1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
11 @c man begin COPYRIGHT
12 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
13 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
15 Permission is granted to copy, distribute and/or modify this document
16 under the terms of the GNU Free Documentation License, Version 1.2 or
17 any later version published by the Free Software Foundation; with the
18 Invariant Sections being ``GNU General Public License'' and ``Funding
19 Free Software'', the Front-Cover texts being (a) (see below), and with
20 the Back-Cover Texts being (b) (see below). A copy of the license is
21 included in the gfdl(7) man page.
23 (a) The FSF's Front-Cover Text is:
27 (b) The FSF's Back-Cover Text is:
29 You have freedom to copy and modify this GNU Manual, like GNU
30 software. Copies published by the Free Software Foundation raise
31 funds for GNU development.
33 @c Set file name and title for the man page.
35 @settitle GNU project C and C++ compiler
37 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
38 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
39 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
40 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
41 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
42 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
43 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
45 Only the most useful options are listed here; see below for the
46 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
49 gpl(7), gfdl(7), fsf-funding(7),
50 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
51 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
52 @file{ld}, @file{binutils} and @file{gdb}.
55 For instructions on reporting bugs, see
56 @w{@uref{http://gcc.gnu.org/bugs.html}}.
59 See the Info entry for @command{gcc}, or
60 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
61 for contributors to GCC@.
66 @chapter GCC Command Options
67 @cindex GCC command options
68 @cindex command options
69 @cindex options, GCC command
71 @c man begin DESCRIPTION
72 When you invoke GCC, it normally does preprocessing, compilation,
73 assembly and linking. The ``overall options'' allow you to stop this
74 process at an intermediate stage. For example, the @option{-c} option
75 says not to run the linker. Then the output consists of object files
76 output by the assembler.
78 Other options are passed on to one stage of processing. Some options
79 control the preprocessor and others the compiler itself. Yet other
80 options control the assembler and linker; most of these are not
81 documented here, since you rarely need to use any of them.
83 @cindex C compilation options
84 Most of the command line options that you can use with GCC are useful
85 for C programs; when an option is only useful with another language
86 (usually C++), the explanation says so explicitly. If the description
87 for a particular option does not mention a source language, you can use
88 that option with all supported languages.
90 @cindex C++ compilation options
91 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
92 options for compiling C++ programs.
94 @cindex grouping options
95 @cindex options, grouping
96 The @command{gcc} program accepts options and file names as operands. Many
97 options have multi-letter names; therefore multiple single-letter options
98 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
101 @cindex order of options
102 @cindex options, order
103 You can mix options and other arguments. For the most part, the order
104 you use doesn't matter. Order does matter when you use several options
105 of the same kind; for example, if you specify @option{-L} more than once,
106 the directories are searched in the order specified.
108 Many options have long names starting with @samp{-f} or with
109 @samp{-W}---for example,
110 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
111 these have both positive and negative forms; the negative form of
112 @option{-ffoo} would be @option{-fno-foo}. This manual documents
113 only one of these two forms, whichever one is not the default.
117 @xref{Option Index}, for an index to GCC's options.
120 * Option Summary:: Brief list of all options, without explanations.
121 * Overall Options:: Controlling the kind of output:
122 an executable, object files, assembler files,
123 or preprocessed source.
124 * Invoking G++:: Compiling C++ programs.
125 * C Dialect Options:: Controlling the variant of C language compiled.
126 * C++ Dialect Options:: Variations on C++.
127 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
129 * Language Independent Options:: Controlling how diagnostics should be
131 * Warning Options:: How picky should the compiler be?
132 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
133 * Optimize Options:: How much optimization?
134 * Preprocessor Options:: Controlling header files and macro definitions.
135 Also, getting dependency information for Make.
136 * Assembler Options:: Passing options to the assembler.
137 * Link Options:: Specifying libraries and so on.
138 * Directory Options:: Where to find header files and libraries.
139 Where to find the compiler executable files.
140 * Spec Files:: How to pass switches to sub-processes.
141 * Target Options:: Running a cross-compiler, or an old version of GCC.
142 * Submodel Options:: Specifying minor hardware or convention variations,
143 such as 68010 vs 68020.
144 * Code Gen Options:: Specifying conventions for function calls, data layout
146 * Environment Variables:: Env vars that affect GCC.
147 * Precompiled Headers:: Compiling a header once, and using it many times.
148 * Running Protoize:: Automatically adding or removing function prototypes.
154 @section Option Summary
156 Here is a summary of all the options, grouped by type. Explanations are
157 in the following sections.
160 @item Overall Options
161 @xref{Overall Options,,Options Controlling the Kind of Output}.
162 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
163 -x @var{language} -v -### --help --target-help --version @@@var{file}}
165 @item C Language Options
166 @xref{C Dialect Options,,Options Controlling C Dialect}.
167 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
168 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
169 -fhosted -ffreestanding -fms-extensions @gol
170 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
171 -fallow-single-precision -fcond-mismatch @gol
172 -fsigned-bitfields -fsigned-char @gol
173 -funsigned-bitfields -funsigned-char}
175 @item C++ Language Options
176 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
177 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
178 -fconserve-space -ffriend-injection -fno-const-strings @gol
179 -fno-elide-constructors @gol
180 -fno-enforce-eh-specs @gol
181 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
182 -fno-implicit-templates @gol
183 -fno-implicit-inline-templates @gol
184 -fno-implement-inlines -fms-extensions @gol
185 -fno-nonansi-builtins -fno-operator-names @gol
186 -fno-optional-diags -fpermissive @gol
187 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
188 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
189 -fno-default-inline -fvisibility-inlines-hidden @gol
190 -Wabi -Wctor-dtor-privacy @gol
191 -Wnon-virtual-dtor -Wreorder @gol
192 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
193 -Wno-non-template-friend -Wold-style-cast @gol
194 -Woverloaded-virtual -Wno-pmf-conversions @gol
197 @item Objective-C and Objective-C++ Language Options
198 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
199 Objective-C and Objective-C++ Dialects}.
201 -fconstant-string-class=@var{class-name} @gol
202 -fgnu-runtime -fnext-runtime @gol
203 -fno-nil-receivers @gol
204 -fobjc-call-cxx-cdtors @gol
205 -fobjc-direct-dispatch @gol
206 -fobjc-exceptions @gol
208 -freplace-objc-classes @gol
211 -Wassign-intercept @gol
212 -Wno-protocol -Wselector @gol
213 -Wstrict-selector-match @gol
214 -Wundeclared-selector}
216 @item Language Independent Options
217 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
218 @gccoptlist{-fmessage-length=@var{n} @gol
219 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
220 -fdiagnostics-show-options
222 @item Warning Options
223 @xref{Warning Options,,Options to Request or Suppress Warnings}.
224 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
225 -w -Wextra -Wall -Waggregate-return -Walways-true -Wno-attributes @gol
226 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
227 -Wconversion -Wno-deprecated-declarations @gol
228 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
229 -Werror -Werror-implicit-function-declaration @gol
230 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
231 -Wno-format-extra-args -Wformat-nonliteral @gol
232 -Wformat-security -Wformat-y2k @gol
233 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
234 -Wimport -Wno-import -Winit-self -Winline @gol
235 -Wno-int-to-pointer-cast @gol
236 -Wno-invalid-offsetof -Winvalid-pch @gol
237 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
238 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
239 -Wmissing-format-attribute -Wmissing-include-dirs @gol
240 -Wmissing-noreturn @gol
241 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
242 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
243 -Wredundant-decls @gol
244 -Wreturn-type -Wsequence-point -Wshadow @gol
245 -Wsign-compare -Wstack-protector @gol
246 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
247 -Wstring-literal-comparison @gol
248 -Wswitch -Wswitch-default -Wswitch-enum @gol
249 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
250 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
251 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
252 -Wunused-value -Wunused-variable -Wvariadic-macros @gol
253 -Wvolatile-register-var -Wwrite-strings}
255 @item C-only Warning Options
256 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
257 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
258 -Wstrict-prototypes -Wtraditional @gol
259 -Wdeclaration-after-statement -Wno-pointer-sign}
261 @item Debugging Options
262 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
263 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
264 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
265 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
266 -fdump-ipa-all -fdump-ipa-cgraph @gol
268 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
273 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
277 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-nrv -fdump-tree-vect @gol
282 -fdump-tree-sink @gol
283 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-salias @gol
285 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
287 -ftree-vectorizer-verbose=@var{n} @gol
288 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
289 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
290 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
291 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
292 -ftest-coverage -ftime-report -fvar-tracking @gol
293 -g -g@var{level} -gcoff -gdwarf-2 @gol
294 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
295 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
296 -print-multi-directory -print-multi-lib @gol
297 -print-prog-name=@var{program} -print-search-dirs -Q @gol
300 @item Optimization Options
301 @xref{Optimize Options,,Options that Control Optimization}.
302 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
303 -falign-labels=@var{n} -falign-loops=@var{n} @gol
304 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
305 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
306 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
307 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
308 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
309 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
310 -fexpensive-optimizations -ffast-math -ffloat-store @gol
311 -fforce-addr -ffunction-sections @gol
312 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
313 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
314 -finline-functions -finline-functions-called-once @gol
315 -finline-limit=@var{n} -fkeep-inline-functions @gol
316 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
317 -fmodulo-sched -fno-branch-count-reg @gol
318 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
319 -fno-function-cse -fno-guess-branch-probability @gol
320 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
321 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
322 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
323 -fomit-frame-pointer -foptimize-register-move @gol
324 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
325 -fprofile-generate -fprofile-use @gol
326 -fregmove -frename-registers @gol
327 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
328 -frerun-cse-after-loop -frerun-loop-opt @gol
329 -frounding-math -frtl-abstract-sequences @gol
330 -fschedule-insns -fschedule-insns2 @gol
331 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
332 -fsched-spec-load-dangerous @gol
333 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
334 -fsched2-use-superblocks @gol
335 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
336 -fsignaling-nans -fsingle-precision-constant @gol
337 -fstack-protector -fstack-protector-all @gol
338 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
339 -funroll-all-loops -funroll-loops -fpeel-loops @gol
340 -fsplit-ivs-in-unroller -funswitch-loops @gol
341 -fvariable-expansion-in-unroller @gol
342 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
343 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
344 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
345 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
346 -ftree-vect-loop-version -ftree-salias -fweb @gol
347 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
348 --param @var{name}=@var{value}
349 -O -O0 -O1 -O2 -O3 -Os}
351 @item Preprocessor Options
352 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
353 @gccoptlist{-A@var{question}=@var{answer} @gol
354 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
355 -C -dD -dI -dM -dN @gol
356 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
357 -idirafter @var{dir} @gol
358 -include @var{file} -imacros @var{file} @gol
359 -iprefix @var{file} -iwithprefix @var{dir} @gol
360 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
361 -isysroot @var{dir} @gol
362 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
363 -P -fworking-directory -remap @gol
364 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
365 -Xpreprocessor @var{option}}
367 @item Assembler Option
368 @xref{Assembler Options,,Passing Options to the Assembler}.
369 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
372 @xref{Link Options,,Options for Linking}.
373 @gccoptlist{@var{object-file-name} -l@var{library} @gol
374 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
375 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
376 -Wl,@var{option} -Xlinker @var{option} @gol
379 @item Directory Options
380 @xref{Directory Options,,Options for Directory Search}.
381 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
382 -specs=@var{file} -I- --sysroot=@var{dir}}
385 @c I wrote this xref this way to avoid overfull hbox. -- rms
386 @xref{Target Options}.
387 @gccoptlist{-V @var{version} -b @var{machine}}
389 @item Machine Dependent Options
390 @xref{Submodel Options,,Hardware Models and Configurations}.
391 @c This list is ordered alphanumerically by subsection name.
392 @c Try and put the significant identifier (CPU or system) first,
393 @c so users have a clue at guessing where the ones they want will be.
396 @gccoptlist{-EB -EL @gol
397 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
398 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
401 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
402 -mabi=@var{name} @gol
403 -mapcs-stack-check -mno-apcs-stack-check @gol
404 -mapcs-float -mno-apcs-float @gol
405 -mapcs-reentrant -mno-apcs-reentrant @gol
406 -msched-prolog -mno-sched-prolog @gol
407 -mlittle-endian -mbig-endian -mwords-little-endian @gol
408 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
409 -mthumb-interwork -mno-thumb-interwork @gol
410 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
411 -mstructure-size-boundary=@var{n} @gol
412 -mabort-on-noreturn @gol
413 -mlong-calls -mno-long-calls @gol
414 -msingle-pic-base -mno-single-pic-base @gol
415 -mpic-register=@var{reg} @gol
416 -mnop-fun-dllimport @gol
417 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
418 -mpoke-function-name @gol
420 -mtpcs-frame -mtpcs-leaf-frame @gol
421 -mcaller-super-interworking -mcallee-super-interworking @gol
425 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
426 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
428 @emph{Blackfin Options}
429 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
430 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
431 -mlow-64k -mno-low64k -mid-shared-library @gol
432 -mno-id-shared-library -mshared-library-id=@var{n} @gol
433 -mlong-calls -mno-long-calls}
436 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
437 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
438 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
439 -mstack-align -mdata-align -mconst-align @gol
440 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
441 -melf -maout -melinux -mlinux -sim -sim2 @gol
442 -mmul-bug-workaround -mno-mul-bug-workaround}
445 @gccoptlist{-mmac -mpush-args}
447 @emph{Darwin Options}
448 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
449 -arch_only -bind_at_load -bundle -bundle_loader @gol
450 -client_name -compatibility_version -current_version @gol
452 -dependency-file -dylib_file -dylinker_install_name @gol
453 -dynamic -dynamiclib -exported_symbols_list @gol
454 -filelist -flat_namespace -force_cpusubtype_ALL @gol
455 -force_flat_namespace -headerpad_max_install_names @gol
456 -image_base -init -install_name -keep_private_externs @gol
457 -multi_module -multiply_defined -multiply_defined_unused @gol
458 -noall_load -no_dead_strip_inits_and_terms @gol
459 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
460 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
461 -private_bundle -read_only_relocs -sectalign @gol
462 -sectobjectsymbols -whyload -seg1addr @gol
463 -sectcreate -sectobjectsymbols -sectorder @gol
464 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
465 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
466 -segprot -segs_read_only_addr -segs_read_write_addr @gol
467 -single_module -static -sub_library -sub_umbrella @gol
468 -twolevel_namespace -umbrella -undefined @gol
469 -unexported_symbols_list -weak_reference_mismatches @gol
470 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
473 @emph{DEC Alpha Options}
474 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
475 -mieee -mieee-with-inexact -mieee-conformant @gol
476 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
477 -mtrap-precision=@var{mode} -mbuild-constants @gol
478 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
479 -mbwx -mmax -mfix -mcix @gol
480 -mfloat-vax -mfloat-ieee @gol
481 -mexplicit-relocs -msmall-data -mlarge-data @gol
482 -msmall-text -mlarge-text @gol
483 -mmemory-latency=@var{time}}
485 @emph{DEC Alpha/VMS Options}
486 @gccoptlist{-mvms-return-codes}
489 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
490 -mhard-float -msoft-float @gol
491 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
492 -mdouble -mno-double @gol
493 -mmedia -mno-media -mmuladd -mno-muladd @gol
494 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
495 -mlinked-fp -mlong-calls -malign-labels @gol
496 -mlibrary-pic -macc-4 -macc-8 @gol
497 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
498 -moptimize-membar -mno-optimize-membar @gol
499 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
500 -mvliw-branch -mno-vliw-branch @gol
501 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
502 -mno-nested-cond-exec -mtomcat-stats @gol
506 @emph{H8/300 Options}
507 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
510 @gccoptlist{-march=@var{architecture-type} @gol
511 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
512 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
513 -mfixed-range=@var{register-range} @gol
514 -mjump-in-delay -mlinker-opt -mlong-calls @gol
515 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
516 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
517 -mno-jump-in-delay -mno-long-load-store @gol
518 -mno-portable-runtime -mno-soft-float @gol
519 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
520 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
521 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
522 -munix=@var{unix-std} -nolibdld -static -threads}
524 @emph{i386 and x86-64 Options}
525 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
526 -mfpmath=@var{unit} @gol
527 -masm=@var{dialect} -mno-fancy-math-387 @gol
528 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
529 -mno-wide-multiply -mrtd -malign-double @gol
530 -mpreferred-stack-boundary=@var{num} @gol
531 -mmmx -msse -msse2 -msse3 -m3dnow @gol
532 -mthreads -mno-align-stringops -minline-all-stringops @gol
533 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
534 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
535 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
536 -mcmodel=@var{code-model} @gol
537 -m32 -m64 -mlarge-data-threshold=@var{num}}
540 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
541 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
542 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
543 -minline-float-divide-max-throughput @gol
544 -minline-int-divide-min-latency @gol
545 -minline-int-divide-max-throughput @gol
546 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
547 -mno-dwarf2-asm -mearly-stop-bits @gol
548 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
549 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
551 @emph{M32R/D Options}
552 @gccoptlist{-m32r2 -m32rx -m32r @gol
554 -malign-loops -mno-align-loops @gol
555 -missue-rate=@var{number} @gol
556 -mbranch-cost=@var{number} @gol
557 -mmodel=@var{code-size-model-type} @gol
558 -msdata=@var{sdata-type} @gol
559 -mno-flush-func -mflush-func=@var{name} @gol
560 -mno-flush-trap -mflush-trap=@var{number} @gol
564 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
566 @emph{M680x0 Options}
567 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
568 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
569 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
570 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
571 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
573 @emph{M68hc1x Options}
574 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
575 -mauto-incdec -minmax -mlong-calls -mshort @gol
576 -msoft-reg-count=@var{count}}
579 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
580 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
581 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
582 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
583 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
586 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
587 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
588 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
589 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
590 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
591 -mdsp -mpaired-single -mips3d @gol
592 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
593 -G@var{num} -membedded-data -mno-embedded-data @gol
594 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
595 -msplit-addresses -mno-split-addresses @gol
596 -mexplicit-relocs -mno-explicit-relocs @gol
597 -mcheck-zero-division -mno-check-zero-division @gol
598 -mdivide-traps -mdivide-breaks @gol
599 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
600 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
601 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
602 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
603 -mfix-sb1 -mno-fix-sb1 @gol
604 -mflush-func=@var{func} -mno-flush-func @gol
605 -mbranch-likely -mno-branch-likely @gol
606 -mfp-exceptions -mno-fp-exceptions @gol
607 -mvr4130-align -mno-vr4130-align}
610 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
611 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
612 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
613 -mno-base-addresses -msingle-exit -mno-single-exit}
615 @emph{MN10300 Options}
616 @gccoptlist{-mmult-bug -mno-mult-bug @gol
617 -mam33 -mno-am33 @gol
618 -mam33-2 -mno-am33-2 @gol
619 -mreturn-pointer-on-d0 @gol
623 @gccoptlist{-mno-crt0 -mbacc -msim @gol
624 -march=@var{cpu-type} }
626 @emph{PDP-11 Options}
627 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
628 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
629 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
630 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
631 -mbranch-expensive -mbranch-cheap @gol
632 -msplit -mno-split -munix-asm -mdec-asm}
634 @emph{PowerPC Options}
635 See RS/6000 and PowerPC Options.
637 @emph{RS/6000 and PowerPC Options}
638 @gccoptlist{-mcpu=@var{cpu-type} @gol
639 -mtune=@var{cpu-type} @gol
640 -mpower -mno-power -mpower2 -mno-power2 @gol
641 -mpowerpc -mpowerpc64 -mno-powerpc @gol
642 -maltivec -mno-altivec @gol
643 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
644 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
645 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
646 -mnew-mnemonics -mold-mnemonics @gol
647 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
648 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
649 -malign-power -malign-natural @gol
650 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
651 -mstring -mno-string -mupdate -mno-update @gol
652 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
653 -mstrict-align -mno-strict-align -mrelocatable @gol
654 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
655 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
656 -mdynamic-no-pic -maltivec -mswdiv @gol
657 -mprioritize-restricted-insns=@var{priority} @gol
658 -msched-costly-dep=@var{dependence_type} @gol
659 -minsert-sched-nops=@var{scheme} @gol
660 -mcall-sysv -mcall-netbsd @gol
661 -maix-struct-return -msvr4-struct-return @gol
662 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
663 -misel -mno-isel @gol
664 -misel=yes -misel=no @gol
666 -mspe=yes -mspe=no @gol
667 -mvrsave -mno-vrsave @gol
668 -mmulhw -mno-mulhw @gol
669 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
670 -mprototype -mno-prototype @gol
671 -msim -mmvme -mads -myellowknife -memb -msdata @gol
672 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
674 @emph{S/390 and zSeries Options}
675 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
676 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
677 -mpacked-stack -mno-packed-stack @gol
678 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
679 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
680 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
681 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
684 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
685 -m4-nofpu -m4-single-only -m4-single -m4 @gol
686 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
687 -m5-64media -m5-64media-nofpu @gol
688 -m5-32media -m5-32media-nofpu @gol
689 -m5-compact -m5-compact-nofpu @gol
690 -mb -ml -mdalign -mrelax @gol
691 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
692 -mieee -misize -mpadstruct -mspace @gol
693 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
694 -mdivsi3_libfunc=@var{name} @gol
695 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
699 @gccoptlist{-mcpu=@var{cpu-type} @gol
700 -mtune=@var{cpu-type} @gol
701 -mcmodel=@var{code-model} @gol
702 -m32 -m64 -mapp-regs -mno-app-regs @gol
703 -mfaster-structs -mno-faster-structs @gol
704 -mfpu -mno-fpu -mhard-float -msoft-float @gol
705 -mhard-quad-float -msoft-quad-float @gol
706 -mimpure-text -mno-impure-text -mlittle-endian @gol
707 -mstack-bias -mno-stack-bias @gol
708 -munaligned-doubles -mno-unaligned-doubles @gol
709 -mv8plus -mno-v8plus -mvis -mno-vis
712 @emph{System V Options}
713 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
715 @emph{TMS320C3x/C4x Options}
716 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
717 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
718 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
719 -mparallel-insns -mparallel-mpy -mpreserve-float}
722 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
723 -mprolog-function -mno-prolog-function -mspace @gol
724 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
725 -mapp-regs -mno-app-regs @gol
726 -mdisable-callt -mno-disable-callt @gol
732 @gccoptlist{-mg -mgnu -munix}
734 @emph{x86-64 Options}
735 See i386 and x86-64 Options.
737 @emph{Xstormy16 Options}
740 @emph{Xtensa Options}
741 @gccoptlist{-mconst16 -mno-const16 @gol
742 -mfused-madd -mno-fused-madd @gol
743 -mtext-section-literals -mno-text-section-literals @gol
744 -mtarget-align -mno-target-align @gol
745 -mlongcalls -mno-longcalls}
747 @emph{zSeries Options}
748 See S/390 and zSeries Options.
750 @item Code Generation Options
751 @xref{Code Gen Options,,Options for Code Generation Conventions}.
752 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
753 -ffixed-@var{reg} -fexceptions @gol
754 -fnon-call-exceptions -funwind-tables @gol
755 -fasynchronous-unwind-tables @gol
756 -finhibit-size-directive -finstrument-functions @gol
757 -fno-common -fno-ident @gol
758 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
759 -fno-jump-tables @gol
760 -freg-struct-return -fshared-data -fshort-enums @gol
761 -fshort-double -fshort-wchar @gol
762 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
763 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
764 -fargument-alias -fargument-noalias @gol
765 -fargument-noalias-global -fleading-underscore @gol
766 -ftls-model=@var{model} @gol
767 -ftrapv -fwrapv -fbounds-check @gol
768 -fvisibility -fopenmp}
772 * Overall Options:: Controlling the kind of output:
773 an executable, object files, assembler files,
774 or preprocessed source.
775 * C Dialect Options:: Controlling the variant of C language compiled.
776 * C++ Dialect Options:: Variations on C++.
777 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
779 * Language Independent Options:: Controlling how diagnostics should be
781 * Warning Options:: How picky should the compiler be?
782 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
783 * Optimize Options:: How much optimization?
784 * Preprocessor Options:: Controlling header files and macro definitions.
785 Also, getting dependency information for Make.
786 * Assembler Options:: Passing options to the assembler.
787 * Link Options:: Specifying libraries and so on.
788 * Directory Options:: Where to find header files and libraries.
789 Where to find the compiler executable files.
790 * Spec Files:: How to pass switches to sub-processes.
791 * Target Options:: Running a cross-compiler, or an old version of GCC.
794 @node Overall Options
795 @section Options Controlling the Kind of Output
797 Compilation can involve up to four stages: preprocessing, compilation
798 proper, assembly and linking, always in that order. GCC is capable of
799 preprocessing and compiling several files either into several
800 assembler input files, or into one assembler input file; then each
801 assembler input file produces an object file, and linking combines all
802 the object files (those newly compiled, and those specified as input)
803 into an executable file.
805 @cindex file name suffix
806 For any given input file, the file name suffix determines what kind of
811 C source code which must be preprocessed.
814 C source code which should not be preprocessed.
817 C++ source code which should not be preprocessed.
820 Objective-C source code. Note that you must link with the @file{libobjc}
821 library to make an Objective-C program work.
824 Objective-C source code which should not be preprocessed.
828 Objective-C++ source code. Note that you must link with the @file{libobjc}
829 library to make an Objective-C++ program work. Note that @samp{.M} refers
830 to a literal capital M@.
833 Objective-C++ source code which should not be preprocessed.
836 C, C++, Objective-C or Objective-C++ header file to be turned into a
841 @itemx @var{file}.cxx
842 @itemx @var{file}.cpp
843 @itemx @var{file}.CPP
844 @itemx @var{file}.c++
846 C++ source code which must be preprocessed. Note that in @samp{.cxx},
847 the last two letters must both be literally @samp{x}. Likewise,
848 @samp{.C} refers to a literal capital C@.
852 Objective-C++ source code which must be preprocessed.
855 Objective-C++ source code which should not be preprocessed.
859 C++ header file to be turned into a precompiled header.
862 @itemx @var{file}.for
863 @itemx @var{file}.FOR
864 Fixed form Fortran source code which should not be preprocessed.
867 @itemx @var{file}.fpp
868 @itemx @var{file}.FPP
869 Fixed form Fortran source code which must be preprocessed (with the traditional
873 @itemx @var{file}.f95
874 Free form Fortran source code which should not be preprocessed.
877 @itemx @var{file}.F95
878 Free form Fortran source code which must be preprocessed (with the
879 traditional preprocessor).
881 @c FIXME: Descriptions of Java file types.
888 Ada source code file which contains a library unit declaration (a
889 declaration of a package, subprogram, or generic, or a generic
890 instantiation), or a library unit renaming declaration (a package,
891 generic, or subprogram renaming declaration). Such files are also
894 @itemx @var{file}.adb
895 Ada source code file containing a library unit body (a subprogram or
896 package body). Such files are also called @dfn{bodies}.
898 @c GCC also knows about some suffixes for languages not yet included:
909 Assembler code which must be preprocessed.
912 An object file to be fed straight into linking.
913 Any file name with no recognized suffix is treated this way.
917 You can specify the input language explicitly with the @option{-x} option:
920 @item -x @var{language}
921 Specify explicitly the @var{language} for the following input files
922 (rather than letting the compiler choose a default based on the file
923 name suffix). This option applies to all following input files until
924 the next @option{-x} option. Possible values for @var{language} are:
926 c c-header c-cpp-output
927 c++ c++-header c++-cpp-output
928 objective-c objective-c-header objective-c-cpp-output
929 objective-c++ objective-c++-header objective-c++-cpp-output
930 assembler assembler-with-cpp
939 Turn off any specification of a language, so that subsequent files are
940 handled according to their file name suffixes (as they are if @option{-x}
941 has not been used at all).
943 @item -pass-exit-codes
944 @opindex pass-exit-codes
945 Normally the @command{gcc} program will exit with the code of 1 if any
946 phase of the compiler returns a non-success return code. If you specify
947 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
948 numerically highest error produced by any phase that returned an error
952 If you only want some of the stages of compilation, you can use
953 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
954 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
955 @command{gcc} is to stop. Note that some combinations (for example,
956 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
961 Compile or assemble the source files, but do not link. The linking
962 stage simply is not done. The ultimate output is in the form of an
963 object file for each source file.
965 By default, the object file name for a source file is made by replacing
966 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
968 Unrecognized input files, not requiring compilation or assembly, are
973 Stop after the stage of compilation proper; do not assemble. The output
974 is in the form of an assembler code file for each non-assembler input
977 By default, the assembler file name for a source file is made by
978 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
980 Input files that don't require compilation are ignored.
984 Stop after the preprocessing stage; do not run the compiler proper. The
985 output is in the form of preprocessed source code, which is sent to the
988 Input files which don't require preprocessing are ignored.
990 @cindex output file option
993 Place output in file @var{file}. This applies regardless to whatever
994 sort of output is being produced, whether it be an executable file,
995 an object file, an assembler file or preprocessed C code.
