make them a single word in the shell.
12609 There are certain @command{gcc} arguments that you cannot use, because they
12610 would produce the wrong kind of output. These include @option{-g},
12611 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12612 the @var{compilation-options}, they are ignored.
12615 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12616 systems) instead of @samp{.c}. This is convenient if you are converting
12617 a C program to C++. This option applies only to @code{protoize}.
12620 Add explicit global declarations. This means inserting explicit
12621 declarations at the beginning of each source file for each function
12622 that is called in the file and was not declared. These declarations
12623 precede the first function definition that contains a call to an
12624 undeclared function. This option applies only to @code{protoize}.
12626 @item -i @var{string}
12627 Indent old-style parameter declarations with the string @var{string}.
12628 This option applies only to @code{protoize}.
12630 @code{unprotoize} converts prototyped function definitions to old-style
12631 function definitions, where the arguments are declared between the
12632 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12633 uses five spaces as the indentation. If you want to indent with just
12634 one space instead, use @option{-i " "}.
12637 Keep the @samp{.X} files. Normally, they are deleted after conversion
12641 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12642 a prototype declaration for each function in each block which calls the
12643 function without any declaration. This option applies only to
12647 Make no real changes. This mode just prints information about the conversions
12648 that would have been done without @option{-n}.
12651 Make no @samp{.save} files. The original files are simply deleted.
12652 Use this option with caution.
12654 @item -p @var{program}
12655 Use the program @var{program} as the compiler. Normally, the name
12656 @file{gcc} is used.
12659 Work quietly. Most warnings are suppressed.
12662 Print the version number, just like @option{-v} for @command{gcc}.
12665 If you need special compiler options to compile one of your program's
12666 source files, then you should generate that file's @samp{.X} file
12667 specially, by running @command{gcc} on that source file with the
12668 appropriate options and the option @option{-aux-info}. Then run
12669 @code{protoize} on the entire set of files. @code{protoize} will use
12670 the existing @samp{.X} file because it is newer than the source file.
12674 gcc -Dfoo=bar file1.c -aux-info file1.X
12679 You need to include the special files along with the rest in the
12680 @code{protoize} command, even though their @samp{.X} files already
12681 exist, because otherwise they won't get converted.
12683 @xref{Protoize Caveats}, for more information on how to use
12684 @code{protoize} successfully.