997 If @option{-o} is not specified, the default is to put an executable
998 file in @file{a.out}, the object file for
999 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1000 assembler file in @file{@var{source}.s}, a precompiled header file in
1001 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1006 Print (on standard error output) the commands executed to run the stages
1007 of compilation. Also print the version number of the compiler driver
1008 program and of the preprocessor and the compiler proper.
1012 Like @option{-v} except the commands are not executed and all command
1013 arguments are quoted. This is useful for shell scripts to capture the
1014 driver-generated command lines.
1018 Use pipes rather than temporary files for communication between the
1019 various stages of compilation. This fails to work on some systems where
1020 the assembler is unable to read from a pipe; but the GNU assembler has
1025 If you are compiling multiple source files, this option tells the driver
1026 to pass all the source files to the compiler at once (for those
1027 languages for which the compiler can handle this). This will allow
1028 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1029 language for which this is supported is C@. If you pass source files for
1030 multiple languages to the driver, using this option, the driver will invoke
1031 the compiler(s) that support IMA once each, passing each compiler all the
1032 source files appropriate for it. For those languages that do not support
1033 IMA this option will be ignored, and the compiler will be invoked once for
1034 each source file in that language. If you use this option in conjunction
1035 with @option{-save-temps}, the compiler will generate multiple
1037 (one for each source file), but only one (combined) @file{.o} or
1042 Print (on the standard output) a description of the command line options
1043 understood by @command{gcc}. If the @option{-v} option is also specified
1044 then @option{--help} will also be passed on to the various processes
1045 invoked by @command{gcc}, so that they can display the command line options
1046 they accept. If the @option{-Wextra} option is also specified then command
1047 line options which have no documentation associated with them will also
1051 @opindex target-help
1052 Print (on the standard output) a description of target specific command
1053 line options for each tool.
1057 Display the version number and copyrights of the invoked GCC@.
1059 @include @value{srcdir}/../libiberty/at-file.texi
1063 @section Compiling C++ Programs
1065 @cindex suffixes for C++ source
1066 @cindex C++ source file suffixes
1067 C++ source files conventionally use one of the suffixes @samp{.C},
1068 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1069 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1070 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1071 files with these names and compiles them as C++ programs even if you
1072 call the compiler the same way as for compiling C programs (usually
1073 with the name @command{gcc}).
1077 However, C++ programs often require class libraries as well as a
1078 compiler that understands the C++ language---and under some
1079 circumstances, you might want to compile programs or header files from
1080 standard input, or otherwise without a suffix that flags them as C++
1081 programs. You might also like to precompile a C header file with a
1082 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1083 program that calls GCC with the default language set to C++, and
1084 automatically specifies linking against the C++ library. On many
1085 systems, @command{g++} is also installed with the name @command{c++}.
1087 @cindex invoking @command{g++}
1088 When you compile C++ programs, you may specify many of the same
1089 command-line options that you use for compiling programs in any
1090 language; or command-line options meaningful for C and related
1091 languages; or options that are meaningful only for C++ programs.
1092 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1093 explanations of options for languages related to C@.
1094 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1095 explanations of options that are meaningful only for C++ programs.
1097 @node C Dialect Options
1098 @section Options Controlling C Dialect
1099 @cindex dialect options
1100 @cindex language dialect options
1101 @cindex options, dialect
1103 The following options control the dialect of C (or languages derived
1104 from C, such as C++, Objective-C and Objective-C++) that the compiler
1108 @cindex ANSI support
1112 In C mode, support all ISO C90 programs. In C++ mode,
1113 remove GNU extensions that conflict with ISO C++.
1115 This turns off certain features of GCC that are incompatible with ISO
1116 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1117 such as the @code{asm} and @code{typeof} keywords, and
1118 predefined macros such as @code{unix} and @code{vax} that identify the
1119 type of system you are using. It also enables the undesirable and
1120 rarely used ISO trigraph feature. For the C compiler,
1121 it disables recognition of C++ style @samp{//} comments as well as
1122 the @code{inline} keyword.
1124 The alternate keywords @code{__asm__}, @code{__extension__},
1125 @code{__inline__} and @code{__typeof__} continue to work despite
1126 @option{-ansi}. You would not want to use them in an ISO C program, of
1127 course, but it is useful to put them in header files that might be included
1128 in compilations done with @option{-ansi}. Alternate predefined macros
1129 such as @code{__unix__} and @code{__vax__} are also available, with or
1130 without @option{-ansi}.
1132 The @option{-ansi} option does not cause non-ISO programs to be
1133 rejected gratuitously. For that, @option{-pedantic} is required in
1134 addition to @option{-ansi}. @xref{Warning Options}.
1136 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1137 option is used. Some header files may notice this macro and refrain
1138 from declaring certain functions or defining certain macros that the
1139 ISO standard doesn't call for; this is to avoid interfering with any
1140 programs that might use these names for other things.
1142 Functions which would normally be built in but do not have semantics
1143 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1144 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1145 built-in functions provided by GCC}, for details of the functions
1150 Determine the language standard. This option is currently only
1151 supported when compiling C or C++. A value for this option must be
1152 provided; possible values are
1157 ISO C90 (same as @option{-ansi}).
1159 @item iso9899:199409
1160 ISO C90 as modified in amendment 1.
1166 ISO C99. Note that this standard is not yet fully supported; see
1167 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1168 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1171 Default, ISO C90 plus GNU extensions (including some C99 features).
1175 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1176 this will become the default. The name @samp{gnu9x} is deprecated.
1179 The 1998 ISO C++ standard plus amendments.
1182 The same as @option{-std=c++98} plus GNU extensions. This is the
1183 default for C++ code.
1186 Even when this option is not specified, you can still use some of the
1187 features of newer standards in so far as they do not conflict with
1188 previous C standards. For example, you may use @code{__restrict__} even
1189 when @option{-std=c99} is not specified.
1191 The @option{-std} options specifying some version of ISO C have the same
1192 effects as @option{-ansi}, except that features that were not in ISO C90
1193 but are in the specified version (for example, @samp{//} comments and
1194 the @code{inline} keyword in ISO C99) are not disabled.
1196 @xref{Standards,,Language Standards Supported by GCC}, for details of
1197 these standard versions.
1199 @item -aux-info @var{filename}
1201 Output to the given filename prototyped declarations for all functions
1202 declared and/or defined in a translation unit, including those in header
1203 files. This option is silently ignored in any language other than C@.
1205 Besides declarations, the file indicates, in comments, the origin of
1206 each declaration (source file and line), whether the declaration was
1207 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1208 @samp{O} for old, respectively, in the first character after the line
1209 number and the colon), and whether it came from a declaration or a
1210 definition (@samp{C} or @samp{F}, respectively, in the following
1211 character). In the case of function definitions, a K&R-style list of
1212 arguments followed by their declarations is also provided, inside
1213 comments, after the declaration.
1217 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1218 keyword, so that code can use these words as identifiers. You can use
1219 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1220 instead. @option{-ansi} implies @option{-fno-asm}.
1222 In C++, this switch only affects the @code{typeof} keyword, since
1223 @code{asm} and @code{inline} are standard keywords. You may want to
1224 use the @option{-fno-gnu-keywords} flag instead, which has the same
1225 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1226 switch only affects the @code{asm} and @code{typeof} keywords, since
1227 @code{inline} is a standard keyword in ISO C99.
1230 @itemx -fno-builtin-@var{function}
1231 @opindex fno-builtin
1232 @cindex built-in functions
1233 Don't recognize built-in functions that do not begin with
1234 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1235 functions provided by GCC}, for details of the functions affected,
1236 including those which are not built-in functions when @option{-ansi} or
1237 @option{-std} options for strict ISO C conformance are used because they
1238 do not have an ISO standard meaning.
1240 GCC normally generates special code to handle certain built-in functions
1241 more efficiently; for instance, calls to @code{alloca} may become single
1242 instructions that adjust the stack directly, and calls to @code{memcpy}
1243 may become inline copy loops. The resulting code is often both smaller
1244 and faster, but since the function calls no longer appear as such, you
1245 cannot set a breakpoint on those calls, nor can you change the behavior
1246 of the functions by linking with a different library. In addition,
1247 when a function is recognized as a built-in function, GCC may use
1248 information about that function to warn about problems with calls to
1249 that function, or to generate more efficient code, even if the
1250 resulting code still contains calls to that function. For example,
1251 warnings are given with @option{-Wformat} for bad calls to
1252 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1253 known not to modify global memory.
1255 With the @option{-fno-builtin-@var{function}} option
1256 only the built-in function @var{function} is
1257 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1258 function is named this is not built-in in this version of GCC, this
1259 option is ignored. There is no corresponding
1260 @option{-fbuiltin-@var{function}} option; if you wish to enable
1261 built-in functions selectively when using @option{-fno-builtin} or
1262 @option{-ffreestanding}, you may define macros such as:
1265 #define abs(n) __builtin_abs ((n))
1266 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1271 @cindex hosted environment
1273 Assert that compilation takes place in a hosted environment. This implies
1274 @option{-fbuiltin}. A hosted environment is one in which the
1275 entire standard library is available, and in which @code{main} has a return
1276 type of @code{int}. Examples are nearly everything except a kernel.
1277 This is equivalent to @option{-fno-freestanding}.
1279 @item -ffreestanding
1280 @opindex ffreestanding
1281 @cindex hosted environment
1283 Assert that compilation takes place in a freestanding environment. This
1284 implies @option{-fno-builtin}. A freestanding environment
1285 is one in which the standard library may not exist, and program startup may
1286 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1287 This is equivalent to @option{-fno-hosted}.
1289 @xref{Standards,,Language Standards Supported by GCC}, for details of
1290 freestanding and hosted environments.
1292 @item -fms-extensions
1293 @opindex fms-extensions
1294 Accept some non-standard constructs used in Microsoft header files.
1296 Some cases of unnamed fields in structures and unions are only
1297 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1298 fields within structs/unions}, for details.
1302 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1303 options for strict ISO C conformance) implies @option{-trigraphs}.
1305 @item -no-integrated-cpp
1306 @opindex no-integrated-cpp
1307 Performs a compilation in two passes: preprocessing and compiling. This
1308 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1309 @option{-B} option. The user supplied compilation step can then add in
1310 an additional preprocessing step after normal preprocessing but before
1311 compiling. The default is to use the integrated cpp (internal cpp)
1313 The semantics of this option will change if "cc1", "cc1plus", and
1314 "cc1obj" are merged.
1316 @cindex traditional C language
1317 @cindex C language, traditional
1319 @itemx -traditional-cpp
1320 @opindex traditional-cpp
1321 @opindex traditional
1322 Formerly, these options caused GCC to attempt to emulate a pre-standard
1323 C compiler. They are now only supported with the @option{-E} switch.
1324 The preprocessor continues to support a pre-standard mode. See the GNU
1325 CPP manual for details.
1327 @item -fcond-mismatch
1328 @opindex fcond-mismatch
1329 Allow conditional expressions with mismatched types in the second and
1330 third arguments. The value of such an expression is void. This option
1331 is not supported for C++.
1333 @item -funsigned-char
1334 @opindex funsigned-char
1335 Let the type @code{char} be unsigned, like @code{unsigned char}.
1337 Each kind of machine has a default for what @code{char} should
1338 be. It is either like @code{unsigned char} by default or like
1339 @code{signed char} by default.
1341 Ideally, a portable program should always use @code{signed char} or
1342 @code{unsigned char} when it depends on the signedness of an object.
1343 But many programs have been written to use plain @code{char} and
1344 expect it to be signed, or expect it to be unsigned, depending on the
1345 machines they were written for. This option, and its inverse, let you
1346 make such a program work with the opposite default.
1348 The type @code{char} is always a distinct type from each of
1349 @code{signed char} or @code{unsigned char}, even though its behavior
1350 is always just like one of those two.
1353 @opindex fsigned-char
1354 Let the type @code{char} be signed, like @code{signed char}.
1356 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1357 the negative form of @option{-funsigned-char}. Likewise, the option
1358 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1360 @item -fsigned-bitfields
1361 @itemx -funsigned-bitfields
1362 @itemx -fno-signed-bitfields
1363 @itemx -fno-unsigned-bitfields
1364 @opindex fsigned-bitfields
1365 @opindex funsigned-bitfields
1366 @opindex fno-signed-bitfields
1367 @opindex fno-unsigned-bitfields
1368 These options control whether a bit-field is signed or unsigned, when the
1369 declaration does not use either @code{signed} or @code{unsigned}. By
1370 default, such a bit-field is signed, because this is consistent: the
1371 basic integer types such as @code{int} are signed types.
1374 @node C++ Dialect Options
1375 @section Options Controlling C++ Dialect
1377 @cindex compiler options, C++
1378 @cindex C++ options, command line
1379 @cindex options, C++
1380 This section describes the command-line options that are only meaningful
1381 for C++ programs; but you can also use most of the GNU compiler options
1382 regardless of what language your program is in. For example, you
1383 might compile a file @code{firstClass.C} like this:
1386 g++ -g -frepo -O -c firstClass.C
1390 In this example, only @option{-frepo} is an option meant
1391 only for C++ programs; you can use the other options with any
1392 language supported by GCC@.
1394 Here is a list of options that are @emph{only} for compiling C++ programs:
1398 @item -fabi-version=@var{n}
1399 @opindex fabi-version
1400 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1401 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1402 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1403 the version that conforms most closely to the C++ ABI specification.
1404 Therefore, the ABI obtained using version 0 will change as ABI bugs
1407 The default is version 2.
1409 @item -fno-access-control
1410 @opindex fno-access-control
1411 Turn off all access checking. This switch is mainly useful for working
1412 around bugs in the access control code.
1416 Check that the pointer returned by @code{operator new} is non-null
1417 before attempting to modify the storage allocated. This check is
1418 normally unnecessary because the C++ standard specifies that
1419 @code{operator new} will only return @code{0} if it is declared
1420 @samp{throw()}, in which case the compiler will always check the
1421 return value even without this option. In all other cases, when
1422 @code{operator new} has a non-empty exception specification, memory
1423 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1424 @samp{new (nothrow)}.
1426 @item -fconserve-space
1427 @opindex fconserve-space
1428 Put uninitialized or runtime-initialized global variables into the
1429 common segment, as C does. This saves space in the executable at the
1430 cost of not diagnosing duplicate definitions. If you compile with this
1431 flag and your program mysteriously crashes after @code{main()} has
1432 completed, you may have an object that is being destroyed twice because
1433 two definitions were merged.
1435 This option is no longer useful on most targets, now that support has
1436 been added for putting variables into BSS without making them common.
1438 @item -ffriend-injection
1439 @opindex ffriend-injection
1440 Inject friend functions into the enclosing namespace, so that they are
1441 visible outside the scope of the class in which they are declared.
1442 Friend functions were documented to work this way in the old Annotated
1443 C++ Reference Manual, and versions of G++ before 4.1 always worked
1444 that way. However, in ISO C++ a friend function which is not declared
1445 in an enclosing scope can only be found using argument dependent
1446 lookup. This option causes friends to be injected as they were in
1449 This option is for compatibility, and may be removed in a future
1452 @item -fno-const-strings
1453 @opindex fno-const-strings
1454 Give string constants type @code{char *} instead of type @code{const
1455 char *}. By default, G++ uses type @code{const char *} as required by
1456 the standard. Even if you use @option{-fno-const-strings}, you cannot
1457 actually modify the value of a string constant.
1459 This option might be removed in a future release of G++. For maximum
1460 portability, you should structure your code so that it works with
1461 string constants that have type @code{const char *}.
1463 @item -fno-elide-constructors
1464 @opindex fno-elide-constructors
1465 The C++ standard allows an implementation to omit creating a temporary
1466 which is only used to initialize another object of the same type.
1467 Specifying this option disables that optimization, and forces G++ to
1468 call the copy constructor in all cases.
1470 @item -fno-enforce-eh-specs
1471 @opindex fno-enforce-eh-specs
1472 Don't generate code to check for violation of exception specifications
1473 at runtime. This option violates the C++ standard, but may be useful
1474 for reducing code size in production builds, much like defining
1475 @samp{NDEBUG}. This does not give user code permission to throw
1476 exceptions in violation of the exception specifications; the compiler
1477 will still optimize based on the specifications, so throwing an
1478 unexpected exception will result in undefined behavior.
1481 @itemx -fno-for-scope
1483 @opindex fno-for-scope
1484 If @option{-ffor-scope} is specified, the scope of variables declared in
1485 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1486 as specified by the C++ standard.
1487 If @option{-fno-for-scope} is specified, the scope of variables declared in
1488 a @i{for-init-statement} extends to the end of the enclosing scope,
1489 as was the case in old versions of G++, and other (traditional)
1490 implementations of C++.
1492 The default if neither flag is given to follow the standard,
1493 but to allow and give a warning for old-style code that would
1494 otherwise be invalid, or have different behavior.
1496 @item -fno-gnu-keywords
1497 @opindex fno-gnu-keywords
1498 Do not recognize @code{typeof} as a keyword, so that code can use this
1499 word as an identifier. You can use the keyword @code{__typeof__} instead.
1500 @option{-ansi} implies @option{-fno-gnu-keywords}.
1502 @item -fno-implicit-templates
1503 @opindex fno-implicit-templates
1504 Never emit code for non-inline templates which are instantiated
1505 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1506 @xref{Template Instantiation}, for more information.
1508 @item -fno-implicit-inline-templates
1509 @opindex fno-implicit-inline-templates
1510 Don't emit code for implicit instantiations of inline templates, either.
1511 The default is to handle inlines differently so that compiles with and
1512 without optimization will need the same set of explicit instantiations.
1514 @item -fno-implement-inlines
1515 @opindex fno-implement-inlines
1516 To save space, do not emit out-of-line copies of inline functions
1517 controlled by @samp{#pragma implementation}. This will cause linker
1518 errors if these functions are not inlined everywhere they are called.
1520 @item -fms-extensions
1521 @opindex fms-extensions
1522 Disable pedantic warnings about constructs used in MFC, such as implicit
1523 int and getting a pointer to member function via non-standard syntax.
1525 @item -fno-nonansi-builtins
1526 @opindex fno-nonansi-builtins
1527 Disable built-in declarations of functions that are not mandated by
1528 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1529 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1531 @item -fno-operator-names
1532 @opindex fno-operator-names
1533 Do not treat the operator name keywords @code{and}, @code{bitand},
1534 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1535 synonyms as keywords.
1537 @item -fno-optional-diags
1538 @opindex fno-optional-diags
1539 Disable diagnostics that the standard says a compiler does not need to
1540 issue. Currently, the only such diagnostic issued by G++ is the one for
1541 a name having multiple meanings within a class.
1544 @opindex fpermissive
1545 Downgrade some diagnostics about nonconformant code from errors to
1546 warnings. Thus, using @option{-fpermissive} will allow some
1547 nonconforming code to compile.
1551 Enable automatic template instantiation at link time. This option also
1552 implies @option{-fno-implicit-templates}. @xref{Template
1553 Instantiation}, for more information.
1557 Disable generation of information about every class with virtual
1558 functions for use by the C++ runtime type identification features
1559 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1560 of the language, you can save some space by using this flag. Note that
1561 exception handling uses the same information, but it will generate it as
1566 Emit statistics about front-end processing at the end of the compilation.
1567 This information is generally only useful to the G++ development team.
1569 @item -ftemplate-depth-@var{n}
1570 @opindex ftemplate-depth
1571 Set the maximum instantiation depth for template classes to @var{n}.
1572 A limit on the template instantiation depth is needed to detect
1573 endless recursions during template class instantiation. ANSI/ISO C++
1574 conforming programs must not rely on a maximum depth greater than 17.
1576 @item -fno-threadsafe-statics
1577 @opindex fno-threadsafe-statics
1578 Do not emit the extra code to use the routines specified in the C++
1579 ABI for thread-safe initialization of local statics. You can use this
1580 option to reduce code size slightly in code that doesn't need to be
1583 @item -fuse-cxa-atexit
1584 @opindex fuse-cxa-atexit
1585 Register destructors for objects with static storage duration with the
1586 @code{__cxa_atexit} function rather than the @code{atexit} function.
1587 This option is required for fully standards-compliant handling of static
1588 destructors, but will only work if your C library supports
1589 @code{__cxa_atexit}.
1591 @item -fvisibility-inlines-hidden
1592 @opindex fvisibility-inlines-hidden
1593 Causes all inlined methods to be marked with
1594 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1595 appear in the export table of a DSO and do not require a PLT indirection
1596 when used within the DSO@. Enabling this option can have a dramatic effect
1597 on load and link times of a DSO as it massively reduces the size of the
1598 dynamic export table when the library makes heavy use of templates. While
1599 it can cause bloating through duplication of code within each DSO where
1600 it is used, often the wastage is less than the considerable space occupied
1601 by a long symbol name in the export table which is typical when using
1602 templates and namespaces. For even more savings, combine with the
1603 @option{-fvisibility=hidden} switch.
1607 Do not use weak symbol support, even if it is provided by the linker.
1608 By default, G++ will use weak symbols if they are available. This
1609 option exists only for testing, and should not be used by end-users;
1610 it will result in inferior code and has no benefits. This option may
1611 be removed in a future release of G++.
1615 Do not search for header files in the standard directories specific to
1616 C++, but do still search the other standard directories. (This option
1617 is used when building the C++ library.)
1620 In addition, these optimization, warning, and code generation options
1621 have meanings only for C++ programs:
1624 @item -fno-default-inline
1625 @opindex fno-default-inline
1626 Do not assume @samp{inline} for functions defined inside a class scope.
1627 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1628 functions will have linkage like inline functions; they just won't be
1631 @item -Wabi @r{(C++ only)}
1633 Warn when G++ generates code that is probably not compatible with the
1634 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1635 all such cases, there are probably some cases that are not warned about,
1636 even though G++ is generating incompatible code. There may also be
1637 cases where warnings are emitted even though the code that is generated
1640 You should rewrite your code to avoid these warnings if you are
1641 concerned about the fact that code generated by G++ may not be binary
1642 compatible with code generated by other compilers.
1644 The known incompatibilities at this point include:
1649 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1650 pack data into the same byte as a base class. For example:
1653 struct A @{ virtual void f(); int f1 : 1; @};
1654 struct B : public A @{ int f2 : 1; @};
1658 In this case, G++ will place @code{B::f2} into the same byte
1659 as@code{A::f1}; other compilers will not. You can avoid this problem
1660 by explicitly padding @code{A} so that its size is a multiple of the
1661 byte size on your platform; that will cause G++ and other compilers to
1662 layout @code{B} identically.
1665 Incorrect handling of tail-padding for virtual bases. G++ does not use
1666 tail padding when laying out virtual bases. For example:
1669 struct A @{ virtual void f(); char c1; @};
1670 struct B @{ B(); char c2; @};
1671 struct C : public A, public virtual B @{@};
1675 In this case, G++ will not place @code{B} into the tail-padding for
1676 @code{A}; other compilers will. You can avoid this problem by
1677 explicitly padding @code{A} so that its size is a multiple of its
1678 alignment (ignoring virtual base classes); that will cause G++ and other
1679 compilers to layout @code{C} identically.
1682 Incorrect handling of bit-fields with declared widths greater than that
1683 of their underlying types, when the bit-fields appear in a union. For
1687 union U @{ int i : 4096; @};
1691 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1692 union too small by the number of bits in an @code{int}.
1695 Empty classes can be placed at incorrect offsets. For example:
1705 struct C : public B, public A @{@};
1709 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1710 it should be placed at offset zero. G++ mistakenly believes that the
1711 @code{A} data member of @code{B} is already at offset zero.
1714 Names of template functions whose types involve @code{typename} or
1715 template template parameters can be mangled incorrectly.
1718 template <typename Q>
1719 void f(typename Q::X) @{@}
1721 template <template <typename> class Q>
1722 void f(typename Q<int>::X) @{@}
1726 Instantiations of these templates may be mangled incorrectly.
1730 @item -Wctor-dtor-privacy @r{(C++ only)}
1731 @opindex Wctor-dtor-privacy
1732 Warn when a class seems unusable because all the constructors or
1733 destructors in that class are private, and it has neither friends nor
1734 public static member functions.
1736 @item -Wnon-virtual-dtor @r{(C++ only)}
1737 @opindex Wnon-virtual-dtor
1738 Warn when a class appears to be polymorphic, thereby requiring a virtual
1739 destructor, yet it declares a non-virtual one.
1740 This warning is enabled by @option{-Wall}.
1742 @item -Wreorder @r{(C++ only)}
1744 @cindex reordering, warning
1745 @cindex warning for reordering of member initializers
1746 Warn when the order of member initializers given in the code does not
1747 match the order in which they must be executed. For instance:
1753 A(): j (0), i (1) @{ @}
1757 The compiler will rearrange the member initializers for @samp{i}
1758 and @samp{j} to match the declaration order of the members, emitting
1759 a warning to that effect. This warning is enabled by @option{-Wall}.
1762 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1765 @item -Weffc++ @r{(C++ only)}
1767 Warn about violations of the following style guidelines from Scott Meyers'
1768 @cite{Effective C++} book:
1772 Item 11: Define a copy constructor and an assignment operator for classes
1773 with dynamically allocated memory.
1776 Item 12: Prefer initialization to assignment in constructors.
1779 Item 14: Make destructors virtual in base classes.
1782 Item 15: Have @code{operator=} return a reference to @code{*this}.
1785 Item 23: Don't try to return a reference when you must return an object.
1789 Also warn about violations of the following style guidelines from
1790 Scott Meyers' @cite{More Effective C++} book:
1794 Item 6: Distinguish between prefix and postfix forms of increment and
1795 decrement operators.
1798 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1802 When selecting this option, be aware that the standard library
1803 headers do not obey all of these guidelines; use @samp{grep -v}
1804 to filter out those warnings.
1806 @item -Wno-deprecated @r{(C++ only)}
1807 @opindex Wno-deprecated
1808 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1810 @item -Wstrict-null-sentinel @r{(C++ only)}
1811 @opindex Wstrict-null-sentinel
1812 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1813 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1814 to @code{__null}. Although it is a null pointer constant not a null pointer,
1815 it is guaranteed to of the same size as a pointer. But this use is
1816 not portable across different compilers.
1818 @item -Wno-non-template-friend @r{(C++ only)}
1819 @opindex Wno-non-template-friend
1820 Disable warnings when non-templatized friend functions are declared
1821 within a template. Since the advent of explicit template specification
1822 support in G++, if the name of the friend is an unqualified-id (i.e.,
1823 @samp{friend foo(int)}), the C++ language specification demands that the
1824 friend declare or define an ordinary, nontemplate function. (Section
1825 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1826 could be interpreted as a particular specialization of a templatized
1827 function. Because this non-conforming behavior is no longer the default
1828 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1829 check existing code for potential trouble spots and is on by default.
1830 This new compiler behavior can be turned off with
1831 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1832 but disables the helpful warning.
1834 @item -Wold-style-cast @r{(C++ only)}
1835 @opindex Wold-style-cast
1836 Warn if an old-style (C-style) cast to a non-void type is used within
1837 a C++ program. The new-style casts (@samp{dynamic_cast},
1838 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1839 less vulnerable to unintended effects and much easier to search for.
1841 @item -Woverloaded-virtual @r{(C++ only)}
1842 @opindex Woverloaded-virtual
1843 @cindex overloaded virtual fn, warning
1844 @cindex warning for overloaded virtual fn
1845 Warn when a function declaration hides virtual functions from a
1846 base class. For example, in:
1853 struct B: public A @{
1858 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1866 will fail to compile.
1868 @item -Wno-pmf-conversions @r{(C++ only)}
1869 @opindex Wno-pmf-conversions
1870 Disable the diagnostic for converting a bound pointer to member function
1873 @item -Wsign-promo @r{(C++ only)}
1874 @opindex Wsign-promo
1875 Warn when overload resolution chooses a promotion from unsigned or
1876 enumerated type to a signed type, over a conversion to an unsigned type of
1877 the same size. Previous versions of G++ would try to preserve
1878 unsignedness, but the standard mandates the current behavior.
1883 A& operator = (int);
1893 In this example, G++ will synthesize a default @samp{A& operator =
1894 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1897 @node Objective-C and Objective-C++ Dialect Options
1898 @section Options Controlling Objective-C and Objective-C++ Dialects
1900 @cindex compiler options, Objective-C and Objective-C++
1901 @cindex Objective-C and Objective-C++ options, command line
1902 @cindex options, Objective-C and Objective-C++
1903 (NOTE: This manual does not describe the Objective-C and Objective-C++
1904 languages themselves. See @xref{Standards,,Language Standards
1905 Supported by GCC}, for references.)
1907 This section describes the command-line options that are only meaningful
1908 for Objective-C and Objective-C++ programs, but you can also use most of
1909 the language-independent GNU compiler options.
1910 For example, you might compile a file @code{some_class.m} like this:
1913 gcc -g -fgnu-runtime -O -c some_class.m
1917 In this example, @option{-fgnu-runtime} is an option meant only for
1918 Objective-C and Objective-C++ programs; you can use the other options with
1919 any language supported by GCC@.
1921 Note that since Objective-C is an extension of the C language, Objective-C
1922 compilations may also use options specific to the C front-end (e.g.,
1923 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1924 C++-specific options (e.g., @option{-Wabi}).
1926 Here is a list of options that are @emph{only} for compiling Objective-C
1927 and Objective-C++ programs:
1930 @item -fconstant-string-class=@var{class-name}
1931 @opindex fconstant-string-class
1932 Use @var{class-name} as the name of the class to instantiate for each
1933 literal string specified with the syntax @code{@@"@dots{}"}. The default
1934 class name is @code{NXConstantString} if the GNU runtime is being used, and
1935 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1936 @option{-fconstant-cfstrings} option, if also present, will override the
1937 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1938 to be laid out as constant CoreFoundation strings.
1941 @opindex fgnu-runtime
1942 Generate object code compatible with the standard GNU Objective-C
1943 runtime. This is the default for most types of systems.
1945 @item -fnext-runtime
1946 @opindex fnext-runtime
1947 Generate output compatible with the NeXT runtime. This is the default
1948 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1949 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1952 @item -fno-nil-receivers
1953 @opindex fno-nil-receivers
1954 Assume that all Objective-C message dispatches (e.g.,
1955 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1956 is not @code{nil}. This allows for more efficient entry points in the runtime
1957 to be used. Currently, this option is only available in conjunction with
1958 the NeXT runtime on Mac OS X 10.3 and later.
1960 @item -fobjc-call-cxx-cdtors
1961 @opindex fobjc-call-cxx-cdtors
1962 For each Objective-C class, check if any of its instance variables is a
1963 C++ object with a non-trivial default constructor. If so, synthesize a
1964 special @code{- (id) .cxx_construct} instance method that will run
1965 non-trivial default constructors on any such instance variables, in order,
1966 and then return @code{self}. Similarly, check if any instance variable
1967 is a C++ object with a non-trivial destructor, and if so, synthesize a
1968 special @code{- (void) .cxx_destruct} method that will run
1969 all such default destructors, in reverse order.
1971 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1972 thusly generated will only operate on instance variables declared in the
1973 current Objective-C class, and not those inherited from superclasses. It
1974 is the responsibility of the Objective-C runtime to invoke all such methods
1975 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1976 will be invoked by the runtime immediately after a new object
1977 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1978 be invoked immediately before the runtime deallocates an object instance.
1980 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1981 support for invoking the @code{- (id) .cxx_construct} and
1982 @code{- (void) .cxx_destruct} methods.
1984 @item -fobjc-direct-dispatch
1985 @opindex fobjc-direct-dispatch
1986 Allow fast jumps to the message dispatcher. On Darwin this is
1987 accomplished via the comm page.
1989 @item -fobjc-exceptions
1990 @opindex fobjc-exceptions
1991 Enable syntactic support for structured exception handling in Objective-C,
1992 similar to what is offered by C++ and Java. This option is
1993 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2002 @@catch (AnObjCClass *exc) @{
2009 @@catch (AnotherClass *exc) @{
2012 @@catch (id allOthers) @{
2022 The @code{@@throw} statement may appear anywhere in an Objective-C or
2023 Objective-C++ program; when used inside of a @code{@@catch} block, the
2024 @code{@@throw} may appear without an argument (as shown above), in which case
2025 the object caught by the @code{@@catch} will be rethrown.
2027 Note that only (pointers to) Objective-C objects may be thrown and
2028 caught using this scheme. When an object is thrown, it will be caught
2029 by the nearest @code{@@catch} clause capable of handling objects of that type,
2030 analogously to how @code{catch} blocks work in C++ and Java. A
2031 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2032 any and all Objective-C exceptions not caught by previous @code{@@catch}
2035 The @code{@@finally} clause, if present, will be executed upon exit from the
2036 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2037 regardless of whether any exceptions are thrown, caught or rethrown
2038 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2039 of the @code{finally} clause in Java.
2041 There are several caveats to using the new exception mechanism:
2045 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2046 idioms provided by the @code{NSException} class, the new
2047 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2048 systems, due to additional functionality needed in the (NeXT) Objective-C
2052 As mentioned above, the new exceptions do not support handling
2053 types other than Objective-C objects. Furthermore, when used from
2054 Objective-C++, the Objective-C exception model does not interoperate with C++
2055 exceptions at this time. This means you cannot @code{@@throw} an exception
2056 from Objective-C and @code{catch} it in C++, or vice versa
2057 (i.e., @code{throw @dots{} @@catch}).
2060 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2061 blocks for thread-safe execution:
2064 @@synchronized (ObjCClass *guard) @{
2069 Upon entering the @code{@@synchronized} block, a thread of execution shall
2070 first check whether a lock has been placed on the corresponding @code{guard}
2071 object by another thread. If it has, the current thread shall wait until
2072 the other thread relinquishes its lock. Once @code{guard} becomes available,
2073 the current thread will place its own lock on it, execute the code contained in
2074 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2075 making @code{guard} available to other threads).
2077 Unlike Java, Objective-C does not allow for entire methods to be marked
2078 @code{@@synchronized}. Note that throwing exceptions out of
2079 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2080 to be unlocked properly.
2084 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2086 @item -freplace-objc-classes
2087 @opindex freplace-objc-classes
2088 Emit a special marker instructing @command{ld(1)} not to statically link in
2089 the resulting object file, and allow @command{dyld(1)} to load it in at
2090 run time instead. This is used in conjunction with the Fix-and-Continue
2091 debugging mode, where the object file in question may be recompiled and
2092 dynamically reloaded in the course of program execution, without the need
2093 to restart the program itself. Currently, Fix-and-Continue functionality
2094 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2099 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2100 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2101 compile time) with static class references that get initialized at load time,
2102 which improves run-time performance. Specifying the @option{-fzero-link} flag
2103 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2104 to be retained. This is useful in Zero-Link debugging mode, since it allows
2105 for individual class implementations to be modified during program execution.
2109 Dump interface declarations for all classes seen in the source file to a
2110 file named @file{@var{sourcename}.decl}.
2112 @item -Wassign-intercept
2113 @opindex Wassign-intercept
2114 Warn whenever an Objective-C assignment is being intercepted by the
2118 @opindex Wno-protocol
2119 If a class is declared to implement a protocol, a warning is issued for
2120 every method in the protocol that is not implemented by the class. The
2121 default behavior is to issue a warning for every method not explicitly
2122 implemented in the class, even if a method implementation is inherited
2123 from the superclass. If you use the @option{-Wno-protocol} option, then
2124 methods inherited from the superclass are considered to be implemented,
2125 and no warning is issued for them.
2129 Warn if multiple methods of different types for the same selector are
2130 found during compilation. The check is performed on the list of methods
2131 in the final stage of compilation. Additionally, a check is performed
2132 for each selector appearing in a @code{@@selector(@dots{})}
2133 expression, and a corresponding method for that selector has been found
2134 during compilation. Because these checks scan the method table only at
2135 the end of compilation, these warnings are not produced if the final
2136 stage of compilation is not reached, for example because an error is
2137 found during compilation, or because the @option{-fsyntax-only} option is
2140 @item -Wstrict-selector-match
2141 @opindex Wstrict-selector-match
2142 Warn if multiple methods with differing argument and/or return types are
2143 found for a given selector when attempting to send a message using this
2144 selector to a receiver of type @code{id} or @code{Class}. When this flag
2145 is off (which is the default behavior), the compiler will omit such warnings
2146 if any differences found are confined to types which share the same size
2149 @item -Wundeclared-selector
2150 @opindex Wundeclared-selector
2151 Warn if a @code{@@selector(@dots{})} expression referring to an
2152 undeclared selector is found. A selector is considered undeclared if no
2153 method with that name has been declared before the
2154 @code{@@selector(@dots{})} expression, either explicitly in an
2155 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2156 an @code{@@implementation} section. This option always performs its
2157 checks as soon as a @code{@@selector(@dots{})} expression is found,
2158 while @option{-Wselector} only performs its checks in the final stage of
2159 compilation. This also enforces the coding style convention
2160 that methods and selectors must be declared before being used.
2162 @item -print-objc-runtime-info
2163 @opindex print-objc-runtime-info
2164 Generate C header describing the largest structure that is passed by
2169 @node Language Independent Options
2170 @section Options to Control Diagnostic Messages Formatting
2171 @cindex options to control diagnostics formatting
2172 @cindex diagnostic messages
2173 @cindex message formatting
2175 Traditionally, diagnostic messages have been formatted irrespective of
2176 the output device's aspect (e.g.@: its width, @dots{}). The options described
2177 below can be used to control the diagnostic messages formatting
2178 algorithm, e.g.@: how many characters per line, how often source location
2179 information should be reported. Right now, only the C++ front end can
2180 honor these options. However it is expected, in the near future, that
2181 the remaining front ends would be able to digest them correctly.
2184 @item -fmessage-length=@var{n}
2185 @opindex fmessage-length
2186 Try to format error messages so that they fit on lines of about @var{n}
2187 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2188 the front ends supported by GCC@. If @var{n} is zero, then no
2189 line-wrapping will be done; each error message will appear on a single
2192 @opindex fdiagnostics-show-location
2193 @item -fdiagnostics-show-location=once
2194 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2195 reporter to emit @emph{once} source location information; that is, in
2196 case the message is too long to fit on a single physical line and has to
2197 be wrapped, the source location won't be emitted (as prefix) again,
2198 over and over, in subsequent continuation lines. This is the default
2201 @item -fdiagnostics-show-location=every-line
2202 Only meaningful in line-wrapping mode. Instructs the diagnostic
2203 messages reporter to emit the same source location information (as
2204 prefix) for physical lines that result from the process of breaking
2205 a message which is too long to fit on a single line.
2207 @item -fdiagnostics-show-options
2208 @opindex fdiagnostics-show-options
2209 This option instructs the diagnostic machinery to add text to each
2210 diagnostic emitted, which indicates which command line option directly
2211 controls that diagnostic, when such an option is known to the
2212 diagnostic machinery.
2216 @node Warning Options
2217 @section Options to Request or Suppress Warnings
2218 @cindex options to control warnings
2219 @cindex warning messages
2220 @cindex messages, warning
2221 @cindex suppressing warnings
2223 Warnings are diagnostic messages that report constructions which
2224 are not inherently erroneous but which are risky or suggest there
2225 may have been an error.
2227 You can request many specific warnings with options beginning @samp{-W},
2228 for example @option{-Wimplicit} to request warnings on implicit
2229 declarations. Each of these specific warning options also has a
2230 negative form beginning @samp{-Wno-} to turn off warnings;
2231 for example, @option{-Wno-implicit}. This manual lists only one of the
2232 two forms, whichever is not the default.
2234 The following options control the amount and kinds of warnings produced
2235 by GCC; for further, language-specific options also refer to
2236 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2240 @cindex syntax checking
2242 @opindex fsyntax-only
2243 Check the code for syntax errors, but don't do anything beyond that.
2247 Issue all the warnings demanded by strict ISO C and ISO C++;
2248 reject all programs that use forbidden extensions, and some other
2249 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2250 version of the ISO C standard specified by any @option{-std} option used.
2252 Valid ISO C and ISO C++ programs should compile properly with or without
2253 this option (though a rare few will require @option{-ansi} or a
2254 @option{-std} option specifying the required version of ISO C)@. However,
2255 without this option, certain GNU extensions and traditional C and C++
2256 features are supported as well. With this option, they are rejected.
2258 @option{-pedantic} does not cause warning messages for use of the
2259 alternate keywords whose names begin and end with @samp{__}. Pedantic
2260 warnings are also disabled in the expression that follows
2261 @code{__extension__}. However, only system header files should use
2262 these escape routes; application programs should avoid them.
2263 @xref{Alternate Keywords}.
2265 Some users try to use @option{-pedantic} to check programs for strict ISO
2266 C conformance. They soon find that it does not do quite what they want:
2267 it finds some non-ISO practices, but not all---only those for which
2268 ISO C @emph{requires} a diagnostic, and some others for which
2269 diagnostics have been added.
2271 A feature to report any failure to conform to ISO C might be useful in
2272 some instances, but would require considerable additional work and would
2273 be quite different from @option{-pedantic}. We don't have plans to
2274 support such a feature in the near future.
2276 Where the standard specified with @option{-std} represents a GNU
2277 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2278 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2279 extended dialect is based. Warnings from @option{-pedantic} are given
2280 where they are required by the base standard. (It would not make sense
2281 for such warnings to be given only for features not in the specified GNU
2282 C dialect, since by definition the GNU dialects of C include all
2283 features the compiler supports with the given option, and there would be
2284 nothing to warn about.)
2286 @item -pedantic-errors
2287 @opindex pedantic-errors
2288 Like @option{-pedantic}, except that errors are produced rather than
2293 Inhibit all warning messages.
2297 Inhibit warning messages about the use of @samp{#import}.
2299 @item -Wchar-subscripts
2300 @opindex Wchar-subscripts
2301 Warn if an array subscript has type @code{char}. This is a common cause
2302 of error, as programmers often forget that this type is signed on some
2304 This warning is enabled by @option{-Wall}.
2308 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2309 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2310 This warning is enabled by @option{-Wall}.
2312 @item -Wfatal-errors
2313 @opindex Wfatal-errors
2314 This option causes the compiler to abort compilation on the first error
2315 occurred rather than trying to keep going and printing further error
2320 @opindex ffreestanding
2321 @opindex fno-builtin
2322 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2323 the arguments supplied have types appropriate to the format string
2324 specified, and that the conversions specified in the format string make
2325 sense. This includes standard functions, and others specified by format
2326 attributes (@pxref{Function Attributes}), in the @code{printf},
2327 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2328 not in the C standard) families (or other target-specific families).
2329 Which functions are checked without format attributes having been
2330 specified depends on the standard version selected, and such checks of
2331 functions without the attribute specified are disabled by
2332 @option{-ffreestanding} or @option{-fno-builtin}.
2334 The formats are checked against the format features supported by GNU
2335 libc version 2.2. These include all ISO C90 and C99 features, as well
2336 as features from the Single Unix Specification and some BSD and GNU
2337 extensions. Other library implementations may not support all these
2338 features; GCC does not support warning about features that go beyond a
2339 particular library's limitations. However, if @option{-pedantic} is used
2340 with @option{-Wformat}, warnings will be given about format features not
2341 in the selected standard version (but not for @code{strfmon} formats,
2342 since those are not in any version of the C standard). @xref{C Dialect
2343 Options,,Options Controlling C Dialect}.
2345 Since @option{-Wformat} also checks for null format arguments for
2346 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2348 @option{-Wformat} is included in @option{-Wall}. For more control over some
2349 aspects of format checking, the options @option{-Wformat-y2k},
2350 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2351 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2352 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2355 @opindex Wformat-y2k
2356 If @option{-Wformat} is specified, also warn about @code{strftime}
2357 formats which may yield only a two-digit year.
2359 @item -Wno-format-extra-args
2360 @opindex Wno-format-extra-args
2361 If @option{-Wformat} is specified, do not warn about excess arguments to a
2362 @code{printf} or @code{scanf} format function. The C standard specifies
2363 that such arguments are ignored.
2365 Where the unused arguments lie between used arguments that are
2366 specified with @samp{$} operand number specifications, normally
2367 warnings are still given, since the implementation could not know what
2368 type to pass to @code{va_arg} to skip the unused arguments. However,
2369 in the case of @code{scanf} formats, this option will suppress the
2370 warning if the unused arguments are all pointers, since the Single
2371 Unix Specification says that such unused arguments are allowed.
2373 @item -Wno-format-zero-length
2374 @opindex Wno-format-zero-length
2375 If @option{-Wformat} is specified, do not warn about zero-length formats.
2376 The C standard specifies that zero-length formats are allowed.
2378 @item -Wformat-nonliteral
2379 @opindex Wformat-nonliteral
2380 If @option{-Wformat} is specified, also warn if the format string is not a
2381 string literal and so cannot be checked, unless the format function
2382 takes its format arguments as a @code{va_list}.
2384 @item -Wformat-security
2385 @opindex Wformat-security
2386 If @option{-Wformat} is specified, also warn about uses of format
2387 functions that represent possible security problems. At present, this
2388 warns about calls to @code{printf} and @code{scanf} functions where the
2389 format string is not a string literal and there are no format arguments,
2390 as in @code{printf (foo);}. This may be a security hole if the format
2391 string came from untrusted input and contains @samp{%n}. (This is
2392 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2393 in future warnings may be added to @option{-Wformat-security} that are not
2394 included in @option{-Wformat-nonliteral}.)
2398 Enable @option{-Wformat} plus format checks not included in
2399 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2400 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2404 Warn about passing a null pointer for arguments marked as
2405 requiring a non-null value by the @code{nonnull} function attribute.
2407 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2408 can be disabled with the @option{-Wno-nonnull} option.
2410 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2412 Warn about uninitialized variables which are initialized with themselves.
2413 Note this option can only be used with the @option{-Wuninitialized} option,
2414 which in turn only works with @option{-O1} and above.
2416 For example, GCC will warn about @code{i} being uninitialized in the
2417 following snippet only when @option{-Winit-self} has been specified:
2428 @item -Wimplicit-int
2429 @opindex Wimplicit-int
2430 Warn when a declaration does not specify a type.
2431 This warning is enabled by @option{-Wall}.
2433 @item -Wimplicit-function-declaration
2434 @itemx -Werror-implicit-function-declaration
2435 @opindex Wimplicit-function-declaration
2436 @opindex Werror-implicit-function-declaration
2437 Give a warning (or error) whenever a function is used before being
2438 declared. The form @option{-Wno-error-implicit-function-declaration}
2440 This warning is enabled by @option{-Wall} (as a warning, not an error).
2444 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2445 This warning is enabled by @option{-Wall}.
2449 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2450 function with external linkage, returning int, taking either zero
2451 arguments, two, or three arguments of appropriate types.
2452 This warning is enabled by @option{-Wall}.
2454 @item -Wmissing-braces
2455 @opindex Wmissing-braces
2456 Warn if an aggregate or union initializer is not fully bracketed. In
2457 the following example, the initializer for @samp{a} is not fully
2458 bracketed, but that for @samp{b} is fully bracketed.
2461 int a[2][2] = @{ 0, 1, 2, 3 @};
2462 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2465 This warning is enabled by @option{-Wall}.
2467 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2468 @opindex Wmissing-include-dirs
2469 Warn if a user-supplied include directory does not exist.
2472 @opindex Wparentheses
2473 Warn if parentheses are omitted in certain contexts, such
2474 as when there is an assignment in a context where a truth value
2475 is expected, or when operators are nested whose precedence people
2476 often get confused about. Only the warning for an assignment used as
2477 a truth value is supported when compiling C++; the other warnings are
2478 only supported when compiling C@.
2480 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2481 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2482 interpretation from that of ordinary mathematical notation.
2484 Also warn about constructions where there may be confusion to which
2485 @code{if} statement an @code{else} branch belongs. Here is an example of
2500 In C, every @code{else} branch belongs to the innermost possible @code{if}
2501 statement, which in this example is @code{if (b)}. This is often not
2502 what the programmer expected, as illustrated in the above example by
2503 indentation the programmer chose. When there is the potential for this
2504 confusion, GCC will issue a warning when this flag is specified.
2505 To eliminate the warning, add explicit braces around the innermost
2506 @code{if} statement so there is no way the @code{else} could belong to
2507 the enclosing @code{if}. The resulting code would look like this:
2523 This warning is enabled by @option{-Wall}.
2525 @item -Wsequence-point
2526 @opindex Wsequence-point
2527 Warn about code that may have undefined semantics because of violations
2528 of sequence point rules in the C standard.
2530 The C standard defines the order in which expressions in a C program are
2531 evaluated in terms of @dfn{sequence points}, which represent a partial
2532 ordering between the execution of parts of the program: those executed
2533 before the sequence point, and those executed after it. These occur
2534 after the evaluation of a full expression (one which is not part of a
2535 larger expression), after the evaluation of the first operand of a
2536 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2537 function is called (but after the evaluation of its arguments and the
2538 expression denoting the called function), and in certain other places.
2539 Other than as expressed by the sequence point rules, the order of
2540 evaluation of subexpressions of an expression is not specified. All
2541 these rules describe only a partial order rather than a total order,
2542 since, for example, if two functions are called within one expression
2543 with no sequence point between them, the order in which the functions
2544 are called is not specified. However, the standards committee have
2545 ruled that function calls do not overlap.
2547 It is not specified when between sequence points modifications to the
2548 values of objects take effect. Programs whose behavior depends on this
2549 have undefined behavior; the C standard specifies that ``Between the
2550 previous and next sequence point an object shall have its stored value
2551 modified at most once by the evaluation of an expression. Furthermore,
2552 the prior value shall be read only to determine the value to be
2553 stored.''. If a program breaks these rules, the results on any
2554 particular implementation are entirely unpredictable.
2556 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2557 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2558 diagnosed by this option, and it may give an occasional false positive
2559 result, but in general it has been found fairly effective at detecting
2560 this sort of problem in programs.
2562 The present implementation of this option only works for C programs. A
2563 future implementation may also work for C++ programs.
2565 The C standard is worded confusingly, therefore there is some debate
2566 over the precise meaning of the sequence point rules in subtle cases.
2567 Links to discussions of the problem, including proposed formal
2568 definitions, may be found on the GCC readings page, at
2569 @w{@uref{http://gcc.gnu.org/readings.html}}.
2571 This warning is enabled by @option{-Wall}.
2574 @opindex Wreturn-type
2575 Warn whenever a function is defined with a return-type that defaults to
2576 @code{int}. Also warn about any @code{return} statement with no
2577 return-value in a function whose return-type is not @code{void}.
2579 For C, also warn if the return type of a function has a type qualifier
2580 such as @code{const}. Such a type qualifier has no effect, since the
2581 value returned by a function is not an lvalue. ISO C prohibits
2582 qualified @code{void} return types on function definitions, so such
2583 return types always receive a warning even without this option.
2585 For C++, a function without return type always produces a diagnostic
2586 message, even when @option{-Wno-return-type} is specified. The only
2587 exceptions are @samp{main} and functions defined in system headers.
2589 This warning is enabled by @option{-Wall}.
2593 Warn whenever a @code{switch} statement has an index of enumerated type
2594 and lacks a @code{case} for one or more of the named codes of that
2595 enumeration. (The presence of a @code{default} label prevents this
2596 warning.) @code{case} labels outside the enumeration range also
2597 provoke warnings when this option is used.
2598 This warning is enabled by @option{-Wall}.
2600 @item -Wswitch-default
2601 @opindex Wswitch-switch
2602 Warn whenever a @code{switch} statement does not have a @code{default}
2606 @opindex Wswitch-enum
2607 Warn whenever a @code{switch} statement has an index of enumerated type
2608 and lacks a @code{case} for one or more of the named codes of that
2609 enumeration. @code{case} labels outside the enumeration range also
2610 provoke warnings when this option is used.
2614 Warn if any trigraphs are encountered that might change the meaning of
2615 the program (trigraphs within comments are not warned about).
2616 This warning is enabled by @option{-Wall}.
2618 @item -Wunused-function
2619 @opindex Wunused-function
2620 Warn whenever a static function is declared but not defined or a
2621 non-inline static function is unused.
2622 This warning is enabled by @option{-Wall}.
2624 @item -Wunused-label
2625 @opindex Wunused-label
2626 Warn whenever a label is declared but not used.
2627 This warning is enabled by @option{-Wall}.
2629 To suppress this warning use the @samp{unused} attribute
2630 (@pxref{Variable Attributes}).
2632 @item -Wunused-parameter
2633 @opindex Wunused-parameter
2634 Warn whenever a function parameter is unused aside from its declaration.
2636 To suppress this warning use the @samp{unused} attribute
2637 (@pxref{Variable Attributes}).
2639 @item -Wunused-variable
2640 @opindex Wunused-variable
2641 Warn whenever a local variable or non-constant static variable is unused
2642 aside from its declaration
2643 This warning is enabled by @option{-Wall}.
2645 To suppress this warning use the @samp{unused} attribute
2646 (@pxref{Variable Attributes}).
2648 @item -Wunused-value
2649 @opindex Wunused-value
2650 Warn whenever a statement computes a result that is explicitly not used.
2651 This warning is enabled by @option{-Wall}.
2653 To suppress this warning cast the expression to @samp{void}.
2657 All the above @option{-Wunused} options combined.
2659 In order to get a warning about an unused function parameter, you must
2660 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2661 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2663 @item -Wuninitialized
2664 @opindex Wuninitialized
2665 Warn if an automatic variable is used without first being initialized or
2666 if a variable may be clobbered by a @code{setjmp} call.
2668 These warnings are possible only in optimizing compilation,
2669 because they require data flow information that is computed only
2670 when optimizing. If you don't specify @option{-O}, you simply won't
2673 If you want to warn about code which uses the uninitialized value of the
2674 variable in its own initializer, use the @option{-Winit-self} option.
2676 These warnings occur for individual uninitialized or clobbered
2677 elements of structure, union or array variables as well as for
2678 variables which are uninitialized or clobbered as a whole. They do
2679 not occur for variables or elements declared @code{volatile}. Because
2680 these warnings depend on optimization, the exact variables or elements
2681 for which there are warnings will depend on the precise optimization
2682 options and version of GCC used.
2684 Note that there may be no warning about a variable that is used only
2685 to compute a value that itself is never used, because such
2686 computations may be deleted by data flow analysis before the warnings
2689 These warnings are made optional because GCC is not smart
2690 enough to see all the reasons why the code might be correct
2691 despite appearing to have an error. Here is one example of how
2712 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2713 always initialized, but GCC doesn't know this. Here is
2714 another common case:
2719 if (change_y) save_y = y, y = new_y;
2721 if (change_y) y = save_y;
2726 This has no bug because @code{save_y} is used only if it is set.
2728 @cindex @code{longjmp} warnings
2729 This option also warns when a non-volatile automatic variable might be
2730 changed by a call to @code{longjmp}. These warnings as well are possible
2731 only in optimizing compilation.
2733 The compiler sees only the calls to @code{setjmp}. It cannot know
2734 where @code{longjmp} will be called; in fact, a signal handler could
2735 call it at any point in the code. As a result, you may get a warning
2736 even when there is in fact no problem because @code{longjmp} cannot
2737 in fact be called at the place which would cause a problem.
2739 Some spurious warnings can be avoided if you declare all the functions
2740 you use that never return as @code{noreturn}. @xref{Function
2743 This warning is enabled by @option{-Wall}.
2745 @item -Wunknown-pragmas
2746 @opindex Wunknown-pragmas
2747 @cindex warning for unknown pragmas
2748 @cindex unknown pragmas, warning
2749 @cindex pragmas, warning of unknown
2750 Warn when a #pragma directive is encountered which is not understood by
2751 GCC@. If this command line option is used, warnings will even be issued
2752 for unknown pragmas in system header files. This is not the case if
2753 the warnings were only enabled by the @option{-Wall} command line option.
2756 @opindex Wno-pragmas
2758 Do not warn about misuses of pragmas, such as incorrect parameters,
2759 invalid syntax, or conflicts between pragmas. See also
2760 @samp{-Wunknown-pragmas}.
2762 @item -Wstrict-aliasing
2763 @opindex Wstrict-aliasing
2764 This option is only active when @option{-fstrict-aliasing} is active.
2765 It warns about code which might break the strict aliasing rules that the
2766 compiler is using for optimization. The warning does not catch all
2767 cases, but does attempt to catch the more common pitfalls. It is
2768 included in @option{-Wall}.
2770 @item -Wstrict-aliasing=2
2771 @opindex Wstrict-aliasing=2
2772 This option is only active when @option{-fstrict-aliasing} is active.
2773 It warns about code which might break the strict aliasing rules that the
2774 compiler is using for optimization. This warning catches more cases than
2775 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2776 cases that are safe.
2780 All of the above @samp{-W} options combined. This enables all the
2781 warnings about constructions that some users consider questionable, and
2782 that are easy to avoid (or modify to prevent the warning), even in
2783 conjunction with macros. This also enables some language-specific
2784 warnings described in @ref{C++ Dialect Options} and
2785 @ref{Objective-C and Objective-C++ Dialect Options}.
2788 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2789 Some of them warn about constructions that users generally do not
2790 consider questionable, but which occasionally you might wish to check
2791 for; others warn about constructions that are necessary or hard to avoid
2792 in some cases, and there is no simple way to modify the code to suppress
2799 (This option used to be called @option{-W}. The older name is still
2800 supported, but the newer name is more descriptive.) Print extra warning
2801 messages for these events:
2805 A function can return either with or without a value. (Falling
2806 off the end of the function body is considered returning without
2807 a value.) For example, this function would evoke such a
2821 An expression-statement or the left-hand side of a comma expression
2822 contains no side effects.
2823 To suppress the warning, cast the unused expression to void.
2824 For example, an expression such as @samp{x[i,j]} will cause a warning,
2825 but @samp{x[(void)i,j]} will not.
2828 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2831 Storage-class specifiers like @code{static} are not the first things in
2832 a declaration. According to the C Standard, this usage is obsolescent.
2835 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2839 A comparison between signed and unsigned values could produce an
2840 incorrect result when the signed value is converted to unsigned.
2841 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2844 An aggregate has an initializer which does not initialize all members.
2845 This warning can be independently controlled by
2846 @option{-Wmissing-field-initializers}.
2849 A function parameter is declared without a type specifier in K&R-style
2857 An empty body occurs in an @samp{if} or @samp{else} statement.
2860 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2861 @samp{>}, or @samp{>=}.
2864 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2867 Any of several floating-point events that often indicate errors, such as
2868 overflow, underflow, loss of precision, etc.
2870 @item @r{(C++ only)}
2871 An enumerator and a non-enumerator both appear in a conditional expression.
2873 @item @r{(C++ only)}
2874 A non-static reference or non-static @samp{const} member appears in a
2875 class without constructors.
2877 @item @r{(C++ only)}
2878 Ambiguous virtual bases.
2880 @item @r{(C++ only)}
2881 Subscripting an array which has been declared @samp{register}.
2883 @item @r{(C++ only)}
2884 Taking the address of a variable which has been declared @samp{register}.
2886 @item @r{(C++ only)}
2887 A base class is not initialized in a derived class' copy constructor.
2890 @item -Wno-div-by-zero
2891 @opindex Wno-div-by-zero
2892 @opindex Wdiv-by-zero
2893 Do not warn about compile-time integer division by zero. Floating point
2894 division by zero is not warned about, as it can be a legitimate way of
2895 obtaining infinities and NaNs.
2897 @item -Wsystem-headers
2898 @opindex Wsystem-headers
2899 @cindex warnings from system headers
2900 @cindex system headers, warnings from
2901 Print warning messages for constructs found in system header files.
2902 Warnings from system headers are normally suppressed, on the assumption
2903 that they usually do not indicate real problems and would only make the
2904 compiler output harder to read. Using this command line option tells
2905 GCC to emit warnings from system headers as if they occurred in user
2906 code. However, note that using @option{-Wall} in conjunction with this
2907 option will @emph{not} warn about unknown pragmas in system
2908 headers---for that, @option{-Wunknown-pragmas} must also be used.
2911 @opindex Wfloat-equal
2912 Warn if floating point values are used in equality comparisons.
2914 The idea behind this is that sometimes it is convenient (for the
2915 programmer) to consider floating-point values as approximations to
2916 infinitely precise real numbers. If you are doing this, then you need
2917 to compute (by analyzing the code, or in some other way) the maximum or
2918 likely maximum error that the computation introduces, and allow for it
2919 when performing comparisons (and when producing output, but that's a
2920 different problem). In particular, instead of testing for equality, you
2921 would check to see whether the two values have ranges that overlap; and
2922 this is done with the relational operators, so equality comparisons are
2925 @item -Wtraditional @r{(C only)}
2926 @opindex Wtraditional
2927 Warn about certain constructs that behave differently in traditional and
2928 ISO C@. Also warn about ISO C constructs that have no traditional C
2929 equivalent, and/or problematic constructs which should be avoided.
2933 Macro parameters that appear within string literals in the macro body.
2934 In traditional C macro replacement takes place within string literals,
2935 but does not in ISO C@.
2938 In traditional C, some preprocessor directives did not exist.
2939 Traditional preprocessors would only consider a line to be a directive
2940 if the @samp{#} appeared in column 1 on the line. Therefore
2941 @option{-Wtraditional} warns about directives that traditional C
2942 understands but would ignore because the @samp{#} does not appear as the
2943 first character on the line. It also suggests you hide directives like
2944 @samp{#pragma} not understood by traditional C by indenting them. Some
2945 traditional implementations would not recognize @samp{#elif}, so it
2946 suggests avoiding it altogether.
2949 A function-like macro that appears without arguments.
2952 The unary plus operator.
2955 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2956 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2957 constants.) Note, these suffixes appear in macros defined in the system
2958 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2959 Use of these macros in user code might normally lead to spurious
2960 warnings, however GCC's integrated preprocessor has enough context to
2961 avoid warning in these cases.
2964 A function declared external in one block and then used after the end of
2968 A @code{switch} statement has an operand of type @code{long}.
2971 A non-@code{static} function declaration follows a @code{static} one.
2972 This construct is not accepted by some traditional C compilers.
2975 The ISO type of an integer constant has a different width or
2976 signedness from its traditional type. This warning is only issued if
2977 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2978 typically represent bit patterns, are not warned about.
2981 Usage of ISO string concatenation is detected.
2984 Initialization of automatic aggregates.
2987 Identifier conflicts with labels. Traditional C lacks a separate
2988 namespace for labels.
2991 Initialization of unions. If the initializer is zero, the warning is
2992 omitted. This is done under the assumption that the zero initializer in
2993 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2994 initializer warnings and relies on default initialization to zero in the
2998 Conversions by prototypes between fixed/floating point values and vice
2999 versa. The absence of these prototypes when compiling with traditional
3000 C would cause serious problems. This is a subset of the possible
3001 conversion warnings, for the full set use @option{-Wconversion}.
3004 Use of ISO C style function definitions. This warning intentionally is
3005 @emph{not} issued for prototype declarations or variadic functions
3006 because these ISO C features will appear in your code when using
3007 libiberty's traditional C compatibility macros, @code{PARAMS} and
3008 @code{VPARAMS}. This warning is also bypassed for nested functions
3009 because that feature is already a GCC extension and thus not relevant to
3010 traditional C compatibility.
3013 @item -Wdeclaration-after-statement @r{(C only)}
3014 @opindex Wdeclaration-after-statement
3015 Warn when a declaration is found after a statement in a block. This
3016 construct, known from C++, was introduced with ISO C99 and is by default
3017 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3018 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3022 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3024 @item -Wno-endif-labels
3025 @opindex Wno-endif-labels
3026 @opindex Wendif-labels
3027 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3031 Warn whenever a local variable shadows another local variable, parameter or
3032 global variable or whenever a built-in function is shadowed.
3034 @item -Wlarger-than-@var{len}
3035 @opindex Wlarger-than
3036 Warn whenever an object of larger than @var{len} bytes is defined.
3038 @item -Wunsafe-loop-optimizations
3039 @opindex Wunsafe-loop-optimizations
3040 Warn if the loop cannot be optimized because the compiler could not
3041 assume anything on the bounds of the loop indices. With
3042 @option{-funsafe-loop-optimizations} warn if the compiler made
3045 @item -Wpointer-arith
3046 @opindex Wpointer-arith
3047 Warn about anything that depends on the ``size of'' a function type or
3048 of @code{void}. GNU C assigns these types a size of 1, for
3049 convenience in calculations with @code{void *} pointers and pointers
3052 @item -Wbad-function-cast @r{(C only)}
3053 @opindex Wbad-function-cast
3054 Warn whenever a function call is cast to a non-matching type.
3055 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3058 Warn about ISO C constructs that are outside of the common subset of
3059 ISO C and ISO C++, e.g.@: request for implicit conversion from
3060 @code{void *} to a pointer to non-@code{void} type.
3064 Warn whenever a pointer is cast so as to remove a type qualifier from
3065 the target type. For example, warn if a @code{const char *} is cast
3066 to an ordinary @code{char *}.
3069 @opindex Wcast-align
3070 Warn whenever a pointer is cast such that the required alignment of the
3071 target is increased. For example, warn if a @code{char *} is cast to
3072 an @code{int *} on machines where integers can only be accessed at
3073 two- or four-byte boundaries.
3075 @item -Wwrite-strings
3076 @opindex Wwrite-strings
3077 When compiling C, give string constants the type @code{const
3078 char[@var{length}]} so that
3079 copying the address of one into a non-@code{const} @code{char *}
3080 pointer will get a warning; when compiling C++, warn about the
3081 deprecated conversion from string constants to @code{char *}.
3082 These warnings will help you find at
3083 compile time code that can try to write into a string constant, but
3084 only if you have been very careful about using @code{const} in
3085 declarations and prototypes. Otherwise, it will just be a nuisance;
3086 this is why we did not make @option{-Wall} request these warnings.
3089 @opindex Wconversion
3090 Warn if a prototype causes a type conversion that is different from what
3091 would happen to the same argument in the absence of a prototype. This
3092 includes conversions of fixed point to floating and vice versa, and
3093 conversions changing the width or signedness of a fixed point argument
3094 except when the same as the default promotion.
3096 Also, warn if a negative integer constant expression is implicitly
3097 converted to an unsigned type. For example, warn about the assignment
3098 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3099 casts like @code{(unsigned) -1}.
3101 @item -Wsign-compare
3102 @opindex Wsign-compare
3103 @cindex warning for comparison of signed and unsigned values
3104 @cindex comparison of signed and unsigned values, warning
3105 @cindex signed and unsigned values, comparison warning
3106 Warn when a comparison between signed and unsigned values could produce
3107 an incorrect result when the signed value is converted to unsigned.
3108 This warning is also enabled by @option{-Wextra}; to get the other warnings
3109 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3111 @item -Waggregate-return
3112 @opindex Waggregate-return
3113 Warn if any functions that return structures or unions are defined or
3114 called. (In languages where you can return an array, this also elicits
3118 @opindex Walways-true
3119 Warn about comparisons which are always true such as testing if
3120 unsigned values are greater than or equal to zero. This warning is
3121 enabled by @option{-Wall}.
3123 @item -Wno-attributes
3124 @opindex Wno-attributes
3125 @opindex Wattributes
3126 Do not warn if an unexpected @code{__attribute__} is used, such as
3127 unrecognized attributes, function attributes applied to variables,
3128 etc. This will not stop errors for incorrect use of supported
3131 @item -Wstrict-prototypes @r{(C only)}
3132 @opindex Wstrict-prototypes
3133 Warn if a function is declared or defined without specifying the
3134 argument types. (An old-style function definition is permitted without
3135 a warning if preceded by a declaration which specifies the argument
3138 @item -Wold-style-definition @r{(C only)}
3139 @opindex Wold-style-definition
3140 Warn if an old-style function definition is used. A warning is given
3141 even if there is a previous prototype.
3143 @item -Wmissing-prototypes @r{(C only)}
3144 @opindex Wmissing-prototypes
3145 Warn if a global function is defined without a previous prototype
3146 declaration. This warning is issued even if the definition itself
3147 provides a prototype. The aim is to detect global functions that fail
3148 to be declared in header files.
3150 @item -Wmissing-declarations @r{(C only)}
3151 @opindex Wmissing-declarations
3152 Warn if a global function is defined without a previous declaration.
3153 Do so even if the definition itself provides a prototype.
3154 Use this option to detect global functions that are not declared in
3157 @item -Wmissing-field-initializers
3158 @opindex Wmissing-field-initializers
3161 Warn if a structure's initializer has some fields missing. For
3162 example, the following code would cause such a warning, because
3163 @code{x.h} is implicitly zero:
3166 struct s @{ int f, g, h; @};
3167 struct s x = @{ 3, 4 @};
3170 This option does not warn about designated initializers, so the following
3171 modification would not trigger a warning:
3174 struct s @{ int f, g, h; @};
3175 struct s x = @{ .f = 3, .g = 4 @};
3178 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3179 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3181 @item -Wmissing-noreturn
3182 @opindex Wmissing-noreturn
3183 Warn about functions which might be candidates for attribute @code{noreturn}.
3184 Note these are only possible candidates, not absolute ones. Care should
3185 be taken to manually verify functions actually do not ever return before
3186 adding the @code{noreturn} attribute, otherwise subtle code generation
3187 bugs could be introduced. You will not get a warning for @code{main} in
3188 hosted C environments.
3190 @item -Wmissing-format-attribute
3191 @opindex Wmissing-format-attribute
3193 Warn about function pointers which might be candidates for @code{format}
3194 attributes. Note these are only possible candidates, not absolute ones.
3195 GCC will guess that function pointers with @code{format} attributes that
3196 are used in assignment, initialization, parameter passing or return
3197 statements should have a corresponding @code{format} attribute in the
3198 resulting type. I.e.@: the left-hand side of the assignment or
3199 initialization, the type of the parameter variable, or the return type
3200 of the containing function respectively should also have a @code{format}
3201 attribute to avoid the warning.
3203 GCC will also warn about function definitions which might be
3204 candidates for @code{format} attributes. Again, these are only
3205 possible candidates. GCC will guess that @code{format} attributes
3206 might be appropriate for any function that calls a function like
3207 @code{vprintf} or @code{vscanf}, but this might not always be the
3208 case, and some functions for which @code{format} attributes are
3209 appropriate may not be detected.
3211 @item -Wno-multichar
3212 @opindex Wno-multichar
3214 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3215 Usually they indicate a typo in the user's code, as they have
3216 implementation-defined values, and should not be used in portable code.
3218 @item -Wnormalized=<none|id|nfc|nfkc>
3219 @opindex Wnormalized
3222 @cindex character set, input normalization
3223 In ISO C and ISO C++, two identifiers are different if they are
3224 different sequences of characters. However, sometimes when characters
3225 outside the basic ASCII character set are used, you can have two
3226 different character sequences that look the same. To avoid confusion,
3227 the ISO 10646 standard sets out some @dfn{normalization rules} which
3228 when applied ensure that two sequences that look the same are turned into
3229 the same sequence. GCC can warn you if you are using identifiers which
3230 have not been normalized; this option controls that warning.
3232 There are four levels of warning that GCC supports. The default is
3233 @option{-Wnormalized=nfc}, which warns about any identifier which is
3234 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3235 recommended form for most uses.
3237 Unfortunately, there are some characters which ISO C and ISO C++ allow
3238 in identifiers that when turned into NFC aren't allowable as
3239 identifiers. That is, there's no way to use these symbols in portable
3240 ISO C or C++ and have all your identifiers in NFC.
3241 @option{-Wnormalized=id} suppresses the warning for these characters.
3242 It is hoped that future versions of the standards involved will correct
3243 this, which is why this option is not the default.
3245 You can switch the warning off for all characters by writing
3246 @option{-Wnormalized=none}. You would only want to do this if you
3247 were using some other normalization scheme (like ``D''), because
3248 otherwise you can easily create bugs that are literally impossible to see.
3250 Some characters in ISO 10646 have distinct meanings but look identical
3251 in some fonts or display methodologies, especially once formatting has
3252 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3253 LETTER N'', will display just like a regular @code{n} which has been
3254 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3255 normalisation scheme to convert all these into a standard form as
3256 well, and GCC will warn if your code is not in NFKC if you use
3257 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3258 about every identifier that contains the letter O because it might be
3259 confused with the digit 0, and so is not the default, but may be
3260 useful as a local coding convention if the programming environment is
3261 unable to be fixed to display these characters distinctly.
3263 @item -Wno-deprecated-declarations
3264 @opindex Wno-deprecated-declarations
3265 Do not warn about uses of functions, variables, and types marked as
3266 deprecated by using the @code{deprecated} attribute.
3267 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3268 @pxref{Type Attributes}.)
3272 Warn if a structure is given the packed attribute, but the packed
3273 attribute has no effect on the layout or size of the structure.
3274 Such structures may be mis-aligned for little benefit. For
3275 instance, in this code, the variable @code{f.x} in @code{struct bar}
3276 will be misaligned even though @code{struct bar} does not itself
3277 have the packed attribute:
3284 @} __attribute__((packed));
3294 Warn if padding is included in a structure, either to align an element
3295 of the structure or to align the whole structure. Sometimes when this
3296 happens it is possible to rearrange the fields of the structure to
3297 reduce the padding and so make the structure smaller.
3299 @item -Wredundant-decls
3300 @opindex Wredundant-decls
3301 Warn if anything is declared more than once in the same scope, even in
3302 cases where multiple declaration is valid and changes nothing.
3304 @item -Wnested-externs @r{(C only)}
3305 @opindex Wnested-externs
3306 Warn if an @code{extern} declaration is encountered within a function.
3308 @item -Wunreachable-code
3309 @opindex Wunreachable-code
3310 Warn if the compiler detects that code will never be executed.
3312 This option is intended to warn when the compiler detects that at
3313 least a whole line of source code will never be executed, because
3314 some condition is never satisfied or because it is after a
3315 procedure that never returns.
3317 It is possible for this option to produce a warning even though there
3318 are circumstances under which part of the affected line can be executed,
3319 so care should be taken when removing apparently-unreachable code.
3321 For instance, when a function is inlined, a warning may mean that the
3322 line is unreachable in only one inlined copy of the function.
3324 This option is not made part of @option{-Wall} because in a debugging
3325 version of a program there is often substantial code which checks
3326 correct functioning of the program and is, hopefully, unreachable
3327 because the program does work. Another common use of unreachable
3328 code is to provide behavior which is selectable at compile-time.
3332 Warn if a function can not be inlined and it was declared as inline.
3333 Even with this option, the compiler will not warn about failures to
3334 inline functions declared in system headers.
3336 The compiler uses a variety of heuristics to determine whether or not
3337 to inline a function. For example, the compiler takes into account
3338 the size of the function being inlined and the amount of inlining
3339 that has already been done in the current function. Therefore,
3340 seemingly insignificant changes in the source program can cause the
3341 warnings produced by @option{-Winline} to appear or disappear.
3343 @item -Wno-invalid-offsetof @r{(C++ only)}
3344 @opindex Wno-invalid-offsetof
3345 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3346 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3347 to a non-POD type is undefined. In existing C++ implementations,
3348 however, @samp{offsetof} typically gives meaningful results even when
3349 applied to certain kinds of non-POD types. (Such as a simple
3350 @samp{struct} that fails to be a POD type only by virtue of having a
3351 constructor.) This flag is for users who are aware that they are
3352 writing nonportable code and who have deliberately chosen to ignore the
3355 The restrictions on @samp{offsetof} may be relaxed in a future version
3356 of the C++ standard.
3358 @item -Wno-int-to-pointer-cast @r{(C only)}
3359 @opindex Wno-int-to-pointer-cast
3360 Suppress warnings from casts to pointer type of an integer of a
3363 @item -Wno-pointer-to-int-cast @r{(C only)}
3364 @opindex Wno-pointer-to-int-cast
3365 Suppress warnings from casts from a pointer to an integer type of a
3369 @opindex Winvalid-pch
3370 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3371 the search path but can't be used.
3375 @opindex Wno-long-long
3376 Warn if @samp{long long} type is used. This is default. To inhibit
3377 the warning messages, use @option{-Wno-long-long}. Flags
3378 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3379 only when @option{-pedantic} flag is used.
3381 @item -Wvariadic-macros
3382 @opindex Wvariadic-macros
3383 @opindex Wno-variadic-macros
3384 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3385 alternate syntax when in pedantic ISO C99 mode. This is default.
3386 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3388 @item -Wvolatile-register-var
3389 @opindex Wvolatile-register-var
3390 @opindex Wno-volatile-register-var
3391 Warn if a register variable is declared volatile. The volatile
3392 modifier does not inhibit all optimizations that may eliminate reads
3393 and/or writes to register variables.
3395 @item -Wdisabled-optimization
3396 @opindex Wdisabled-optimization
3397 Warn if a requested optimization pass is disabled. This warning does
3398 not generally indicate that there is anything wrong with your code; it
3399 merely indicates that GCC's optimizers were unable to handle the code
3400 effectively. Often, the problem is that your code is too big or too
3401 complex; GCC will refuse to optimize programs when the optimization
3402 itself is likely to take inordinate amounts of time.
3404 @item -Wno-pointer-sign
3405 @opindex Wno-pointer-sign
3406 Don't warn for pointer argument passing or assignment with different signedness.
3407 Only useful in the negative form since this warning is enabled by default.
3408 This option is only supported for C and Objective-C@.
3412 Make all warnings into errors.
3414 @item -Wstack-protector
3415 @opindex Wstack-protector
3416 This option is only active when @option{-fstack-protector} is active. It
3417 warns about functions that will not be protected against stack smashing.
3419 @item -Wstring-literal-comparison
3420 @opindex Wstring-literal-comparison
3421 Warn about suspicious comparisons to string literal constants. In C,
3422 direct comparisons against the memory address of a string literal, such
3423 as @code{if (x == "abc")}, typically indicate a programmer error, and
3424 even when intentional, result in unspecified behavior and are not portable.
3425 Usually these warnings alert that the programmer intended to use
3426 @code{strcmp}. This warning is enabled by @option{-Wall}.
3430 @node Debugging Options
3431 @section Options for Debugging Your Program or GCC
3432 @cindex options, debugging
3433 @cindex debugging information options
3435 GCC has various special options that are used for debugging
3436 either your program or GCC:
3441 Produce debugging information in the operating system's native format
3442 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3445 On most systems that use stabs format, @option{-g} enables use of extra
3446 debugging information that only GDB can use; this extra information
3447 makes debugging work better in GDB but will probably make other debuggers
3449 refuse to read the program. If you want to control for certain whether
3450 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3451 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3453 GCC allows you to use @option{-g} with
3454 @option{-O}. The shortcuts taken by optimized code may occasionally
3455 produce surprising results: some variables you declared may not exist
3456 at all; flow of control may briefly move where you did not expect it;
3457 some statements may not be executed because they compute constant
3458 results or their values were already at hand; some statements may
3459 execute in different places because they were moved out of loops.
3461 Nevertheless it proves possible to debug optimized output. This makes
3462 it reasonable to use the optimizer for programs that might have bugs.
3464 The following options are useful when GCC is generated with the
3465 capability for more than one debugging format.
3469 Produce debugging information for use by GDB@. This means to use the
3470 most expressive format available (DWARF 2, stabs, or the native format
3471 if neither of those are supported), including GDB extensions if at all
3476 Produce debugging information in stabs format (if that is supported),
3477 without GDB extensions. This is the format used by DBX on most BSD
3478 systems. On MIPS, Alpha and System V Release 4 systems this option
3479 produces stabs debugging output which is not understood by DBX or SDB@.
3480 On System V Release 4 systems this option requires the GNU assembler.
3482 @item -feliminate-unused-debug-symbols
3483 @opindex feliminate-unused-debug-symbols
3484 Produce debugging information in stabs format (if that is supported),
3485 for only symbols that are actually used.
3489 Produce debugging information in stabs format (if that is supported),
3490 using GNU extensions understood only by the GNU debugger (GDB)@. The
3491 use of these extensions is likely to make other debuggers crash or
3492 refuse to read the program.
3496 Produce debugging information in COFF format (if that is supported).
3497 This is the format used by SDB on most System V systems prior to
3502 Produce debugging information in XCOFF format (if that is supported).
3503 This is the format used by the DBX debugger on IBM RS/6000 systems.
3507 Produce debugging information in XCOFF format (if that is supported),
3508 using GNU extensions understood only by the GNU debugger (GDB)@. The
3509 use of these extensions is likely to make other debuggers crash or
3510 refuse to read the program, and may cause assemblers other than the GNU
3511 assembler (GAS) to fail with an error.
3515 Produce debugging information in DWARF version 2 format (if that is
3516 supported). This is the format used by DBX on IRIX 6. With this
3517 option, GCC uses features of DWARF version 3 when they are useful;
3518 version 3 is upward compatible with version 2, but may still cause
3519 problems for older debuggers.
3523 Produce debugging information in VMS debug format (if that is
3524 supported). This is the format used by DEBUG on VMS systems.
3527 @itemx -ggdb@var{level}
3528 @itemx -gstabs@var{level}
3529 @itemx -gcoff@var{level}
3530 @itemx -gxcoff@var{level}
3531 @itemx -gvms@var{level}
3532 Request debugging information and also use @var{level} to specify how
3533 much information. The default level is 2.
3535 Level 1 produces minimal information, enough for making backtraces in
3536 parts of the program that you don't plan to debug. This includes
3537 descriptions of functions and external variables, but no information
3538 about local variables and no line numbers.
3540 Level 3 includes extra information, such as all the macro definitions
3541 present in the program. Some debuggers support macro expansion when
3542 you use @option{-g3}.
3544 @option{-gdwarf-2} does not accept a concatenated debug level, because
3545 GCC used to support an option @option{-gdwarf} that meant to generate
3546 debug information in version 1 of the DWARF format (which is very
3547 different from version 2), and it would have been too confusing. That
3548 debug format is long obsolete, but the option cannot be changed now.
3549 Instead use an additional @option{-g@var{level}} option to change the
3550 debug level for DWARF2.
3552 @item -feliminate-dwarf2-dups
3553 @opindex feliminate-dwarf2-dups
3554 Compress DWARF2 debugging information by eliminating duplicated
3555 information about each symbol. This option only makes sense when
3556 generating DWARF2 debugging information with @option{-gdwarf-2}.
3558 @cindex @command{prof}
3561 Generate extra code to write profile information suitable for the
3562 analysis program @command{prof}. You must use this option when compiling
3563 the source files you want data about, and you must also use it when
3566 @cindex @command{gprof}
3569 Generate extra code to write profile information suitable for the
3570 analysis program @command{gprof}. You must use this option when compiling
3571 the source files you want data about, and you must also use it when
3576 Makes the compiler print out each function name as it is compiled, and
3577 print some statistics about each pass when it finishes.
3580 @opindex ftime-report
3581 Makes the compiler print some statistics about the time consumed by each
3582 pass when it finishes.
3585 @opindex fmem-report
3586 Makes the compiler print some statistics about permanent memory
3587 allocation when it finishes.
3589 @item -fprofile-arcs
3590 @opindex fprofile-arcs
3591 Add code so that program flow @dfn{arcs} are instrumented. During
3592 execution the program records how many times each branch and call is
3593 executed and how many times it is taken or returns. When the compiled
3594 program exits it saves this data to a file called
3595 @file{@var{auxname}.gcda} for each source file. The data may be used for
3596 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3597 test coverage analysis (@option{-ftest-coverage}). Each object file's
3598 @var{auxname} is generated from the name of the output file, if
3599 explicitly specified and it is not the final executable, otherwise it is
3600 the basename of the source file. In both cases any suffix is removed
3601 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3602 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3603 @xref{Cross-profiling}.
3605 @cindex @command{gcov}
3609 This option is used to compile and link code instrumented for coverage
3610 analysis. The option is a synonym for @option{-fprofile-arcs}
3611 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3612 linking). See the documentation for those options for more details.
3617 Compile the source files with @option{-fprofile-arcs} plus optimization
3618 and code generation options. For test coverage analysis, use the
3619 additional @option{-ftest-coverage} option. You do not need to profile
3620 every source file in a program.
3623 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3624 (the latter implies the former).
3627 Run the program on a representative workload to generate the arc profile
3628 information. This may be repeated any number of times. You can run
3629 concurrent instances of your program, and provided that the file system
3630 supports locking, the data files will be correctly updated. Also
3631 @code{fork} calls are detected and correctly handled (double counting
3635 For profile-directed optimizations, compile the source files again with
3636 the same optimization and code generation options plus
3637 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3638 Control Optimization}).
3641 For test coverage analysis, use @command{gcov} to produce human readable
3642 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3643 @command{gcov} documentation for further information.
3647 With @option{-fprofile-arcs}, for each function of your program GCC
3648 creates a program flow graph, then finds a spanning tree for the graph.
3649 Only arcs that are not on the spanning tree have to be instrumented: the
3650 compiler adds code to count the number of times that these arcs are
3651 executed. When an arc is the only exit or only entrance to a block, the
3652 instrumentation code can be added to the block; otherwise, a new basic
3653 block must be created to hold the instrumentation code.
3656 @item -ftest-coverage
3657 @opindex ftest-coverage
3658 Produce a notes file that the @command{gcov} code-coverage utility
3659 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3660 show program coverage. Each source file's note file is called
3661 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3662 above for a description of @var{auxname} and instructions on how to
3663 generate test coverage data. Coverage data will match the source files
3664 more closely, if you do not optimize.
3666 @item -d@var{letters}
3667 @item -fdump-rtl-@var{pass}
3669 Says to make debugging dumps during compilation at times specified by
3670 @var{letters}. This is used for debugging the RTL-based passes of the
3671 compiler. The file names for most of the dumps are made by appending a
3672 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3673 from the name of the output file, if explicitly specified and it is not
3674 an executable, otherwise it is the basename of the source file.
3676 Most debug dumps can be enabled either passing a letter to the @option{-d}
3677 option, or with a long @option{-fdump-rtl} switch; here are the possible
3678 letters for use in @var{letters} and @var{pass}, and their meanings:
3683 Annotate the assembler output with miscellaneous debugging information.
3686 @itemx -fdump-rtl-bp
3688 @opindex fdump-rtl-bp
3689 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3692 @itemx -fdump-rtl-bbro
3694 @opindex fdump-rtl-bbro
3695 Dump after block reordering, to @file{@var{file}.30.bbro}.
3698 @itemx -fdump-rtl-combine
3700 @opindex fdump-rtl-combine
3701 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3704 @itemx -fdump-rtl-ce1
3705 @itemx -fdump-rtl-ce2
3707 @opindex fdump-rtl-ce1
3708 @opindex fdump-rtl-ce2
3709 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3710 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3711 and @option{-fdump-rtl-ce2} enable dumping after the second if
3712 conversion, to the file @file{@var{file}.18.ce2}.
3715 @itemx -fdump-rtl-btl
3716 @itemx -fdump-rtl-dbr
3718 @opindex fdump-rtl-btl
3719 @opindex fdump-rtl-dbr
3720 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3721 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3722 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3723 scheduling, to @file{@var{file}.36.dbr}.
3727 Dump all macro definitions, at the end of preprocessing, in addition to
3731 @itemx -fdump-rtl-ce3
3733 @opindex fdump-rtl-ce3
3734 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3737 @itemx -fdump-rtl-cfg
3738 @itemx -fdump-rtl-life
3740 @opindex fdump-rtl-cfg
3741 @opindex fdump-rtl-life
3742 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3743 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3744 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3745 to @file{@var{file}.16.life}.
3748 @itemx -fdump-rtl-greg
3750 @opindex fdump-rtl-greg
3751 Dump after global register allocation, to @file{@var{file}.23.greg}.
3754 @itemx -fdump-rtl-gcse
3755 @itemx -fdump-rtl-bypass
3757 @opindex fdump-rtl-gcse
3758 @opindex fdump-rtl-bypass
3759 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3760 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3761 enable dumping after jump bypassing and control flow optimizations, to
3762 @file{@var{file}.07.bypass}.
3765 @itemx -fdump-rtl-eh
3767 @opindex fdump-rtl-eh
3768 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3771 @itemx -fdump-rtl-sibling
3773 @opindex fdump-rtl-sibling
3774 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3777 @itemx -fdump-rtl-jump
3779 @opindex fdump-rtl-jump
3780 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3783 @itemx -fdump-rtl-stack
3785 @opindex fdump-rtl-stack
3786 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3789 @itemx -fdump-rtl-lreg
3791 @opindex fdump-rtl-lreg
3792 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3795 @itemx -fdump-rtl-loop
3796 @itemx -fdump-rtl-loop2
3798 @opindex fdump-rtl-loop
3799 @opindex fdump-rtl-loop2
3800 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3801 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3802 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3803 @file{@var{file}.13.loop2}.
3806 @itemx -fdump-rtl-sms
3808 @opindex fdump-rtl-sms
3809 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3812 @itemx -fdump-rtl-mach
3814 @opindex fdump-rtl-mach
3815 Dump after performing the machine dependent reorganization pass, to
3816 @file{@var{file}.35.mach}.
3819 @itemx -fdump-rtl-rnreg
3821 @opindex fdump-rtl-rnreg
3822 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3825 @itemx -fdump-rtl-regmove
3827 @opindex fdump-rtl-regmove
3828 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3831 @itemx -fdump-rtl-postreload
3833 @opindex fdump-rtl-postreload
3834 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3837 @itemx -fdump-rtl-expand
3839 @opindex fdump-rtl-expand
3840 Dump after RTL generation, to @file{@var{file}.00.expand}.
3843 @itemx -fdump-rtl-sched2
3845 @opindex fdump-rtl-sched2
3846 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3849 @itemx -fdump-rtl-cse
3851 @opindex fdump-rtl-cse
3852 Dump after CSE (including the jump optimization that sometimes follows
3853 CSE), to @file{@var{file}.04.cse}.
3856 @itemx -fdump-rtl-sched
3858 @opindex fdump-rtl-sched
3859 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3862 @itemx -fdump-rtl-cse2
3864 @opindex fdump-rtl-cse2
3865 Dump after the second CSE pass (including the jump optimization that
3866 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3869 @itemx -fdump-rtl-tracer
3871 @opindex fdump-rtl-tracer
3872 Dump after running tracer, to @file{@var{file}.12.tracer}.
3875 @itemx -fdump-rtl-vpt
3876 @itemx -fdump-rtl-vartrack
3878 @opindex fdump-rtl-vpt
3879 @opindex fdump-rtl-vartrack
3880 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3881 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3882 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3883 to @file{@var{file}.34.vartrack}.
3886 @itemx -fdump-rtl-flow2
3888 @opindex fdump-rtl-flow2
3889 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3892 @itemx -fdump-rtl-peephole2
3894 @opindex fdump-rtl-peephole2
3895 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3898 @itemx -fdump-rtl-web
3900 @opindex fdump-rtl-web
3901 Dump after live range splitting, to @file{@var{file}.14.web}.
3904 @itemx -fdump-rtl-all
3906 @opindex fdump-rtl-all
3907 Produce all the dumps listed above.
3911 Produce a core dump whenever an error occurs.
3915 Print statistics on memory usage, at the end of the run, to
3920 Annotate the assembler output with a comment indicating which
3921 pattern and alternative was used. The length of each instruction is
3926 Dump the RTL in the assembler output as a comment before each instruction.
3927 Also turns on @option{-dp} annotation.
3931 For each of the other indicated dump files (either with @option{-d} or
3932 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3933 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3937 Just generate RTL for a function instead of compiling it. Usually used
3938 with @samp{r} (@option{-fdump-rtl-expand}).
3942 Dump debugging information during parsing, to standard error.
3945 @item -fdump-unnumbered
3946 @opindex fdump-unnumbered
3947 When doing debugging dumps (see @option{-d} option above), suppress instruction
3948 numbers and line number note output. This makes it more feasible to
3949 use diff on debugging dumps for compiler invocations with different
3950 options, in particular with and without @option{-g}.
3952 @item -fdump-translation-unit @r{(C++ only)}
3953 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3954 @opindex fdump-translation-unit
3955 Dump a representation of the tree structure for the entire translation
3956 unit to a file. The file name is made by appending @file{.tu} to the
3957 source file name. If the @samp{-@var{options}} form is used, @var{options}
3958 controls the details of the dump as described for the
3959 @option{-fdump-tree} options.
3961 @item -fdump-class-hierarchy @r{(C++ only)}
3962 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3963 @opindex fdump-class-hierarchy
3964 Dump a representation of each class's hierarchy and virtual function
3965 table layout to a file. The file name is made by appending @file{.class}
3966 to the source file name. If the @samp{-@var{options}} form is used,
3967 @var{options} controls the details of the dump as described for the
3968 @option{-fdump-tree} options.
3970 @item -fdump-ipa-@var{switch}
3972 Control the dumping at various stages of inter-procedural analysis
3973 language tree to a file. The file name is generated by appending a switch
3974 specific suffix to the source file name. The following dumps are possible:
3978 Enables all inter-procedural analysis dumps; currently the only produced
3979 dump is the @samp{cgraph} dump.
3982 Dumps information about call-graph optimization, unused function removal,
3983 and inlining decisions.
3986 @item -fdump-tree-@var{switch}
3987 @itemx -fdump-tree-@var{switch}-@var{options}
3989 Control the dumping at various stages of processing the intermediate
3990 language tree to a file. The file name is generated by appending a switch
3991 specific suffix to the source file name. If the @samp{-@var{options}}
3992 form is used, @var{options} is a list of @samp{-} separated options that
3993 control the details of the dump. Not all options are applicable to all
3994 dumps, those which are not meaningful will be ignored. The following
3995 options are available
3999 Print the address of each node. Usually this is not meaningful as it
4000 changes according to the environment and source file. Its primary use
4001 is for tying up a dump file with a debug environment.
4003 Inhibit dumping of members of a scope or body of a function merely
4004 because that scope has been reached. Only dump such items when they
4005 are directly reachable by some other path. When dumping pretty-printed
4006 trees, this option inhibits dumping the bodies of control structures.
4008 Print a raw representation of the tree. By default, trees are
4009 pretty-printed into a C-like representation.
4011 Enable more detailed dumps (not honored by every dump option).
4013 Enable dumping various statistics about the pass (not honored by every dump
4016 Enable showing basic block boundaries (disabled in raw dumps).
4018 Enable showing virtual operands for every statement.
4020 Enable showing line numbers for statements.
4022 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4024 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4027 The following tree dumps are possible:
4031 Dump before any tree based optimization, to @file{@var{file}.original}.
4034 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4037 Dump after function inlining, to @file{@var{file}.inlined}.
4040 @opindex fdump-tree-gimple
4041 Dump each function before and after the gimplification pass to a file. The
4042 file name is made by appending @file{.gimple} to the source file name.
4045 @opindex fdump-tree-cfg
4046 Dump the control flow graph of each function to a file. The file name is
4047 made by appending @file{.cfg} to the source file name.
4050 @opindex fdump-tree-vcg
4051 Dump the control flow graph of each function to a file in VCG format. The
4052 file name is made by appending @file{.vcg} to the source file name. Note
4053 that if the file contains more than one function, the generated file cannot
4054 be used directly by VCG@. You will need to cut and paste each function's
4055 graph into its own separate file first.
4058 @opindex fdump-tree-ch
4059 Dump each function after copying loop headers. The file name is made by
4060 appending @file{.ch} to the source file name.
4063 @opindex fdump-tree-ssa
4064 Dump SSA related information to a file. The file name is made by appending
4065 @file{.ssa} to the source file name.
4068 @opindex fdump-tree-salias
4069 Dump structure aliasing variable information to a file. This file name
4070 is made by appending @file{.salias} to the source file name.
4073 @opindex fdump-tree-alias
4074 Dump aliasing information for each function. The file name is made by
4075 appending @file{.alias} to the source file name.
4078 @opindex fdump-tree-ccp
4079 Dump each function after CCP@. The file name is made by appending
4080 @file{.ccp} to the source file name.
4083 @opindex fdump-tree-storeccp
4084 Dump each function after STORE-CCP. The file name is made by appending
4085 @file{.storeccp} to the source file name.
4088 @opindex fdump-tree-pre
4089 Dump trees after partial redundancy elimination. The file name is made
4090 by appending @file{.pre} to the source file name.
4093 @opindex fdump-tree-fre
4094 Dump trees after full redundancy elimination. The file name is made
4095 by appending @file{.fre} to the source file name.
4098 @opindex fdump-tree-copyprop
4099 Dump trees after copy propagation. The file name is made
4100 by appending @file{.copyprop} to the source file name.
4102 @item store_copyprop
4103 @opindex fdump-tree-store_copyprop
4104 Dump trees after store copy-propagation. The file name is made
4105 by appending @file{.store_copyprop} to the source file name.
4108 @opindex fdump-tree-dce
4109 Dump each function after dead code elimination. The file name is made by
4110 appending @file{.dce} to the source file name.
4113 @opindex fdump-tree-mudflap
4114 Dump each function after adding mudflap instrumentation. The file name is
4115 made by appending @file{.mudflap} to the source file name.
4118 @opindex fdump-tree-sra
4119 Dump each function after performing scalar replacement of aggregates. The
4120 file name is made by appending @file{.sra} to the source file name.
4123 @opindex fdump-tree-sink
4124 Dump each function after performing code sinking. The file name is made
4125 by appending @file{.sink} to the source file name.
4128 @opindex fdump-tree-dom
4129 Dump each function after applying dominator tree optimizations. The file
4130 name is made by appending @file{.dom} to the source file name.
4133 @opindex fdump-tree-dse
4134 Dump each function after applying dead store elimination. The file
4135 name is made by appending @file{.dse} to the source file name.
4138 @opindex fdump-tree-phiopt
4139 Dump each function after optimizing PHI nodes into straightline code. The file
4140 name is made by appending @file{.phiopt} to the source file name.
4143 @opindex fdump-tree-forwprop
4144 Dump each function after forward propagating single use variables. The file
4145 name is made by appending @file{.forwprop} to the source file name.
4148 @opindex fdump-tree-copyrename
4149 Dump each function after applying the copy rename optimization. The file
4150 name is made by appending @file{.copyrename} to the source file name.
4153 @opindex fdump-tree-nrv
4154 Dump each function after applying the named return value optimization on
4155 generic trees. The file name is made by appending @file{.nrv} to the source
4159 @opindex fdump-tree-vect
4160 Dump each function after applying vectorization of loops. The file name is
4161 made by appending @file{.vect} to the source file name.
4164 @opindex fdump-tree-vrp
4165 Dump each function after Value Range Propagation (VRP). The file name
4166 is made by appending @file{.vrp} to the source file name.
4169 @opindex fdump-tree-all
4170 Enable all the available tree dumps with the flags provided in this option.
4173 @item -ftree-vectorizer-verbose=@var{n}
4174 @opindex ftree-vectorizer-verbose
4175 This option controls the amount of debugging output the vectorizer prints.
4176 This information is written to standard error, unless @option{-fdump-tree-all}
4177 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4178 usual dump listing file, @file{.vect}.
4180 @item -frandom-seed=@var{string}
4181 @opindex frandom-string
4182 This option provides a seed that GCC uses when it would otherwise use
4183 random numbers. It is used to generate certain symbol names
4184 that have to be different in every compiled file. It is also used to
4185 place unique stamps in coverage data files and the object files that
4186 produce them. You can use the @option{-frandom-seed} option to produce
4187 reproducibly identical object files.
4189 The @var{string} should be different for every file you compile.
4191 @item -fsched-verbose=@var{n}
4192 @opindex fsched-verbose
4193 On targets that use instruction scheduling, this option controls the
4194 amount of debugging output the scheduler prints. This information is
4195 written to standard error, unless @option{-dS} or @option{-dR} is
4196 specified, in which case it is output to the usual dump
4197 listing file, @file{.sched} or @file{.sched2} respectively. However
4198 for @var{n} greater than nine, the output is always printed to standard
4201 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4202 same information as @option{-dRS}. For @var{n} greater than one, it
4203 also output basic block probabilities, detailed ready list information
4204 and unit/insn info. For @var{n} greater than two, it includes RTL
4205 at abort point, control-flow and regions info. And for @var{n} over
4206 four, @option{-fsched-verbose} also includes dependence info.
4210 Store the usual ``temporary'' intermediate files permanently; place them
4211 in the current directory and name them based on the source file. Thus,
4212 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4213 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4214 preprocessed @file{foo.i} output file even though the compiler now
4215 normally uses an integrated preprocessor.
4217 When used in combination with the @option{-x} command line option,
4218 @option{-save-temps} is sensible enough to avoid over writing an
4219 input source file with the same extension as an intermediate file.
4220 The corresponding intermediate file may be obtained by renaming the
4221 source file before using @option{-save-temps}.
4225 Report the CPU time taken by each subprocess in the compilation
4226 sequence. For C source files, this is the compiler proper and assembler
4227 (plus the linker if linking is done). The output looks like this:
4234 The first number on each line is the ``user time'', that is time spent
4235 executing the program itself. The second number is ``system time'',
4236 time spent executing operating system routines on behalf of the program.
4237 Both numbers are in seconds.
4239 @item -fvar-tracking
4240 @opindex fvar-tracking
4241 Run variable tracking pass. It computes where variables are stored at each
4242 position in code. Better debugging information is then generated
4243 (if the debugging information format supports this information).
4245 It is enabled by default when compiling with optimization (@option{-Os},
4246 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4247 the debug info format supports it.
4249 @item -print-file-name=@var{library}
4250 @opindex print-file-name
4251 Print the full absolute name of the library file @var{library} that
4252 would be used when linking---and don't do anything else. With this
4253 option, GCC does not compile or link anything; it just prints the
4256 @item -print-multi-directory
4257 @opindex print-multi-directory
4258 Print the directory name corresponding to the multilib selected by any
4259 other switches present in the command line. This directory is supposed
4260 to exist in @env{GCC_EXEC_PREFIX}.
4262 @item -print-multi-lib
4263 @opindex print-multi-lib
4264 Print the mapping from multilib directory names to compiler switches
4265 that enable them. The directory name is separated from the switches by
4266 @samp{;}, and each switch starts with an @samp{@@} instead of the
4267 @samp{-}, without spaces between multiple switches. This is supposed to
4268 ease shell-processing.
4270 @item -print-prog-name=@var{program}
4271 @opindex print-prog-name
4272 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4274 @item -print-libgcc-file-name
4275 @opindex print-libgcc-file-name
4276 Same as @option{-print-file-name=libgcc.a}.
4278 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4279 but you do want to link with @file{libgcc.a}. You can do
4282 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4285 @item -print-search-dirs
4286 @opindex print-search-dirs
4287 Print the name of the configured installation directory and a list of
4288 program and library directories @command{gcc} will search---and don't do anything else.
4290 This is useful when @command{gcc} prints the error message
4291 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4292 To resolve this you either need to put @file{cpp0} and the other compiler
4293 components where @command{gcc} expects to find them, or you can set the environment
4294 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4295 Don't forget the trailing @samp{/}.
4296 @xref{Environment Variables}.
4299 @opindex dumpmachine
4300 Print the compiler's target machine (for example,
4301 @samp{i686-pc-linux-gnu})---and don't do anything else.
4304 @opindex dumpversion
4305 Print the compiler version (for example, @samp{3.0})---and don't do
4310 Print the compiler's built-in specs---and don't do anything else. (This
4311 is used when GCC itself is being built.) @xref{Spec Files}.
4313 @item -feliminate-unused-debug-types
4314 @opindex feliminate-unused-debug-types
4315 Normally, when producing DWARF2 output, GCC will emit debugging
4316 information for all types declared in a compilation
4317 unit, regardless of whether or not they are actually used
4318 in that compilation unit. Sometimes this is useful, such as
4319 if, in the debugger, you want to cast a value to a type that is
4320 not actually used in your program (but is declared). More often,
4321 however, this results in a significant amount of wasted space.
4322 With this option, GCC will avoid producing debug symbol output
4323 for types that are nowhere used in the source file being compiled.
4326 @node Optimize Options
4327 @section Options That Control Optimization
4328 @cindex optimize options
4329 @cindex options, optimization
4331 These options control various sorts of optimizations.
4333 Without any optimization option, the compiler's goal is to reduce the
4334 cost of compilation and to make debugging produce the expected
4335 results. Statements are independent: if you stop the program with a
4336 breakpoint between statements, you can then assign a new value to any
4337 variable or change the program counter to any other statement in the
4338 function and get exactly the results you would expect from the source
4341 Turning on optimization flags makes the compiler attempt to improve
4342 the performance and/or code size at the expense of compilation time
4343 and possibly the ability to debug the program.
4345 The compiler performs optimization based on the knowledge it has of
4346 the program. Optimization levels @option{-O2} and above, in
4347 particular, enable @emph{unit-at-a-time} mode, which allows the
4348 compiler to consider information gained from later functions in
4349 the file when compiling a function. Compiling multiple files at
4350 once to a single output file in @emph{unit-at-a-time} mode allows
4351 the compiler to use information gained from all of the files when
4352 compiling each of them.
4354 Not all optimizations are controlled directly by a flag. Only
4355 optimizations that have a flag are listed.
4362 Optimize. Optimizing compilation takes somewhat more time, and a lot
4363 more memory for a large function.
4365 With @option{-O}, the compiler tries to reduce code size and execution
4366 time, without performing any optimizations that take a great deal of
4369 @option{-O} turns on the following optimization flags:
4370 @gccoptlist{-fdefer-pop @gol
4371 -fdelayed-branch @gol
4372 -fguess-branch-probability @gol
4373 -fcprop-registers @gol
4374 -floop-optimize @gol
4375 -fif-conversion @gol
4376 -fif-conversion2 @gol
4379 -ftree-dominator-opts @gol
4384 -ftree-copyrename @gol
4389 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4390 where doing so does not interfere with debugging.
4394 Optimize even more. GCC performs nearly all supported optimizations
4395 that do not involve a space-speed tradeoff. The compiler does not
4396 perform loop unrolling or function inlining when you specify @option{-O2}.
4397 As compared to @option{-O}, this option increases both compilation time
4398 and the performance of the generated code.
4400 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4401 also turns on the following optimization flags:
4402 @gccoptlist{-fthread-jumps @gol
4404 -foptimize-sibling-calls @gol
4405 -fcse-follow-jumps -fcse-skip-blocks @gol
4406 -fgcse -fgcse-lm @gol
4407 -fexpensive-optimizations @gol
4408 -fstrength-reduce @gol
4409 -frerun-cse-after-loop -frerun-loop-opt @gol
4412 -fschedule-insns -fschedule-insns2 @gol
4413 -fsched-interblock -fsched-spec @gol
4415 -fstrict-aliasing @gol
4416 -fdelete-null-pointer-checks @gol
4417 -freorder-blocks -freorder-functions @gol
4418 -funit-at-a-time @gol
4419 -falign-functions -falign-jumps @gol
4420 -falign-loops -falign-labels @gol
4424 Please note the warning under @option{-fgcse} about
4425 invoking @option{-O2} on programs that use computed gotos.
4429 Optimize yet more. @option{-O3} turns on all optimizations specified by
4430 @option{-O2} and also turns on the @option{-finline-functions},
4431 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4435 Do not optimize. This is the default.
4439 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4440 do not typically increase code size. It also performs further
4441 optimizations designed to reduce code size.
4443 @option{-Os} disables the following optimization flags:
4444 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4445 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4446 -fprefetch-loop-arrays -ftree-vect-loop-version}
4448 If you use multiple @option{-O} options, with or without level numbers,
4449 the last such option is the one that is effective.
4452 Options of the form @option{-f@var{flag}} specify machine-independent
4453 flags. Most flags have both positive and negative forms; the negative
4454 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4455 below, only one of the forms is listed---the one you typically will
4456 use. You can figure out the other form by either removing @samp{no-}
4459 The following options control specific optimizations. They are either
4460 activated by @option{-O} options or are related to ones that are. You
4461 can use the following flags in the rare cases when ``fine-tuning'' of
4462 optimizations to be performed is desired.
4465 @item -fno-default-inline
4466 @opindex fno-default-inline
4467 Do not make member functions inline by default merely because they are
4468 defined inside the class scope (C++ only). Otherwise, when you specify
4469 @w{@option{-O}}, member functions defined inside class scope are compiled
4470 inline by default; i.e., you don't need to add @samp{inline} in front of
4471 the member function name.
4473 @item -fno-defer-pop
4474 @opindex fno-defer-pop
4475 Always pop the arguments to each function call as soon as that function
4476 returns. For machines which must pop arguments after a function call,
4477 the compiler normally lets arguments accumulate on the stack for several
4478 function calls and pops them all at once.
4480 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4484 Force memory operands to be copied into registers before doing
4485 arithmetic on them. This produces better code by making all memory
4486 references potential common subexpressions. When they are not common
4487 subexpressions, instruction combination should eliminate the separate
4488 register-load. This option is now a nop and will be removed in 4.2.
4491 @opindex fforce-addr
4492 Force memory address constants to be copied into registers before
4493 doing arithmetic on them.
4495 @item -fomit-frame-pointer
4496 @opindex fomit-frame-pointer
4497 Don't keep the frame pointer in a register for functions that
4498 don't need one. This avoids the instructions to save, set up and
4499 restore frame pointers; it also makes an extra register available
4500 in many functions. @strong{It also makes debugging impossible on
4503 On some machines, such as the VAX, this flag has no effect, because
4504 the standard calling sequence automatically handles the frame pointer
4505 and nothing is saved by pretending it doesn't exist. The
4506 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4507 whether a target machine supports this flag. @xref{Registers,,Register
4508 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4510 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4512 @item -foptimize-sibling-calls
4513 @opindex foptimize-sibling-calls
4514 Optimize sibling and tail recursive calls.
4516 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4520 Don't pay attention to the @code{inline} keyword. Normally this option
4521 is used to keep the compiler from expanding any functions inline.
4522 Note that if you are not optimizing, no functions can be expanded inline.
4524 @item -finline-functions
4525 @opindex finline-functions
4526 Integrate all simple functions into their callers. The compiler
4527 heuristically decides which functions are simple enough to be worth
4528 integrating in this way.
4530 If all calls to a given function are integrated, and the function is
4531 declared @code{static}, then the function is normally not output as
4532 assembler code in its own right.
4534 Enabled at level @option{-O3}.
4536 @item -finline-functions-called-once
4537 @opindex finline-functions-called-once
4538 Consider all @code{static} functions called once for inlining into their
4539 caller even if they are not marked @code{inline}. If a call to a given
4540 function is integrated, then the function is not output as assembler code
4543 Enabled if @option{-funit-at-a-time} is enabled.
4545 @item -fearly-inlining
4546 @opindex fearly-inlining
4547 Inline functions marked by @code{always_inline} and functions whose body seems
4548 smaller than the function call overhead early before doing
4549 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4550 makes profiling significantly cheaper and usually inlining faster on programs
4551 having large chains of nested wrapper functions.
4555 @item -finline-limit=@var{n}
4556 @opindex finline-limit
4557 By default, GCC limits the size of functions that can be inlined. This flag
4558 allows the control of this limit for functions that are explicitly marked as
4559 inline (i.e., marked with the inline keyword or defined within the class
4560 definition in c++). @var{n} is the size of functions that can be inlined in
4561 number of pseudo instructions (not counting parameter handling). The default
4562 value of @var{n} is 600.
4563 Increasing this value can result in more inlined code at
4564 the cost of compilation time and memory consumption. Decreasing usually makes
4565 the compilation faster and less code will be inlined (which presumably
4566 means slower programs). This option is particularly useful for programs that
4567 use inlining heavily such as those based on recursive templates with C++.
4569 Inlining is actually controlled by a number of parameters, which may be
4570 specified individually by using @option{--param @var{name}=@var{value}}.
4571 The @option{-finline-limit=@var{n}} option sets some of these parameters
4575 @item max-inline-insns-single
4576 is set to @var{n}/2.
4577 @item max-inline-insns-auto
4578 is set to @var{n}/2.
4579 @item min-inline-insns
4580 is set to 130 or @var{n}/4, whichever is smaller.
4581 @item max-inline-insns-rtl
4585 See below for a documentation of the individual
4586 parameters controlling inlining.
4588 @emph{Note:} pseudo instruction represents, in this particular context, an
4589 abstract measurement of function's size. In no way does it represent a count
4590 of assembly instructions and as such its exact meaning might change from one
4591 release to an another.
4593 @item -fkeep-inline-functions
4594 @opindex fkeep-inline-functions
4595 In C, emit @code{static} functions that are declared @code{inline}
4596 into the object file, even if the function has been inlined into all
4597 of its callers. This switch does not affect functions using the
4598 @code{extern inline} extension in GNU C@. In C++, emit any and all
4599 inline functions into the object file.
4601 @item -fkeep-static-consts
4602 @opindex fkeep-static-consts
4603 Emit variables declared @code{static const} when optimization isn't turned
4604 on, even if the variables aren't referenced.
4606 GCC enables this option by default. If you want to force the compiler to
4607 check if the variable was referenced, regardless of whether or not
4608 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4610 @item -fmerge-constants
4611 Attempt to merge identical constants (string constants and floating point
4612 constants) across compilation units.
4614 This option is the default for optimized compilation if the assembler and
4615 linker support it. Use @option{-fno-merge-constants} to inhibit this
4618 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4620 @item -fmerge-all-constants
4621 Attempt to merge identical constants and identical variables.
4623 This option implies @option{-fmerge-constants}. In addition to
4624 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4625 arrays or initialized constant variables with integral or floating point
4626 types. Languages like C or C++ require each non-automatic variable to
4627 have distinct location, so using this option will result in non-conforming
4630 @item -fmodulo-sched
4631 @opindex fmodulo-sched
4632 Perform swing modulo scheduling immediately before the first scheduling
4633 pass. This pass looks at innermost loops and reorders their
4634 instructions by overlapping different iterations.
4636 @item -fno-branch-count-reg
4637 @opindex fno-branch-count-reg
4638 Do not use ``decrement and branch'' instructions on a count register,
4639 but instead generate a sequence of instructions that decrement a
4640 register, compare it against zero, then branch based upon the result.
4641 This option is only meaningful on architectures that support such
4642 instructions, which include x86, PowerPC, IA-64 and S/390.
4644 The default is @option{-fbranch-count-reg}, enabled when
4645 @option{-fstrength-reduce} is enabled.
4647 @item -fno-function-cse
4648 @opindex fno-function-cse
4649 Do not put function addresses in registers; make each instruction that
4650 calls a constant function contain the function's address explicitly.
4652 This option results in less efficient code, but some strange hacks
4653 that alter the assembler output may be confused by the optimizations
4654 performed when this option is not used.
4656 The default is @option{-ffunction-cse}
4658 @item -fno-zero-initialized-in-bss
4659 @opindex fno-zero-initialized-in-bss
4660 If the target supports a BSS section, GCC by default puts variables that
4661 are initialized to zero into BSS@. This can save space in the resulting
4664 This option turns off this behavior because some programs explicitly
4665 rely on variables going to the data section. E.g., so that the
4666 resulting executable can find the beginning of that section and/or make
4667 assumptions based on that.
4669 The default is @option{-fzero-initialized-in-bss}.
4671 @item -fbounds-check
4672 @opindex fbounds-check
4673 For front-ends that support it, generate additional code to check that
4674 indices used to access arrays are within the declared range. This is
4675 currently only supported by the Java and Fortran front-ends, where
4676 this option defaults to true and false respectively.
4678 @item -fmudflap -fmudflapth -fmudflapir
4682 @cindex bounds checking
4684 For front-ends that support it (C and C++), instrument all risky
4685 pointer/array dereferencing operations, some standard library
4686 string/heap functions, and some other associated constructs with
4687 range/validity tests. Modules so instrumented should be immune to
4688 buffer overflows, invalid heap use, and some other classes of C/C++
4689 programming errors. The instrumentation relies on a separate runtime
4690 library (@file{libmudflap}), which will be linked into a program if
4691 @option{-fmudflap} is given at link time. Run-time behavior of the
4692 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4693 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4696 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4697 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4698 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4699 instrumentation should ignore pointer reads. This produces less
4700 instrumentation (and therefore faster execution) and still provides
4701 some protection against outright memory corrupting writes, but allows
4702 erroneously read data to propagate within a program.
4706 @cindex openmp parallel
4707 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
4708 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
4709 compiler generates parallel code according to the OpenMP Application
4710 Program Interface v2.5. To generate the final exectuable, the runtime
4711 library @code{libgomp} must be linked in using @option{-lgomp}.
4713 @item -fstrength-reduce
4714 @opindex fstrength-reduce
4715 Perform the optimizations of loop strength reduction and
4716 elimination of iteration variables.
4718 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4720 @item -fthread-jumps
4721 @opindex fthread-jumps
4722 Perform optimizations where we check to see if a jump branches to a
4723 location where another comparison subsumed by the first is found. If
4724 so, the first branch is redirected to either the destination of the
4725 second branch or a point immediately following it, depending on whether
4726 the condition is known to be true or false.
4728 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4730 @item -fcse-follow-jumps
4731 @opindex fcse-follow-jumps
4732 In common subexpression elimination, scan through jump instructions
4733 when the target of the jump is not reached by any other path. For
4734 example, when CSE encounters an @code{if} statement with an
4735 @code{else} clause, CSE will follow the jump when the condition
4738 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4740 @item -fcse-skip-blocks
4741 @opindex fcse-skip-blocks
4742 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4743 follow jumps which conditionally skip over blocks. When CSE
4744 encounters a simple @code{if} statement with no else clause,
4745 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4746 body of the @code{if}.
4748 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4750 @item -frerun-cse-after-loop
4751 @opindex frerun-cse-after-loop
4752 Re-run common subexpression elimination after loop optimizations has been
4755 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4757 @item -frerun-loop-opt
4758 @opindex frerun-loop-opt
4759 Run the loop optimizer twice.
4761 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4765 Perform a global common subexpression elimination pass.
4766 This pass also performs global constant and copy propagation.
4768 @emph{Note:} When compiling a program using computed gotos, a GCC
4769 extension, you may get better runtime performance if you disable
4770 the global common subexpression elimination pass by adding
4771 @option{-fno-gcse} to the command line.
4773 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4777 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4778 attempt to move loads which are only killed by stores into themselves. This
4779 allows a loop containing a load/store sequence to be changed to a load outside
4780 the loop, and a copy/store within the loop.
4782 Enabled by default when gcse is enabled.
4786 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4787 global common subexpression elimination. This pass will attempt to move
4788 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4789 loops containing a load/store sequence can be changed to a load before
4790 the loop and a store after the loop.
4792 Not enabled at any optimization level.
4796 When @option{-fgcse-las} is enabled, the global common subexpression
4797 elimination pass eliminates redundant loads that come after stores to the
4798 same memory location (both partial and full redundancies).
4800 Not enabled at any optimization level.
4802 @item -fgcse-after-reload
4803 @opindex fgcse-after-reload
4804 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4805 pass is performed after reload. The purpose of this pass is to cleanup
4808 @item -floop-optimize
4809 @opindex floop-optimize
4810 Perform loop optimizations: move constant expressions out of loops, simplify
4811 exit test conditions and optionally do strength-reduction as well.
4813 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4815 @item -floop-optimize2
4816 @opindex floop-optimize2
4817 Perform loop optimizations using the new loop optimizer. The optimizations
4818 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4821 @item -funsafe-loop-optimizations
4822 @opindex funsafe-loop-optimizations
4823 If given, the loop optimizer will assume that loop indices do not
4824 overflow, and that the loops with nontrivial exit condition are not
4825 infinite. This enables a wider range of loop optimizations even if
4826 the loop optimizer itself cannot prove that these assumptions are valid.
4827 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4828 if it finds this kind of loop.
4830 @item -fcrossjumping
4831 @opindex crossjumping
4832 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4833 resulting code may or may not perform better than without cross-jumping.
4835 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4837 @item -fif-conversion
4838 @opindex if-conversion
4839 Attempt to transform conditional jumps into branch-less equivalents. This
4840 include use of conditional moves, min, max, set flags and abs instructions, and
4841 some tricks doable by standard arithmetics. The use of conditional execution
4842 on chips where it is available is controlled by @code{if-conversion2}.
4844 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4846 @item -fif-conversion2
4847 @opindex if-conversion2
4848 Use conditional execution (where available) to transform conditional jumps into
4849 branch-less equivalents.
4851 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4853 @item -fdelete-null-pointer-checks
4854 @opindex fdelete-null-pointer-checks
4855 Use global dataflow analysis to identify and eliminate useless checks
4856 for null pointers. The compiler assumes that dereferencing a null
4857 pointer would have halted the program. If a pointer is checked after
4858 it has already been dereferenced, it cannot be null.
4860 In some environments, this assumption is not true, and programs can
4861 safely dereference null pointers. Use
4862 @option{-fno-delete-null-pointer-checks} to disable this optimization
4863 for programs which depend on that behavior.
4865 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4867 @item -fexpensive-optimizations
4868 @opindex fexpensive-optimizations
4869 Perform a number of minor optimizations that are relatively expensive.
4871 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4873 @item -foptimize-register-move
4875 @opindex foptimize-register-move
4877 Attempt to reassign register numbers in move instructions and as
4878 operands of other simple instructions in order to maximize the amount of
4879 register tying. This is especially helpful on machines with two-operand
4882 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4885 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4887 @item -fdelayed-branch
4888 @opindex fdelayed-branch
4889 If supported for the target machine, attempt to reorder instructions
4890 to exploit instruction slots available after delayed branch
4893 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4895 @item -fschedule-insns
4896 @opindex fschedule-insns
4897 If supported for the target machine, attempt to reorder instructions to
4898 eliminate execution stalls due to required data being unavailable. This
4899 helps machines that have slow floating point or memory load instructions
4900 by allowing other instructions to be issued until the result of the load
4901 or floating point instruction is required.
4903 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4905 @item -fschedule-insns2
4906 @opindex fschedule-insns2
4907 Similar to @option{-fschedule-insns}, but requests an additional pass of
4908 instruction scheduling after register allocation has been done. This is
4909 especially useful on machines with a relatively small number of
4910 registers and where memory load instructions take more than one cycle.
4912 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4914 @item -fno-sched-interblock
4915 @opindex fno-sched-interblock
4916 Don't schedule instructions across basic blocks. This is normally
4917 enabled by default when scheduling before register allocation, i.e.@:
4918 with @option{-fschedule-insns} or at @option{-O2} or higher.
4920 @item -fno-sched-spec
4921 @opindex fno-sched-spec
4922 Don't allow speculative motion of non-load instructions. This is normally
4923 enabled by default when scheduling before register allocation, i.e.@:
4924 with @option{-fschedule-insns} or at @option{-O2} or higher.
4926 @item -fsched-spec-load
4927 @opindex fsched-spec-load
4928 Allow speculative motion of some load instructions. This only makes
4929 sense when scheduling before register allocation, i.e.@: with
4930 @option{-fschedule-insns} or at @option{-O2} or higher.
4932 @item -fsched-spec-load-dangerous
4933 @opindex fsched-spec-load-dangerous
4934 Allow speculative motion of more load instructions. This only makes
4935 sense when scheduling before register allocation, i.e.@: with
4936 @option{-fschedule-insns} or at @option{-O2} or higher.
4938 @item -fsched-stalled-insns=@var{n}
4939 @opindex fsched-stalled-insns
4940 Define how many insns (if any) can be moved prematurely from the queue
4941 of stalled insns into the ready list, during the second scheduling pass.
4943 @item -fsched-stalled-insns-dep=@var{n}
4944 @opindex fsched-stalled-insns-dep
4945 Define how many insn groups (cycles) will be examined for a dependency
4946 on a stalled insn that is candidate for premature removal from the queue
4947 of stalled insns. Has an effect only during the second scheduling pass,
4948 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4950 @item -fsched2-use-superblocks
4951 @opindex fsched2-use-superblocks
4952 When scheduling after register allocation, do use superblock scheduling
4953 algorithm. Superblock scheduling allows motion across basic block boundaries
4954 resulting on faster schedules. This option is experimental, as not all machine
4955 descriptions used by GCC model the CPU closely enough to avoid unreliable
4956 results from the algorithm.
4958 This only makes sense when scheduling after register allocation, i.e.@: with
4959 @option{-fschedule-insns2} or at @option{-O2} or higher.
4961 @item -fsched2-use-traces
4962 @opindex fsched2-use-traces
4963 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4964 allocation and additionally perform code duplication in order to increase the
4965 size of superblocks using tracer pass. See @option{-ftracer} for details on
4968 This mode should produce faster but significantly longer programs. Also
4969 without @option{-fbranch-probabilities} the traces constructed may not
4970 match the reality and hurt the performance. This only makes
4971 sense when scheduling after register allocation, i.e.@: with
4972 @option{-fschedule-insns2} or at @option{-O2} or higher.
4974 @item -freschedule-modulo-scheduled-loops
4975 @opindex fscheduling-in-modulo-scheduled-loops
4976 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4977 we may want to prevent the later scheduling passes from changing its schedule, we use this
4978 option to control that.
4980 @item -fcaller-saves
4981 @opindex fcaller-saves
4982 Enable values to be allocated in registers that will be clobbered by
4983 function calls, by emitting extra instructions to save and restore the
4984 registers around such calls. Such allocation is done only when it
4985 seems to result in better code than would otherwise be produced.
4987 This option is always enabled by default on certain machines, usually
4988 those which have no call-preserved registers to use instead.
4990 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4993 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4994 enabled by default at @option{-O2} and @option{-O3}.
4997 Perform Full Redundancy Elimination (FRE) on trees. The difference
4998 between FRE and PRE is that FRE only considers expressions
4999 that are computed on all paths leading to the redundant computation.
5000 This analysis faster than PRE, though it exposes fewer redundancies.
5001 This flag is enabled by default at @option{-O} and higher.
5003 @item -ftree-copy-prop
5004 Perform copy propagation on trees. This pass eliminates unnecessary
5005 copy operations. This flag is enabled by default at @option{-O} and
5008 @item -ftree-store-copy-prop
5009 Perform copy propagation of memory loads and stores. This pass
5010 eliminates unnecessary copy operations in memory references
5011 (structures, global variables, arrays, etc). This flag is enabled by
5012 default at @option{-O2} and higher.
5015 Perform structural alias analysis on trees. This flag
5016 is enabled by default at @option{-O} and higher.
5019 Perform forward store motion on trees. This flag is
5020 enabled by default at @option{-O} and higher.
5023 Perform sparse conditional constant propagation (CCP) on trees. This
5024 pass only operates on local scalar variables and is enabled by default
5025 at @option{-O} and higher.
5027 @item -ftree-store-ccp
5028 Perform sparse conditional constant propagation (CCP) on trees. This
5029 pass operates on both local scalar variables and memory stores and
5030 loads (global variables, structures, arrays, etc). This flag is
5031 enabled by default at @option{-O2} and higher.
5034 Perform dead code elimination (DCE) on trees. This flag is enabled by
5035 default at @option{-O} and higher.
5037 @item -ftree-dominator-opts
5038 Perform a variety of simple scalar cleanups (constant/copy
5039 propagation, redundancy elimination, range propagation and expression
5040 simplification) based on a dominator tree traversal. This also
5041 performs jump threading (to reduce jumps to jumps). This flag is
5042 enabled by default at @option{-O} and higher.
5045 Perform loop header copying on trees. This is beneficial since it increases
5046 effectiveness of code motion optimizations. It also saves one jump. This flag
5047 is enabled by default at @option{-O} and higher. It is not enabled
5048 for @option{-Os}, since it usually increases code size.
5050 @item -ftree-loop-optimize
5051 Perform loop optimizations on trees. This flag is enabled by default
5052 at @option{-O} and higher.
5054 @item -ftree-loop-linear
5055 Perform linear loop transformations on tree. This flag can improve cache
5056 performance and allow further loop optimizations to take place.
5058 @item -ftree-loop-im
5059 Perform loop invariant motion on trees. This pass moves only invariants that
5060 would be hard to handle at RTL level (function calls, operations that expand to
5061 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5062 operands of conditions that are invariant out of the loop, so that we can use
5063 just trivial invariantness analysis in loop unswitching. The pass also includes
5066 @item -ftree-loop-ivcanon
5067 Create a canonical counter for number of iterations in the loop for that
5068 determining number of iterations requires complicated analysis. Later
5069 optimizations then may determine the number easily. Useful especially
5070 in connection with unrolling.
5073 Perform induction variable optimizations (strength reduction, induction
5074 variable merging and induction variable elimination) on trees.
5077 Perform scalar replacement of aggregates. This pass replaces structure
5078 references with scalars to prevent committing structures to memory too
5079 early. This flag is enabled by default at @option{-O} and higher.
5081 @item -ftree-copyrename
5082 Perform copy renaming on trees. This pass attempts to rename compiler
5083 temporaries to other variables at copy locations, usually resulting in
5084 variable names which more closely resemble the original variables. This flag
5085 is enabled by default at @option{-O} and higher.
5088 Perform temporary expression replacement during the SSA->normal phase. Single
5089 use/single def temporaries are replaced at their use location with their
5090 defining expression. This results in non-GIMPLE code, but gives the expanders
5091 much more complex trees to work on resulting in better RTL generation. This is
5092 enabled by default at @option{-O} and higher.
5095 Perform live range splitting during the SSA->normal phase. Distinct live
5096 ranges of a variable are split into unique variables, allowing for better
5097 optimization later. This is enabled by default at @option{-O} and higher.
5099 @item -ftree-vectorize
5100 Perform loop vectorization on trees.
5102 @item -ftree-vect-loop-version
5103 @opindex ftree-vect-loop-version
5104 Perform loop versioning when doing loop vectorization on trees. When a loop
5105 appears to be vectorizable except that data alignment or data dependence cannot
5106 be determined at compile time then vectorized and non-vectorized versions of
5107 the loop are generated along with runtime checks for alignment or dependence
5108 to control which version is executed. This option is enabled by default
5109 except at level @option{-Os} where it is disabled.
5112 Perform Value Range Propagation on trees. This is similar to the
5113 constant propagation pass, but instead of values, ranges of values are
5114 propagated. This allows the optimizers to remove unnecessary range
5115 checks like array bound checks and null pointer checks. This is
5116 enabled by default at @option{-O2} and higher. Null pointer check
5117 elimination is only done if @option{-fdelete-null-pointer-checks} is
5122 Perform tail duplication to enlarge superblock size. This transformation
5123 simplifies the control flow of the function allowing other optimizations to do
5126 @item -funroll-loops
5127 @opindex funroll-loops
5128 Unroll loops whose number of iterations can be determined at compile
5129 time or upon entry to the loop. @option{-funroll-loops} implies both
5130 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
5131 option makes code larger, and may or may not make it run faster.
5133 @item -funroll-all-loops
5134 @opindex funroll-all-loops
5135 Unroll all loops, even if their number of iterations is uncertain when
5136 the loop is entered. This usually makes programs run more slowly.
5137 @option{-funroll-all-loops} implies the same options as
5138 @option{-funroll-loops},
5140 @item -fsplit-ivs-in-unroller
5141 @opindex -fsplit-ivs-in-unroller
5142 Enables expressing of values of induction variables in later iterations
5143 of the unrolled loop using the value in the first iteration. This breaks
5144 long dependency chains, thus improving efficiency of the scheduling passes.
5146 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5147 same effect. However in cases the loop body is more complicated than
5148 a single basic block, this is not reliable. It also does not work at all
5149 on some of the architectures due to restrictions in the CSE pass.
5151 This optimization is enabled by default.
5153 @item -fvariable-expansion-in-unroller
5154 @opindex -fvariable-expansion-in-unroller
5155 With this option, the compiler will create multiple copies of some
5156 local variables when unrolling a loop which can result in superior code.
5158 @item -fprefetch-loop-arrays
5159 @opindex fprefetch-loop-arrays
5160 If supported by the target machine, generate instructions to prefetch
5161 memory to improve the performance of loops that access large arrays.
5163 These options may generate better or worse code; results are highly
5164 dependent on the structure of loops within the source code.
5167 @itemx -fno-peephole2
5168 @opindex fno-peephole
5169 @opindex fno-peephole2
5170 Disable any machine-specific peephole optimizations. The difference
5171 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5172 are implemented in the compiler; some targets use one, some use the
5173 other, a few use both.
5175 @option{-fpeephole} is enabled by default.
5176 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5178 @item -fno-guess-branch-probability
5179 @opindex fno-guess-branch-probability
5180 Do not guess branch probabilities using heuristics.
5182 GCC will use heuristics to guess branch probabilities if they are
5183 not provided by profiling feedback (@option{-fprofile-arcs}). These
5184 heuristics are based on the control flow graph. If some branch probabilities
5185 are specified by @samp{__builtin_expect}, then the heuristics will be
5186 used to guess branch probabilities for the rest of the control flow graph,
5187 taking the @samp{__builtin_expect} info into account. The interactions
5188 between the heuristics and @samp{__builtin_expect} can be complex, and in
5189 some cases, it may be useful to disable the heuristics so that the effects
5190 of @samp{__builtin_expect} are easier to understand.
5192 The default is @option{-fguess-branch-probability} at levels
5193 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5195 @item -freorder-blocks
5196 @opindex freorder-blocks
5197 Reorder basic blocks in the compiled function in order to reduce number of
5198 taken branches and improve code locality.
5200 Enabled at levels @option{-O2}, @option{-O3}.
5202 @item -freorder-blocks-and-partition
5203 @opindex freorder-blocks-and-partition
5204 In addition to reordering basic blocks in the compiled function, in order
5205 to reduce number of taken branches, partitions hot and cold basic blocks
5206 into separate sections of the assembly and .o files, to improve
5207 paging and cache locality performance.
5209 This optimization is automatically turned off in the presence of
5210 exception handling, for linkonce sections, for functions with a user-defined
5211 section attribute and on any architecture that does not support named
5214 @item -freorder-functions
5215 @opindex freorder-functions
5216 Reorder functions in the object file in order to
5217 improve code locality. This is implemented by using special
5218 subsections @code{.text.hot} for most frequently executed functions and
5219 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5220 the linker so object file format must support named sections and linker must
5221 place them in a reasonable way.
5223 Also profile feedback must be available in to make this option effective. See
5224 @option{-fprofile-arcs} for details.
5226 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5228 @item -fstrict-aliasing
5229 @opindex fstrict-aliasing
5230 Allows the compiler to assume the strictest aliasing rules applicable to
5231 the language being compiled. For C (and C++), this activates
5232 optimizations based on the type of expressions. In particular, an
5233 object of one type is assumed never to reside at the same address as an
5234 object of a different type, unless the types are almost the same. For
5235 example, an @code{unsigned int} can alias an @code{int}, but not a
5236 @code{void*} or a @code{double}. A character type may alias any other
5239 Pay special attention to code like this:
5252 The practice of reading from a different union member than the one most
5253 recently written to (called ``type-punning'') is common. Even with
5254 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5255 is accessed through the union type. So, the code above will work as
5256 expected. However, this code might not:
5267 Every language that wishes to perform language-specific alias analysis
5268 should define a function that computes, given an @code{tree}
5269 node, an alias set for the node. Nodes in different alias sets are not
5270 allowed to alias. For an example, see the C front-end function
5271 @code{c_get_alias_set}.
5273 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5275 @item -falign-functions
5276 @itemx -falign-functions=@var{n}
5277 @opindex falign-functions
5278 Align the start of functions to the next power-of-two greater than
5279 @var{n}, skipping up to @var{n} bytes. For instance,
5280 @option{-falign-functions=32} aligns functions to the next 32-byte
5281 boundary, but @option{-falign-functions=24} would align to the next
5282 32-byte boundary only if this can be done by skipping 23 bytes or less.
5284 @option{-fno-align-functions} and @option{-falign-functions=1} are
5285 equivalent and mean that functions will not be aligned.
5287 Some assemblers only support this flag when @var{n} is a power of two;
5288 in that case, it is rounded up.
5290 If @var{n} is not specified or is zero, use a machine-dependent default.
5292 Enabled at levels @option{-O2}, @option{-O3}.
5294 @item -falign-labels
5295 @itemx -falign-labels=@var{n}
5296 @opindex falign-labels
5297 Align all branch targets to a power-of-two boundary, skipping up to
5298 @var{n} bytes like @option{-falign-functions}. This option can easily
5299 make code slower, because it must insert dummy operations for when the
5300 branch target is reached in the usual flow of the code.
5302 @option{-fno-align-labels} and @option{-falign-labels=1} are
5303 equivalent and mean that labels will not be aligned.
5305 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5306 are greater than this value, then their values are used instead.
5308 If @var{n} is not specified or is zero, use a machine-dependent default
5309 which is very likely to be @samp{1}, meaning no alignment.
5311 Enabled at levels @option{-O2}, @option{-O3}.
5314 @itemx -falign-loops=@var{n}
5315 @opindex falign-loops
5316 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5317 like @option{-falign-functions}. The hope is that the loop will be
5318 executed many times, which will make up for any execution of the dummy
5321 @option{-fno-align-loops} and @option{-falign-loops=1} are
5322 equivalent and mean that loops will not be aligned.
5324 If @var{n} is not specified or is zero, use a machine-dependent default.
5326 Enabled at levels @option{-O2}, @option{-O3}.
5329 @itemx -falign-jumps=@var{n}
5330 @opindex falign-jumps
5331 Align branch targets to a power-of-two boundary, for branch targets
5332 where the targets can only be reached by jumping, skipping up to @var{n}
5333 bytes like @option{-falign-functions}. In this case, no dummy operations
5336 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5337 equivalent and mean that loops will not be aligned.
5339 If @var{n} is not specified or is zero, use a machine-dependent default.
5341 Enabled at levels @option{-O2}, @option{-O3}.
5343 @item -funit-at-a-time
5344 @opindex funit-at-a-time
5345 Parse the whole compilation unit before starting to produce code.
5346 This allows some extra optimizations to take place but consumes
5347 more memory (in general). There are some compatibility issues
5348 with @emph{unit-at-a-time} mode:
5351 enabling @emph{unit-at-a-time} mode may change the order
5352 in which functions, variables, and top-level @code{asm} statements
5353 are emitted, and will likely break code relying on some particular
5354 ordering. The majority of such top-level @code{asm} statements,
5355 though, can be replaced by @code{section} attributes. The
5356 @option{fno-toplevel-reorder} option may be used to keep the ordering
5357 used in the input file, at the cost of some optimizations.
5360 @emph{unit-at-a-time} mode removes unreferenced static variables
5361 and functions. This may result in undefined references
5362 when an @code{asm} statement refers directly to variables or functions
5363 that are otherwise unused. In that case either the variable/function
5364 shall be listed as an operand of the @code{asm} statement operand or,
5365 in the case of top-level @code{asm} statements the attribute @code{used}
5366 shall be used on the declaration.
5369 Static functions now can use non-standard passing conventions that
5370 may break @code{asm} statements calling functions directly. Again,
5371 attribute @code{used} will prevent this behavior.
5374 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5375 but this scheme may not be supported by future releases of GCC@.
5377 Enabled at levels @option{-O2}, @option{-O3}.
5379 @item -fno-toplevel-reorder
5380 Do not reorder top-level functions, variables, and @code{asm}
5381 statements. Output them in the same order that they appear in the
5382 input file. When this option is used, unreferenced static variables
5383 will not be removed. This option is intended to support existing code
5384 which relies on a particular ordering. For new code, it is better to
5389 Constructs webs as commonly used for register allocation purposes and assign
5390 each web individual pseudo register. This allows the register allocation pass
5391 to operate on pseudos directly, but also strengthens several other optimization
5392 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5393 however, make debugging impossible, since variables will no longer stay in a
5396 Enabled by default with @option{-funroll-loops}.
5398 @item -fwhole-program
5399 @opindex fwhole-program
5400 Assume that the current compilation unit represents whole program being
5401 compiled. All public functions and variables with the exception of @code{main}
5402 and those merged by attribute @code{externally_visible} become static functions
5403 and in a affect gets more aggressively optimized by interprocedural optimizers.
5404 While this option is equivalent to proper use of @code{static} keyword for
5405 programs consisting of single file, in combination with option
5406 @option{--combine} this flag can be used to compile most of smaller scale C
5407 programs since the functions and variables become local for the whole combined
5408 compilation unit, not for the single source file itself.
5411 @item -fno-cprop-registers
5412 @opindex fno-cprop-registers
5413 After register allocation and post-register allocation instruction splitting,
5414 we perform a copy-propagation pass to try to reduce scheduling dependencies
5415 and occasionally eliminate the copy.
5417 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5419 @item -fprofile-generate
5420 @opindex fprofile-generate
5422 Enable options usually used for instrumenting application to produce
5423 profile useful for later recompilation with profile feedback based
5424 optimization. You must use @option{-fprofile-generate} both when
5425 compiling and when linking your program.
5427 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5430 @opindex fprofile-use
5431 Enable profile feedback directed optimizations, and optimizations
5432 generally profitable only with profile feedback available.
5434 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5435 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer},
5436 @code{-fno-loop-optimize}.
5440 The following options control compiler behavior regarding floating
5441 point arithmetic. These options trade off between speed and
5442 correctness. All must be specifically enabled.
5446 @opindex ffloat-store
5447 Do not store floating point variables in registers, and inhibit other
5448 options that might change whether a floating point value is taken from a
5451 @cindex floating point precision
5452 This option prevents undesirable excess precision on machines such as
5453 the 68000 where the floating registers (of the 68881) keep more
5454 precision than a @code{double} is supposed to have. Similarly for the
5455 x86 architecture. For most programs, the excess precision does only
5456 good, but a few programs rely on the precise definition of IEEE floating
5457 point. Use @option{-ffloat-store} for such programs, after modifying
5458 them to store all pertinent intermediate computations into variables.
5462 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5463 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5464 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5465 and @option{fcx-limited-range}.
5467 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5469 This option should never be turned on by any @option{-O} option since
5470 it can result in incorrect output for programs which depend on
5471 an exact implementation of IEEE or ISO rules/specifications for
5474 @item -fno-math-errno
5475 @opindex fno-math-errno
5476 Do not set ERRNO after calling math functions that are executed
5477 with a single instruction, e.g., sqrt. A program that relies on
5478 IEEE exceptions for math error handling may want to use this flag
5479 for speed while maintaining IEEE arithmetic compatibility.
5481 This option should never be turned on by any @option{-O} option since
5482 it can result in incorrect output for programs which depend on
5483 an exact implementation of IEEE or ISO rules/specifications for
5486 The default is @option{-fmath-errno}.
5488 On Darwin systems, the math library never sets @code{errno}. There is therefore
5489 no reason for the compiler to consider the possibility that it might,
5490 and @option{-fno-math-errno} is the default.
5492 @item -funsafe-math-optimizations
5493 @opindex funsafe-math-optimizations
5494 Allow optimizations for floating-point arithmetic that (a) assume
5495 that arguments and results are valid and (b) may violate IEEE or
5496 ANSI standards. When used at link-time, it may include libraries
5497 or startup files that change the default FPU control word or other
5498 similar optimizations.
5500 This option should never be turned on by any @option{-O} option since
5501 it can result in incorrect output for programs which depend on
5502 an exact implementation of IEEE or ISO rules/specifications for
5505 The default is @option{-fno-unsafe-math-optimizations}.
5507 @item -ffinite-math-only
5508 @opindex ffinite-math-only
5509 Allow optimizations for floating-point arithmetic that assume
5510 that arguments and results are not NaNs or +-Infs.
5512 This option should never be turned on by any @option{-O} option since
5513 it can result in incorrect output for programs which depend on
5514 an exact implementation of IEEE or ISO rules/specifications.
5516 The default is @option{-fno-finite-math-only}.
5518 @item -fno-trapping-math
5519 @opindex fno-trapping-math
5520 Compile code assuming that floating-point operations cannot generate
5521 user-visible traps. These traps include division by zero, overflow,
5522 underflow, inexact result and invalid operation. This option implies
5523 @option{-fno-signaling-nans}. Setting this option may allow faster
5524 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5526 This option should never be turned on by any @option{-O} option since
5527 it can result in incorrect output for programs which depend on
5528 an exact implementation of IEEE or ISO rules/specifications for
5531 The default is @option{-ftrapping-math}.
5533 @item -frounding-math
5534 @opindex frounding-math
5535 Disable transformations and optimizations that assume default floating
5536 point rounding behavior. This is round-to-zero for all floating point
5537 to integer conversions, and round-to-nearest for all other arithmetic
5538 truncations. This option should be specified for programs that change
5539 the FP rounding mode dynamically, or that may be executed with a
5540 non-default rounding mode. This option disables constant folding of
5541 floating point expressions at compile-time (which may be affected by
5542 rounding mode) and arithmetic transformations that are unsafe in the
5543 presence of sign-dependent rounding modes.
5545 The default is @option{-fno-rounding-math}.
5547 This option is experimental and does not currently guarantee to
5548 disable all GCC optimizations that are affected by rounding mode.
5549 Future versions of GCC may provide finer control of this setting
5550 using C99's @code{FENV_ACCESS} pragma. This command line option
5551 will be used to specify the default state for @code{FENV_ACCESS}.
5553 @item -frtl-abstract-sequences
5554 @opindex frtl-abstract-sequences
5555 It is a size optimization method. This option is to find identical
5556 sequences of code, which can be turned into pseudo-procedures and
5557 then replace all occurrences with calls to the newly created
5558 subroutine. It is kind of an opposite of @option{-finline-functions}.
5559 This optimization runs at RTL level.
5561 @item -fsignaling-nans
5562 @opindex fsignaling-nans
5563 Compile code assuming that IEEE signaling NaNs may generate user-visible
5564 traps during floating-point operations. Setting this option disables
5565 optimizations that may change the number of exceptions visible with
5566 signaling NaNs. This option implies @option{-ftrapping-math}.
5568 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5571 The default is @option{-fno-signaling-nans}.
5573 This option is experimental and does not currently guarantee to
5574 disable all GCC optimizations that affect signaling NaN behavior.
5576 @item -fsingle-precision-constant
5577 @opindex fsingle-precision-constant
5578 Treat floating point constant as single precision constant instead of
5579 implicitly converting it to double precision constant.
5581 @item -fcx-limited-range
5582 @itemx -fno-cx-limited-range
5583 @opindex fcx-limited-range
5584 @opindex fno-cx-limited-range
5585 When enabled, this option states that a range reduction step is not
5586 needed when performing complex division. The default is
5587 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5589 This option controls the default setting of the ISO C99
5590 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5595 The following options control optimizations that may improve
5596 performance, but are not enabled by any @option{-O} options. This
5597 section includes experimental options that may produce broken code.
5600 @item -fbranch-probabilities
5601 @opindex fbranch-probabilities
5602 After running a program compiled with @option{-fprofile-arcs}
5603 (@pxref{Debugging Options,, Options for Debugging Your Program or
5604 @command{gcc}}), you can compile it a second time using
5605 @option{-fbranch-probabilities}, to improve optimizations based on
5606 the number of times each branch was taken. When the program
5607 compiled with @option{-fprofile-arcs} exits it saves arc execution
5608 counts to a file called @file{@var{sourcename}.gcda} for each source
5609 file The information in this data file is very dependent on the
5610 structure of the generated code, so you must use the same source code
5611 and the same optimization options for both compilations.
5613 With @option{-fbranch-probabilities}, GCC puts a
5614 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5615 These can be used to improve optimization. Currently, they are only
5616 used in one place: in @file{reorg.c}, instead of guessing which path a
5617 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5618 exactly determine which path is taken more often.
5620 @item -fprofile-values
5621 @opindex fprofile-values
5622 If combined with @option{-fprofile-arcs}, it adds code so that some
5623 data about values of expressions in the program is gathered.
5625 With @option{-fbranch-probabilities}, it reads back the data gathered
5626 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5627 notes to instructions for their later usage in optimizations.
5629 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5633 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5634 a code to gather information about values of expressions.
5636 With @option{-fbranch-probabilities}, it reads back the data gathered
5637 and actually performs the optimizations based on them.
5638 Currently the optimizations include specialization of division operation
5639 using the knowledge about the value of the denominator.
5641 @item -frename-registers
5642 @opindex frename-registers
5643 Attempt to avoid false dependencies in scheduled code by making use
5644 of registers left over after register allocation. This optimization
5645 will most benefit processors with lots of registers. Depending on the
5646 debug information format adopted by the target, however, it can
5647 make debugging impossible, since variables will no longer stay in
5648 a ``home register''.
5650 Enabled by default with @option{-funroll-loops}.
5654 Perform tail duplication to enlarge superblock size. This transformation
5655 simplifies the control flow of the function allowing other optimizations to do
5658 Enabled with @option{-fprofile-use}.
5660 @item -funroll-loops
5661 @opindex funroll-loops
5662 Unroll loops whose number of iterations can be determined at compile time or
5663 upon entry to the loop. @option{-funroll-loops} implies
5664 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5665 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5666 small constant number of iterations). This option makes code larger, and may
5667 or may not make it run faster.
5669 Enabled with @option{-fprofile-use}.
5671 @item -funroll-all-loops
5672 @opindex funroll-all-loops
5673 Unroll all loops, even if their number of iterations is uncertain when
5674 the loop is entered. This usually makes programs run more slowly.
5675 @option{-funroll-all-loops} implies the same options as
5676 @option{-funroll-loops}.
5679 @opindex fpeel-loops
5680 Peels the loops for that there is enough information that they do not
5681 roll much (from profile feedback). It also turns on complete loop peeling
5682 (i.e.@: complete removal of loops with small constant number of iterations).
5684 Enabled with @option{-fprofile-use}.
5686 @item -fmove-loop-invariants
5687 @opindex fmove-loop-invariants
5688 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5689 at level @option{-O1}
5691 @item -funswitch-loops
5692 @opindex funswitch-loops
5693 Move branches with loop invariant conditions out of the loop, with duplicates
5694 of the loop on both branches (modified according to result of the condition).
5696 @item -fprefetch-loop-arrays
5697 @opindex fprefetch-loop-arrays
5698 If supported by the target machine, generate instructions to prefetch
5699 memory to improve the performance of loops that access large arrays.
5701 Disabled at level @option{-Os}.
5703 @item -ffunction-sections
5704 @itemx -fdata-sections
5705 @opindex ffunction-sections
5706 @opindex fdata-sections
5707 Place each function or data item into its own section in the output
5708 file if the target supports arbitrary sections. The name of the
5709 function or the name of the data item determines the section's name
5712 Use these options on systems where the linker can perform optimizations
5713 to improve locality of reference in the instruction space. Most systems
5714 using the ELF object format and SPARC processors running Solaris 2 have
5715 linkers with such optimizations. AIX may have these optimizations in
5718 Only use these options when there are significant benefits from doing
5719 so. When you specify these options, the assembler and linker will
5720 create larger object and executable files and will also be slower.
5721 You will not be able to use @code{gprof} on all systems if you
5722 specify this option and you may have problems with debugging if
5723 you specify both this option and @option{-g}.
5725 @item -fbranch-target-load-optimize
5726 @opindex fbranch-target-load-optimize
5727 Perform branch target register load optimization before prologue / epilogue
5729 The use of target registers can typically be exposed only during reload,
5730 thus hoisting loads out of loops and doing inter-block scheduling needs
5731 a separate optimization pass.
5733 @item -fbranch-target-load-optimize2
5734 @opindex fbranch-target-load-optimize2
5735 Perform branch target register load optimization after prologue / epilogue
5738 @item -fbtr-bb-exclusive
5739 @opindex fbtr-bb-exclusive
5740 When performing branch target register load optimization, don't reuse
5741 branch target registers in within any basic block.
5743 @item -fstack-protector
5744 Emit extra code to check for buffer overflows, such as stack smashing
5745 attacks. This is done by adding a guard variable to functions with
5746 vulnerable objects. This includes functions that call alloca, and
5747 functions with buffers larger than 8 bytes. The guards are initialized
5748 when a function is entered and then checked when the function exits.
5749 If a guard check fails, an error message is printed and the program exits.
5751 @item -fstack-protector-all
5752 Like @option{-fstack-protector} except that all functions are protected.
5754 @item --param @var{name}=@var{value}
5756 In some places, GCC uses various constants to control the amount of
5757 optimization that is done. For example, GCC will not inline functions
5758 that contain more that a certain number of instructions. You can
5759 control some of these constants on the command-line using the
5760 @option{--param} option.
5762 The names of specific parameters, and the meaning of the values, are
5763 tied to the internals of the compiler, and are subject to change
5764 without notice in future releases.
5766 In each case, the @var{value} is an integer. The allowable choices for
5767 @var{name} are given in the following table:
5770 @item salias-max-implicit-fields
5771 The maximum number of fields in a variable without direct
5772 structure accesses for which structure aliasing will consider trying
5773 to track each field. The default is 5
5775 @item salias-max-array-elements
5776 The maximum number of elements an array can have and its elements
5777 still be tracked individually by structure aliasing. The default is 4
5779 @item sra-max-structure-size
5780 The maximum structure size, in bytes, at which the scalar replacement
5781 of aggregates (SRA) optimization will perform block copies. The
5782 default value, 0, implies that GCC will select the most appropriate
5785 @item sra-field-structure-ratio
5786 The threshold ratio (as a percentage) between instantiated fields and
5787 the complete structure size. We say that if the ratio of the number
5788 of bytes in instantiated fields to the number of bytes in the complete
5789 structure exceeds this parameter, then block copies are not used. The
5792 @item max-crossjump-edges
5793 The maximum number of incoming edges to consider for crossjumping.
5794 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5795 the number of edges incoming to each block. Increasing values mean
5796 more aggressive optimization, making the compile time increase with
5797 probably small improvement in executable size.
5799 @item min-crossjump-insns
5800 The minimum number of instructions which must be matched at the end
5801 of two blocks before crossjumping will be performed on them. This
5802 value is ignored in the case where all instructions in the block being
5803 crossjumped from are matched. The default value is 5.
5805 @item max-grow-copy-bb-insns
5806 The maximum code size expansion factor when copying basic blocks
5807 instead of jumping. The expansion is relative to a jump instruction.
5808 The default value is 8.
5810 @item max-goto-duplication-insns
5811 The maximum number of instructions to duplicate to a block that jumps
5812 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5813 passes, GCC factors computed gotos early in the compilation process,
5814 and unfactors them as late as possible. Only computed jumps at the
5815 end of a basic blocks with no more than max-goto-duplication-insns are
5816 unfactored. The default value is 8.
5818 @item max-delay-slot-insn-search
5819 The maximum number of instructions to consider when looking for an
5820 instruction to fill a delay slot. If more than this arbitrary number of
5821 instructions is searched, the time savings from filling the delay slot
5822 will be minimal so stop searching. Increasing values mean more
5823 aggressive optimization, making the compile time increase with probably
5824 small improvement in executable run time.
5826 @item max-delay-slot-live-search
5827 When trying to fill delay slots, the maximum number of instructions to
5828 consider when searching for a block with valid live register
5829 information. Increasing this arbitrarily chosen value means more
5830 aggressive optimization, increasing the compile time. This parameter
5831 should be removed when the delay slot code is rewritten to maintain the
5834 @item max-gcse-memory
5835 The approximate maximum amount of memory that will be allocated in
5836 order to perform the global common subexpression elimination
5837 optimization. If more memory than specified is required, the
5838 optimization will not be done.
5840 @item max-gcse-passes
5841 The maximum number of passes of GCSE to run. The default is 1.
5843 @item max-pending-list-length
5844 The maximum number of pending dependencies scheduling will allow
5845 before flushing the current state and starting over. Large functions
5846 with few branches or calls can create excessively large lists which
5847 needlessly consume memory and resources.
5849 @item max-inline-insns-single
5850 Several parameters control the tree inliner used in gcc.
5851 This number sets the maximum number of instructions (counted in GCC's
5852 internal representation) in a single function that the tree inliner
5853 will consider for inlining. This only affects functions declared
5854 inline and methods implemented in a class declaration (C++).
5855 The default value is 450.
5857 @item max-inline-insns-auto
5858 When you use @option{-finline-functions} (included in @option{-O3}),
5859 a lot of functions that would otherwise not be considered for inlining
5860 by the compiler will be investigated. To those functions, a different
5861 (more restrictive) limit compared to functions declared inline can
5863 The default value is 90.
5865 @item large-function-insns
5866 The limit specifying really large functions. For functions larger than this
5867 limit after inlining inlining is constrained by
5868 @option{--param large-function-growth}. This parameter is useful primarily
5869 to avoid extreme compilation time caused by non-linear algorithms used by the
5871 This parameter is ignored when @option{-funit-at-a-time} is not used.
5872 The default value is 2700.
5874 @item large-function-growth
5875 Specifies maximal growth of large function caused by inlining in percents.
5876 This parameter is ignored when @option{-funit-at-a-time} is not used.
5877 The default value is 100 which limits large function growth to 2.0 times
5880 @item large-unit-insns
5881 The limit specifying large translation unit. Growth caused by inlining of
5882 units larger than this limit is limited by @option{--param inline-unit-growth}.
5883 For small units this might be too tight (consider unit consisting of function A
5884 that is inline and B that just calls A three time. If B is small relative to
5885 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5886 large units consisting of small inlininable functions however the overall unit
5887 growth limit is needed to avoid exponential explosion of code size. Thus for
5888 smaller units, the size is increased to @option{--param large-unit-insns}
5889 before applying @option{--param inline-unit-growth}. The default is 10000
5891 @item inline-unit-growth
5892 Specifies maximal overall growth of the compilation unit caused by inlining.
5893 This parameter is ignored when @option{-funit-at-a-time} is not used.
5894 The default value is 50 which limits unit growth to 1.5 times the original
5897 @item max-inline-insns-recursive
5898 @itemx max-inline-insns-recursive-auto
5899 Specifies maximum number of instructions out-of-line copy of self recursive inline
5900 function can grow into by performing recursive inlining.
5902 For functions declared inline @option{--param max-inline-insns-recursive} is
5903 taken into acount. For function not declared inline, recursive inlining
5904 happens only when @option{-finline-functions} (included in @option{-O3}) is
5905 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5906 default value is 450.
5908 @item max-inline-recursive-depth
5909 @itemx max-inline-recursive-depth-auto
5910 Specifies maximum recursion depth used by the recursive inlining.
5912 For functions declared inline @option{--param max-inline-recursive-depth} is
5913 taken into acount. For function not declared inline, recursive inlining
5914 happens only when @option{-finline-functions} (included in @option{-O3}) is
5915 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5916 default value is 450.
5918 @item min-inline-recursive-probability
5919 Recursive inlining is profitable only for function having deep recursion
5920 in average and can hurt for function having little recursion depth by
5921 increasing the prologue size or complexity of function body to other
5924 When profile feedback is available (see @option{-fprofile-generate}) the actual
5925 recursion depth can be guessed from probability that function will recurse via
5926 given call expression. This parameter limits inlining only to call expression
5927 whose probability exceeds given threshold (in percents). The default value is
5930 @item inline-call-cost
5931 Specify cost of call instruction relative to simple arithmetics operations
5932 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5933 functions and at the same time increases size of leaf function that is believed to
5934 reduce function size by being inlined. In effect it increases amount of
5935 inlining for code having large abstraction penalty (many functions that just
5936 pass the arguments to other functions) and decrease inlining for code with low
5937 abstraction penalty. The default value is 16.
5939 @item max-unrolled-insns
5940 The maximum number of instructions that a loop should have if that loop
5941 is unrolled, and if the loop is unrolled, it determines how many times
5942 the loop code is unrolled.
5944 @item max-average-unrolled-insns
5945 The maximum number of instructions biased by probabilities of their execution
5946 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5947 it determines how many times the loop code is unrolled.
5949 @item max-unroll-times
5950 The maximum number of unrollings of a single loop.
5952 @item max-peeled-insns
5953 The maximum number of instructions that a loop should have if that loop
5954 is peeled, and if the loop is peeled, it determines how many times
5955 the loop code is peeled.
5957 @item max-peel-times
5958 The maximum number of peelings of a single loop.
5960 @item max-completely-peeled-insns
5961 The maximum number of insns of a completely peeled loop.
5963 @item max-completely-peel-times
5964 The maximum number of iterations of a loop to be suitable for complete peeling.
5966 @item max-unswitch-insns
5967 The maximum number of insns of an unswitched loop.
5969 @item max-unswitch-level
5970 The maximum number of branches unswitched in a single loop.
5973 The minimum cost of an expensive expression in the loop invariant motion.
5975 @item iv-consider-all-candidates-bound
5976 Bound on number of candidates for induction variables below that
5977 all candidates are considered for each use in induction variable
5978 optimizations. Only the most relevant candidates are considered
5979 if there are more candidates, to avoid quadratic time complexity.
5981 @item iv-max-considered-uses
5982 The induction variable optimizations give up on loops that contain more
5983 induction variable uses.
5985 @item iv-always-prune-cand-set-bound
5986 If number of candidates in the set is smaller than this value,
5987 we always try to remove unnecessary ivs from the set during its
5988 optimization when a new iv is added to the set.
5990 @item scev-max-expr-size
5991 Bound on size of expressions used in the scalar evolutions analyzer.
5992 Large expressions slow the analyzer.
5994 @item vect-max-version-checks
5995 The maximum number of runtime checks that can be performed when doing
5996 loop versioning in the vectorizer. See option ftree-vect-loop-version
5997 for more information.
5999 @item max-iterations-to-track
6001 The maximum number of iterations of a loop the brute force algorithm
6002 for analysis of # of iterations of the loop tries to evaluate.
6004 @item hot-bb-count-fraction
6005 Select fraction of the maximal count of repetitions of basic block in program
6006 given basic block needs to have to be considered hot.
6008 @item hot-bb-frequency-fraction
6009 Select fraction of the maximal frequency of executions of basic block in
6010 function given basic block needs to have to be considered hot
6012 @item max-predicted-iterations
6013 The maximum number of loop iterations we predict statically. This is useful
6014 in cases where function contain single loop with known bound and other loop
6015 with unknown. We predict the known number of iterations correctly, while
6016 the unknown number of iterations average to roughly 10. This means that the
6017 loop without bounds would appear artificially cold relative to the other one.
6019 @item tracer-dynamic-coverage
6020 @itemx tracer-dynamic-coverage-feedback
6022 This value is used to limit superblock formation once the given percentage of
6023 executed instructions is covered. This limits unnecessary code size
6026 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6027 feedback is available. The real profiles (as opposed to statically estimated
6028 ones) are much less balanced allowing the threshold to be larger value.
6030 @item tracer-max-code-growth
6031 Stop tail duplication once code growth has reached given percentage. This is
6032 rather hokey argument, as most of the duplicates will be eliminated later in
6033 cross jumping, so it may be set to much higher values than is the desired code
6036 @item tracer-min-branch-ratio
6038 Stop reverse growth when the reverse probability of best edge is less than this
6039 threshold (in percent).
6041 @item tracer-min-branch-ratio
6042 @itemx tracer-min-branch-ratio-feedback
6044 Stop forward growth if the best edge do have probability lower than this
6047 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6048 compilation for profile feedback and one for compilation without. The value
6049 for compilation with profile feedback needs to be more conservative (higher) in
6050 order to make tracer effective.
6052 @item max-cse-path-length
6054 Maximum number of basic blocks on path that cse considers. The default is 10.
6057 The maximum instructions CSE process before flushing. The default is 1000.
6059 @item global-var-threshold
6061 Counts the number of function calls (@var{n}) and the number of
6062 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6063 single artificial variable will be created to represent all the
6064 call-clobbered variables at function call sites. This artificial
6065 variable will then be made to alias every call-clobbered variable.
6066 (done as @code{int * size_t} on the host machine; beware overflow).
6068 @item max-aliased-vops
6070 Maximum number of virtual operands allowed to represent aliases
6071 before triggering the alias grouping heuristic. Alias grouping
6072 reduces compile times and memory consumption needed for aliasing at
6073 the expense of precision loss in alias information.
6075 @item ggc-min-expand
6077 GCC uses a garbage collector to manage its own memory allocation. This
6078 parameter specifies the minimum percentage by which the garbage
6079 collector's heap should be allowed to expand between collections.
6080 Tuning this may improve compilation speed; it has no effect on code
6083 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6084 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6085 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6086 GCC is not able to calculate RAM on a particular platform, the lower
6087 bound of 30% is used. Setting this parameter and
6088 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6089 every opportunity. This is extremely slow, but can be useful for
6092 @item ggc-min-heapsize
6094 Minimum size of the garbage collector's heap before it begins bothering
6095 to collect garbage. The first collection occurs after the heap expands
6096 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6097 tuning this may improve compilation speed, and has no effect on code
6100 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6101 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6102 with a lower bound of 4096 (four megabytes) and an upper bound of
6103 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6104 particular platform, the lower bound is used. Setting this parameter
6105 very large effectively disables garbage collection. Setting this
6106 parameter and @option{ggc-min-expand} to zero causes a full collection
6107 to occur at every opportunity.
6109 @item max-reload-search-insns
6110 The maximum number of instruction reload should look backward for equivalent
6111 register. Increasing values mean more aggressive optimization, making the
6112 compile time increase with probably slightly better performance. The default
6115 @item max-cselib-memory-location
6116 The maximum number of memory locations cselib should take into acount.
6117 Increasing values mean more aggressive optimization, making the compile time
6118 increase with probably slightly better performance. The default value is 500.
6120 @item max-flow-memory-location
6121 Similar as @option{max-cselib-memory-location} but for dataflow liveness.
6122 The default value is 100.
6124 @item reorder-blocks-duplicate
6125 @itemx reorder-blocks-duplicate-feedback
6127 Used by basic block reordering pass to decide whether to use unconditional
6128 branch or duplicate the code on its destination. Code is duplicated when its
6129 estimated size is smaller than this value multiplied by the estimated size of
6130 unconditional jump in the hot spots of the program.
6132 The @option{reorder-block-duplicate-feedback} is used only when profile
6133 feedback is available and may be set to higher values than
6134 @option{reorder-block-duplicate} since information about the hot spots is more
6137 @item max-sched-region-blocks
6138 The maximum number of blocks in a region to be considered for
6139 interblock scheduling. The default value is 10.
6141 @item max-sched-region-insns
6142 The maximum number of insns in a region to be considered for
6143 interblock scheduling. The default value is 100.
6145 @item min-sched-prob
6146 The minimum probability of reaching a source block for interblock
6147 speculative scheduling. The default value is 40.
6149 @item max-last-value-rtl
6151 The maximum size measured as number of RTLs that can be recorded in an expression
6152 in combiner for a pseudo register as last known value of that register. The default
6155 @item integer-share-limit
6156 Small integer constants can use a shared data structure, reducing the
6157 compiler's memory usage and increasing its speed. This sets the maximum
6158 value of a shared integer constant's. The default value is 256.
6160 @item min-virtual-mappings
6161 Specifies the minimum number of virtual mappings in the incremental
6162 SSA updater that should be registered to trigger the virtual mappings
6163 heuristic defined by virtual-mappings-ratio. The default value is
6166 @item virtual-mappings-ratio
6167 If the number of virtual mappings is virtual-mappings-ratio bigger
6168 than the number of virtual symbols to be updated, then the incremental
6169 SSA updater switches to a full update for those symbols. The default
6172 @item ssp-buffer-size
6173 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6174 protection when @option{-fstack-protection} is used.
6176 @item max-jump-thread-duplication-stmts
6177 Maximum number of statements allowed in a block that needs to be
6178 duplicated when threading jumps.
6182 @node Preprocessor Options
6183 @section Options Controlling the Preprocessor
6184 @cindex preprocessor options
6185 @cindex options, preprocessor
6187 These options control the C preprocessor, which is run on each C source
6188 file before actual compilation.
6190 If you use the @option{-E} option, nothing is done except preprocessing.
6191 Some of these options make sense only together with @option{-E} because
6192 they cause the preprocessor output to be unsuitable for actual
6197 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6198 and pass @var{option} directly through to the preprocessor. If
6199 @var{option} contains commas, it is split into multiple options at the
6200 commas. However, many options are modified, translated or interpreted
6201 by the compiler driver before being passed to the preprocessor, and
6202 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6203 interface is undocumented and subject to change, so whenever possible
6204 you should avoid using @option{-Wp} and let the driver handle the
6207 @item -Xpreprocessor @var{option}
6208 @opindex preprocessor
6209 Pass @var{option} as an option to the preprocessor. You can use this to
6210 supply system-specific preprocessor options which GCC does not know how to
6213 If you want to pass an option that takes an argument, you must use
6214 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6217 @include cppopts.texi
6219 @node Assembler Options
6220 @section Passing Options to the Assembler
6222 @c prevent bad page break with this line
6223 You can pass options to the assembler.
6226 @item -Wa,@var{option}
6228 Pass @var{option} as an option to the assembler. If @var{option}
6229 contains commas, it is split into multiple options at the commas.
6231 @item -Xassembler @var{option}
6233 Pass @var{option} as an option to the assembler. You can use this to
6234 supply system-specific assembler options which GCC does not know how to
6237 If you want to pass an option that takes an argument, you must use
6238 @option{-Xassembler} twice, once for the option and once for the argument.
6243 @section Options for Linking
6244 @cindex link options
6245 @cindex options, linking
6247 These options come into play when the compiler links object files into
6248 an executable output file. They are meaningless if the compiler is
6249 not doing a link step.
6253 @item @var{object-file-name}
6254 A file name that does not end in a special recognized suffix is
6255 considered to name an object file or library. (Object files are
6256 distinguished from libraries by the linker according to the file
6257 contents.) If linking is done, these object files are used as input
6266 If any of these options is used, then the linker is not run, and
6267 object file names should not be used as arguments. @xref{Overall
6271 @item -l@var{library}
6272 @itemx -l @var{library}
6274 Search the library named @var{library} when linking. (The second
6275 alternative with the library as a separate argument is only for
6276 POSIX compliance and is not recommended.)
6278 It makes a difference where in the command you write this option; the
6279 linker searches and processes libraries and object files in the order they
6280 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6281 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6282 to functions in @samp{z}, those functions may not be loaded.
6284 The linker searches a standard list of directories for the library,
6285 which is actually a file named @file{lib@var{library}.a}. The linker
6286 then uses this file as if it had been specified precisely by name.
6288 The directories searched include several standard system directories
6289 plus any that you specify with @option{-L}.
6291 Normally the files found this way are library files---archive files
6292 whose members are object files. The linker handles an archive file by
6293 scanning through it for members which define symbols that have so far
6294 been referenced but not defined. But if the file that is found is an
6295 ordinary object file, it is linked in the usual fashion. The only
6296 difference between using an @option{-l} option and specifying a file name
6297 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6298 and searches several directories.
6302 You need this special case of the @option{-l} option in order to
6303 link an Objective-C or Objective-C++ program.
6306 @opindex nostartfiles
6307 Do not use the standard system startup files when linking.
6308 The standard system libraries are used normally, unless @option{-nostdlib}
6309 or @option{-nodefaultlibs} is used.
6311 @item -nodefaultlibs
6312 @opindex nodefaultlibs
6313 Do not use the standard system libraries when linking.
6314 Only the libraries you specify will be passed to the linker.
6315 The standard startup files are used normally, unless @option{-nostartfiles}
6316 is used. The compiler may generate calls to @code{memcmp},
6317 @code{memset}, @code{memcpy} and @code{memmove}.
6318 These entries are usually resolved by entries in
6319 libc. These entry points should be supplied through some other
6320 mechanism when this option is specified.
6324 Do not use the standard system startup files or libraries when linking.
6325 No startup files and only the libraries you specify will be passed to
6326 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6327 @code{memcpy} and @code{memmove}.
6328 These entries are usually resolved by entries in
6329 libc. These entry points should be supplied through some other
6330 mechanism when this option is specified.
6332 @cindex @option{-lgcc}, use with @option{-nostdlib}
6333 @cindex @option{-nostdlib} and unresolved references
6334 @cindex unresolved references and @option{-nostdlib}
6335 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6336 @cindex @option{-nodefaultlibs} and unresolved references
6337 @cindex unresolved references and @option{-nodefaultlibs}
6338 One of the standard libraries bypassed by @option{-nostdlib} and
6339 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6340 that GCC uses to overcome shortcomings of particular machines, or special
6341 needs for some languages.
6342 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6343 Collection (GCC) Internals},
6344 for more discussion of @file{libgcc.a}.)
6345 In most cases, you need @file{libgcc.a} even when you want to avoid
6346 other standard libraries. In other words, when you specify @option{-nostdlib}
6347 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6348 This ensures that you have no unresolved references to internal GCC
6349 library subroutines. (For example, @samp{__main}, used to ensure C++
6350 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6351 GNU Compiler Collection (GCC) Internals}.)
6355 Produce a position independent executable on targets which support it.
6356 For predictable results, you must also specify the same set of options
6357 that were used to generate code (@option{-fpie}, @option{-fPIE},
6358 or model suboptions) when you specify this option.
6362 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6363 that support it. This instructs the linker to add all symbols, not
6364 only used ones, to the dynamic symbol table. This option is needed
6365 for some uses of @code{dlopen} or to allow obtaining backtraces
6366 from within a program.
6370 Remove all symbol table and relocation information from the executable.
6374 On systems that support dynamic linking, this prevents linking with the shared
6375 libraries. On other systems, this option has no effect.
6379 Produce a shared object which can then be linked with other objects to
6380 form an executable. Not all systems support this option. For predictable
6381 results, you must also specify the same set of options that were used to
6382 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6383 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6384 needs to build supplementary stub code for constructors to work. On
6385 multi-libbed systems, @samp{gcc -shared} must select the correct support
6386 libraries to link against. Failing to supply the correct flags may lead
6387 to subtle defects. Supplying them in cases where they are not necessary
6390 @item -shared-libgcc
6391 @itemx -static-libgcc
6392 @opindex shared-libgcc
6393 @opindex static-libgcc
6394 On systems that provide @file{libgcc} as a shared library, these options
6395 force the use of either the shared or static version respectively.
6396 If no shared version of @file{libgcc} was built when the compiler was
6397 configured, these options have no effect.
6399 There are several situations in which an application should use the
6400 shared @file{libgcc} instead of the static version. The most common
6401 of these is when the application wishes to throw and catch exceptions
6402 across different shared libraries. In that case, each of the libraries
6403 as well as the application itself should use the shared @file{libgcc}.
6405 Therefore, the G++ and GCJ drivers automatically add
6406 @option{-shared-libgcc} whenever you build a shared library or a main
6407 executable, because C++ and Java programs typically use exceptions, so
6408 this is the right thing to do.
6410 If, instead, you use the GCC driver to create shared libraries, you may
6411 find that they will not always be linked with the shared @file{libgcc}.
6412 If GCC finds, at its configuration time, that you have a non-GNU linker
6413 or a GNU linker that does not support option @option{--eh-frame-hdr},
6414 it will link the shared version of @file{libgcc} into shared libraries
6415 by default. Otherwise, it will take advantage of the linker and optimize
6416 away the linking with the shared version of @file{libgcc}, linking with
6417 the static version of libgcc by default. This allows exceptions to
6418 propagate through such shared libraries, without incurring relocation
6419 costs at library load time.
6421 However, if a library or main executable is supposed to throw or catch
6422 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6423 for the languages used in the program, or using the option
6424 @option{-shared-libgcc}, such that it is linked with the shared
6429 Bind references to global symbols when building a shared object. Warn
6430 about any unresolved references (unless overridden by the link editor
6431 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6434 @item -Xlinker @var{option}
6436 Pass @var{option} as an option to the linker. You can use this to
6437 supply system-specific linker options which GCC does not know how to
6440 If you want to pass an option that takes an argument, you must use
6441 @option{-Xlinker} twice, once for the option and once for the argument.
6442 For example, to pass @option{-assert definitions}, you must write
6443 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6444 @option{-Xlinker "-assert definitions"}, because this passes the entire
6445 string as a single argument, which is not what the linker expects.
6447 @item -Wl,@var{option}
6449 Pass @var{option} as an option to the linker. If @var{option} contains
6450 commas, it is split into multiple options at the commas.
6452 @item -u @var{symbol}
6454 Pretend the symbol @var{symbol} is undefined, to force linking of
6455 library modules to define it. You can use @option{-u} multiple times with
6456 different symbols to force loading of additional library modules.
6459 @node Directory Options
6460 @section Options for Directory Search
6461 @cindex directory options
6462 @cindex options, directory search
6465 These options specify directories to search for header files, for
6466 libraries and for parts of the compiler:
6471 Add the directory @var{dir} to the head of the list of directories to be
6472 searched for header files. This can be used to override a system header
6473 file, substituting your own version, since these directories are
6474 searched before the system header file directories. However, you should
6475 not use this option to add directories that contain vendor-supplied
6476 system header files (use @option{-isystem} for that). If you use more than
6477 one @option{-I} option, the directories are scanned in left-to-right
6478 order; the standard system directories come after.
6480 If a standard system include directory, or a directory specified with
6481 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6482 option will be ignored. The directory will still be searched but as a
6483 system directory at its normal position in the system include chain.
6484 This is to ensure that GCC's procedure to fix buggy system headers and
6485 the ordering for the include_next directive are not inadvertently changed.
6486 If you really need to change the search order for system directories,
6487 use the @option{-nostdinc} and/or @option{-isystem} options.
6489 @item -iquote@var{dir}
6491 Add the directory @var{dir} to the head of the list of directories to
6492 be searched for header files only for the case of @samp{#include
6493 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6494 otherwise just like @option{-I}.
6498 Add directory @var{dir} to the list of directories to be searched
6501 @item -B@var{prefix}
6503 This option specifies where to find the executables, libraries,
6504 include files, and data files of the compiler itself.
6506 The compiler driver program runs one or more of the subprograms
6507 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6508 @var{prefix} as a prefix for each program it tries to run, both with and
6509 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6511 For each subprogram to be run, the compiler driver first tries the
6512 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6513 was not specified, the driver tries two standard prefixes, which are
6514 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6515 those results in a file name that is found, the unmodified program
6516 name is searched for using the directories specified in your
6517 @env{PATH} environment variable.
6519 The compiler will check to see if the path provided by the @option{-B}
6520 refers to a directory, and if necessary it will add a directory
6521 separator character at the end of the path.
6523 @option{-B} prefixes that effectively specify directory names also apply
6524 to libraries in the linker, because the compiler translates these
6525 options into @option{-L} options for the linker. They also apply to
6526 includes files in the preprocessor, because the compiler translates these
6527 options into @option{-isystem} options for the preprocessor. In this case,
6528 the compiler appends @samp{include} to the prefix.
6530 The run-time support file @file{libgcc.a} can also be searched for using
6531 the @option{-B} prefix, if needed. If it is not found there, the two
6532 standard prefixes above are tried, and that is all. The file is left
6533 out of the link if it is not found by those means.
6535 Another way to specify a prefix much like the @option{-B} prefix is to use
6536 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6539 As a special kludge, if the path provided by @option{-B} is
6540 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6541 9, then it will be replaced by @file{[dir/]include}. This is to help
6542 with boot-strapping the compiler.
6544 @item -specs=@var{file}
6546 Process @var{file} after the compiler reads in the standard @file{specs}
6547 file, in order to override the defaults that the @file{gcc} driver
6548 program uses when determining what switches to pass to @file{cc1},
6549 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6550 @option{-specs=@var{file}} can be specified on the command line, and they
6551 are processed in order, from left to right.
6553 @item --sysroot=@var{dir}
6555 Use @var{dir} as the logical root directory for headers and libraries.
6556 For example, if the compiler would normally search for headers in
6557 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6558 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6560 If you use both this option and the @option{-isysroot} option, then
6561 the @option{--sysroot} option will apply to libraries, but the
6562 @option{-isysroot} option will apply to header files.
6564 The GNU linker (beginning with version 2.16) has the necessary support
6565 for this option. If your linker does not support this option, the
6566 header file aspect of @option{--sysroot} will still work, but the
6567 library aspect will not.
6571 This option has been deprecated. Please use @option{-iquote} instead for
6572 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6573 Any directories you specify with @option{-I} options before the @option{-I-}
6574 option are searched only for the case of @samp{#include "@var{file}"};
6575 they are not searched for @samp{#include <@var{file}>}.
6577 If additional directories are specified with @option{-I} options after
6578 the @option{-I-}, these directories are searched for all @samp{#include}
6579 directives. (Ordinarily @emph{all} @option{-I} directories are used
6582 In addition, the @option{-I-} option inhibits the use of the current
6583 directory (where the current input file came from) as the first search
6584 directory for @samp{#include "@var{file}"}. There is no way to
6585 override this effect of @option{-I-}. With @option{-I.} you can specify
6586 searching the directory which was current when the compiler was
6587 invoked. That is not exactly the same as what the preprocessor does
6588 by default, but it is often satisfactory.
6590 @option{-I-} does not inhibit the use of the standard system directories
6591 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6598 @section Specifying subprocesses and the switches to pass to them
6601 @command{gcc} is a driver program. It performs its job by invoking a
6602 sequence of other programs to do the work of compiling, assembling and
6603 linking. GCC interprets its command-line parameters and uses these to
6604 deduce which programs it should invoke, and which command-line options
6605 it ought to place on their command lines. This behavior is controlled
6606 by @dfn{spec strings}. In most cases there is one spec string for each
6607 program that GCC can invoke, but a few programs have multiple spec
6608 strings to control their behavior. The spec strings built into GCC can
6609 be overridden by using the @option{-specs=} command-line switch to specify
6612 @dfn{Spec files} are plaintext files that are used to construct spec
6613 strings. They consist of a sequence of directives separated by blank
6614 lines. The type of directive is determined by the first non-whitespace
6615 character on the line and it can be one of the following:
6618 @item %@var{command}
6619 Issues a @var{command} to the spec file processor. The commands that can
6623 @item %include <@var{file}>
6625 Search for @var{file} and insert its text at the current point in the
6628 @item %include_noerr <@var{file}>
6629 @cindex %include_noerr
6630 Just like @samp{%include}, but do not generate an error message if the include
6631 file cannot be found.
6633 @item %rename @var{old_name} @var{new_name}
6635 Rename the spec string @var{old_name} to @var{new_name}.
6639 @item *[@var{spec_name}]:
6640 This tells the compiler to create, override or delete the named spec
6641 string. All lines after this directive up to the next directive or
6642 blank line are considered to be the text for the spec string. If this
6643 results in an empty string then the spec will be deleted. (Or, if the
6644 spec did not exist, then nothing will happened.) Otherwise, if the spec
6645 does not currently exist a new spec will be created. If the spec does
6646 exist then its contents will be overridden by the text of this
6647 directive, unless the first character of that text is the @samp{+}
6648 character, in which case the text will be appended to the spec.
6650 @item [@var{suffix}]:
6651 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6652 and up to the next directive or blank line are considered to make up the
6653 spec string for the indicated suffix. When the compiler encounters an
6654 input file with the named suffix, it will processes the spec string in
6655 order to work out how to compile that file. For example:
6662 This says that any input file whose name ends in @samp{.ZZ} should be
6663 passed to the program @samp{z-compile}, which should be invoked with the
6664 command-line switch @option{-input} and with the result of performing the
6665 @samp{%i} substitution. (See below.)
6667 As an alternative to providing a spec string, the text that follows a
6668 suffix directive can be one of the following:
6671 @item @@@var{language}
6672 This says that the suffix is an alias for a known @var{language}. This is
6673 similar to using the @option{-x} command-line switch to GCC to specify a
6674 language explicitly. For example:
6681 Says that .ZZ files are, in fact, C++ source files.
6684 This causes an error messages saying:
6687 @var{name} compiler not installed on this system.
6691 GCC already has an extensive list of suffixes built into it.
6692 This directive will add an entry to the end of the list of suffixes, but
6693 since the list is searched from the end backwards, it is effectively
6694 possible to override earlier entries using this technique.
6698 GCC has the following spec strings built into it. Spec files can
6699 override these strings or create their own. Note that individual
6700 targets can also add their own spec strings to this list.
6703 asm Options to pass to the assembler
6704 asm_final Options to pass to the assembler post-processor
6705 cpp Options to pass to the C preprocessor
6706 cc1 Options to pass to the C compiler
6707 cc1plus Options to pass to the C++ compiler
6708 endfile Object files to include at the end of the link
6709 link Options to pass to the linker
6710 lib Libraries to include on the command line to the linker
6711 libgcc Decides which GCC support library to pass to the linker
6712 linker Sets the name of the linker
6713 predefines Defines to be passed to the C preprocessor
6714 signed_char Defines to pass to CPP to say whether @code{char} is signed
6716 startfile Object files to include at the start of the link
6719 Here is a small example of a spec file:
6725 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6728 This example renames the spec called @samp{lib} to @samp{old_lib} and
6729 then overrides the previous definition of @samp{lib} with a new one.
6730 The new definition adds in some extra command-line options before
6731 including the text of the old definition.
6733 @dfn{Spec strings} are a list of command-line options to be passed to their
6734 corresponding program. In addition, the spec strings can contain
6735 @samp{%}-prefixed sequences to substitute variable text or to
6736 conditionally insert text into the command line. Using these constructs
6737 it is possible to generate quite complex command lines.
6739 Here is a table of all defined @samp{%}-sequences for spec
6740 strings. Note that spaces are not generated automatically around the
6741 results of expanding these sequences. Therefore you can concatenate them
6742 together or combine them with constant text in a single argument.
6746 Substitute one @samp{%} into the program name or argument.
6749 Substitute the name of the input file being processed.
6752 Substitute the basename of the input file being processed.
6753 This is the substring up to (and not including) the last period
6754 and not including the directory.
6757 This is the same as @samp{%b}, but include the file suffix (text after
6761 Marks the argument containing or following the @samp{%d} as a
6762 temporary file name, so that that file will be deleted if GCC exits
6763 successfully. Unlike @samp{%g}, this contributes no text to the
6766 @item %g@var{suffix}
6767 Substitute a file name that has suffix @var{suffix} and is chosen
6768 once per compilation, and mark the argument in the same way as
6769 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6770 name is now chosen in a way that is hard to predict even when previously
6771 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6772 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6773 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6774 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6775 was simply substituted with a file name chosen once per compilation,
6776 without regard to any appended suffix (which was therefore treated
6777 just like ordinary text), making such attacks more likely to succeed.
6779 @item %u@var{suffix}
6780 Like @samp{%g}, but generates a new temporary file name even if
6781 @samp{%u@var{suffix}} was already seen.
6783 @item %U@var{suffix}
6784 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6785 new one if there is no such last file name. In the absence of any
6786 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6787 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6788 would involve the generation of two distinct file names, one
6789 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6790 simply substituted with a file name chosen for the previous @samp{%u},
6791 without regard to any appended suffix.
6793 @item %j@var{suffix}
6794 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6795 writable, and if save-temps is off; otherwise, substitute the name
6796 of a temporary file, just like @samp{%u}. This temporary file is not
6797 meant for communication between processes, but rather as a junk
6800 @item %|@var{suffix}
6801 @itemx %m@var{suffix}
6802 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6803 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6804 all. These are the two most common ways to instruct a program that it
6805 should read from standard input or write to standard output. If you
6806 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6807 construct: see for example @file{f/lang-specs.h}.
6809 @item %.@var{SUFFIX}
6810 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6811 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6812 terminated by the next space or %.
6815 Marks the argument containing or following the @samp{%w} as the
6816 designated output file of this compilation. This puts the argument
6817 into the sequence of arguments that @samp{%o} will substitute later.
6820 Substitutes the names of all the output files, with spaces
6821 automatically placed around them. You should write spaces
6822 around the @samp{%o} as well or the results are undefined.
6823 @samp{%o} is for use in the specs for running the linker.
6824 Input files whose names have no recognized suffix are not compiled
6825 at all, but they are included among the output files, so they will
6829 Substitutes the suffix for object files. Note that this is
6830 handled specially when it immediately follows @samp{%g, %u, or %U},
6831 because of the need for those to form complete file names. The
6832 handling is such that @samp{%O} is treated exactly as if it had already
6833 been substituted, except that @samp{%g, %u, and %U} do not currently
6834 support additional @var{suffix} characters following @samp{%O} as they would
6835 following, for example, @samp{.o}.
6838 Substitutes the standard macro predefinitions for the
6839 current target machine. Use this when running @code{cpp}.
6842 Like @samp{%p}, but puts @samp{__} before and after the name of each
6843 predefined macro, except for macros that start with @samp{__} or with
6844 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6848 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6849 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6850 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6854 Current argument is the name of a library or startup file of some sort.
6855 Search for that file in a standard list of directories and substitute
6856 the full name found.
6859 Print @var{str} as an error message. @var{str} is terminated by a newline.
6860 Use this when inconsistent options are detected.
6863 Substitute the contents of spec string @var{name} at this point.
6866 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6868 @item %x@{@var{option}@}
6869 Accumulate an option for @samp{%X}.
6872 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6876 Output the accumulated assembler options specified by @option{-Wa}.
6879 Output the accumulated preprocessor options specified by @option{-Wp}.
6882 Process the @code{asm} spec. This is used to compute the
6883 switches to be passed to the assembler.
6886 Process the @code{asm_final} spec. This is a spec string for
6887 passing switches to an assembler post-processor, if such a program is
6891 Process the @code{link} spec. This is the spec for computing the
6892 command line passed to the linker. Typically it will make use of the
6893 @samp{%L %G %S %D and %E} sequences.
6896 Dump out a @option{-L} option for each directory that GCC believes might
6897 contain startup files. If the target supports multilibs then the
6898 current multilib directory will be prepended to each of these paths.
6901 Process the @code{lib} spec. This is a spec string for deciding which
6902 libraries should be included on the command line to the linker.
6905 Process the @code{libgcc} spec. This is a spec string for deciding
6906 which GCC support library should be included on the command line to the linker.
6909 Process the @code{startfile} spec. This is a spec for deciding which
6910 object files should be the first ones passed to the linker. Typically
6911 this might be a file named @file{crt0.o}.
6914 Process the @code{endfile} spec. This is a spec string that specifies
6915 the last object files that will be passed to the linker.
6918 Process the @code{cpp} spec. This is used to construct the arguments
6919 to be passed to the C preprocessor.
6922 Process the @code{cc1} spec. This is used to construct the options to be
6923 passed to the actual C compiler (@samp{cc1}).
6926 Process the @code{cc1plus} spec. This is used to construct the options to be
6927 passed to the actual C++ compiler (@samp{cc1plus}).
6930 Substitute the variable part of a matched option. See below.
6931 Note that each comma in the substituted string is replaced by
6935 Remove all occurrences of @code{-S} from the command line. Note---this
6936 command is position dependent. @samp{%} commands in the spec string
6937 before this one will see @code{-S}, @samp{%} commands in the spec string
6938 after this one will not.
6940 @item %:@var{function}(@var{args})
6941 Call the named function @var{function}, passing it @var{args}.
6942 @var{args} is first processed as a nested spec string, then split
6943 into an argument vector in the usual fashion. The function returns
6944 a string which is processed as if it had appeared literally as part
6945 of the current spec.
6947 The following built-in spec functions are provided:
6950 @item @code{if-exists}
6951 The @code{if-exists} spec function takes one argument, an absolute
6952 pathname to a file. If the file exists, @code{if-exists} returns the
6953 pathname. Here is a small example of its usage:
6957 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6960 @item @code{if-exists-else}
6961 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6962 spec function, except that it takes two arguments. The first argument is
6963 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6964 returns the pathname. If it does not exist, it returns the second argument.
6965 This way, @code{if-exists-else} can be used to select one file or another,
6966 based on the existence of the first. Here is a small example of its usage:
6970 crt0%O%s %:if-exists(crti%O%s) \
6971 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6974 @item @code{replace-outfile}
6975 The @code{replace-outfile} spec function takes two arguments. It looks for the
6976 first argument in the outfiles array and replaces it with the second argument. Here
6977 is a small example of its usage:
6980 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6986 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6987 If that switch was not specified, this substitutes nothing. Note that
6988 the leading dash is omitted when specifying this option, and it is
6989 automatically inserted if the substitution is performed. Thus the spec
6990 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6991 and would output the command line option @option{-foo}.
6993 @item %W@{@code{S}@}
6994 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6997 @item %@{@code{S}*@}
6998 Substitutes all the switches specified to GCC whose names start
6999 with @code{-S}, but which also take an argument. This is used for
7000 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7001 GCC considers @option{-o foo} as being
7002 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7003 text, including the space. Thus two arguments would be generated.
7005 @item %@{@code{S}*&@code{T}*@}
7006 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7007 (the order of @code{S} and @code{T} in the spec is not significant).
7008 There can be any number of ampersand-separated variables; for each the
7009 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7011 @item %@{@code{S}:@code{X}@}
7012 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7014 @item %@{!@code{S}:@code{X}@}
7015 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7017 @item %@{@code{S}*:@code{X}@}
7018 Substitutes @code{X} if one or more switches whose names start with
7019 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7020 once, no matter how many such switches appeared. However, if @code{%*}
7021 appears somewhere in @code{X}, then @code{X} will be substituted once
7022 for each matching switch, with the @code{%*} replaced by the part of
7023 that switch that matched the @code{*}.
7025 @item %@{.@code{S}:@code{X}@}
7026 Substitutes @code{X}, if processing a file with suffix @code{S}.
7028 @item %@{!.@code{S}:@code{X}@}
7029 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7031 @item %@{@code{S}|@code{P}:@code{X}@}
7032 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7033 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7034 although they have a stronger binding than the @samp{|}. If @code{%*}
7035 appears in @code{X}, all of the alternatives must be starred, and only
7036 the first matching alternative is substituted.
7038 For example, a spec string like this:
7041 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7044 will output the following command-line options from the following input
7045 command-line options:
7050 -d fred.c -foo -baz -boggle
7051 -d jim.d -bar -baz -boggle
7054 @item %@{S:X; T:Y; :D@}
7056 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7057 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7058 be as many clauses as you need. This may be combined with @code{.},
7059 @code{!}, @code{|}, and @code{*} as needed.
7064 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7065 construct may contain other nested @samp{%} constructs or spaces, or
7066 even newlines. They are processed as usual, as described above.
7067 Trailing white space in @code{X} is ignored. White space may also
7068 appear anywhere on the left side of the colon in these constructs,
7069 except between @code{.} or @code{*} and the corresponding word.
7071 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7072 handled specifically in these constructs. If another value of
7073 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7074 @option{-W} switch is found later in the command line, the earlier
7075 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7076 just one letter, which passes all matching options.
7078 The character @samp{|} at the beginning of the predicate text is used to
7079 indicate that a command should be piped to the following command, but
7080 only if @option{-pipe} is specified.
7082 It is built into GCC which switches take arguments and which do not.
7083 (You might think it would be useful to generalize this to allow each
7084 compiler's spec to say which switches take arguments. But this cannot
7085 be done in a consistent fashion. GCC cannot even decide which input
7086 files have been specified without knowing which switches take arguments,
7087 and it must know which input files to compile in order to tell which
7090 GCC also knows implicitly that arguments starting in @option{-l} are to be
7091 treated as compiler output files, and passed to the linker in their
7092 proper position among the other output files.
7094 @c man begin OPTIONS
7096 @node Target Options
7097 @section Specifying Target Machine and Compiler Version
7098 @cindex target options
7099 @cindex cross compiling
7100 @cindex specifying machine version
7101 @cindex specifying compiler version and target machine
7102 @cindex compiler version, specifying
7103 @cindex target machine, specifying
7105 The usual way to run GCC is to run the executable called @file{gcc}, or
7106 @file{<machine>-gcc} when cross-compiling, or
7107 @file{<machine>-gcc-<version>} to run a version other than the one that
7108 was installed last. Sometimes this is inconvenient, so GCC provides
7109 options that will switch to another cross-compiler or version.
7112 @item -b @var{machine}
7114 The argument @var{machine} specifies the target machine for compilation.
7116 The value to use for @var{machine} is the same as was specified as the
7117 machine type when configuring GCC as a cross-compiler. For
7118 example, if a cross-compiler was configured with @samp{configure
7119 arm-elf}, meaning to compile for an arm processor with elf binaries,
7120 then you would specify @option{-b arm-elf} to run that cross compiler.
7121 Because there are other options beginning with @option{-b}, the
7122 configuration must contain a hyphen.
7124 @item -V @var{version}
7126 The argument @var{version} specifies which version of GCC to run.
7127 This is useful when multiple versions are installed. For example,
7128 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7131 The @option{-V} and @option{-b} options work by running the
7132 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7133 use them if you can just run that directly.
7135 @node Submodel Options
7136 @section Hardware Models and Configurations
7137 @cindex submodel options
7138 @cindex specifying hardware config
7139 @cindex hardware models and configurations, specifying
7140 @cindex machine dependent options
7142 Earlier we discussed the standard option @option{-b} which chooses among
7143 different installed compilers for completely different target
7144 machines, such as VAX vs.@: 68000 vs.@: 80386.
7146 In addition, each of these target machine types can have its own
7147 special options, starting with @samp{-m}, to choose among various
7148 hardware models or configurations---for example, 68010 vs 68020,
7149 floating coprocessor or none. A single installed version of the
7150 compiler can compile for any model or configuration, according to the
7153 Some configurations of the compiler also support additional special
7154 options, usually for compatibility with other compilers on the same
7157 @c This list is ordered alphanumerically by subsection name.
7158 @c It should be the same order and spelling as these options are listed
7159 @c in Machine Dependent Options
7165 * Blackfin Options::
7169 * DEC Alpha Options::
7170 * DEC Alpha/VMS Options::
7174 * i386 and x86-64 Options::
7187 * RS/6000 and PowerPC Options::
7188 * S/390 and zSeries Options::
7191 * System V Options::
7192 * TMS320C3x/C4x Options::
7196 * Xstormy16 Options::
7202 @subsection ARC Options
7205 These options are defined for ARC implementations:
7210 Compile code for little endian mode. This is the default.
7214 Compile code for big endian mode.
7217 @opindex mmangle-cpu
7218 Prepend the name of the cpu to all public symbol names.
7219 In multiple-processor systems, there are many ARC variants with different
7220 instruction and register set characteristics. This flag prevents code
7221 compiled for one cpu to be linked with code compiled for another.
7222 No facility exists for handling variants that are ``almost identical''.
7223 This is an all or nothing option.
7225 @item -mcpu=@var{cpu}
7227 Compile code for ARC variant @var{cpu}.
7228 Which variants are supported depend on the configuration.
7229 All variants support @option{-mcpu=base}, this is the default.
7231 @item -mtext=@var{text-section}
7232 @itemx -mdata=@var{data-section}
7233 @itemx -mrodata=@var{readonly-data-section}
7237 Put functions, data, and readonly data in @var{text-section},
7238 @var{data-section}, and @var{readonly-data-section} respectively
7239 by default. This can be overridden with the @code{section} attribute.
7240 @xref{Variable Attributes}.
7245 @subsection ARM Options
7248 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7252 @item -mabi=@var{name}
7254 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7255 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7258 @opindex mapcs-frame
7259 Generate a stack frame that is compliant with the ARM Procedure Call
7260 Standard for all functions, even if this is not strictly necessary for
7261 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7262 with this option will cause the stack frames not to be generated for
7263 leaf functions. The default is @option{-mno-apcs-frame}.
7267 This is a synonym for @option{-mapcs-frame}.
7270 @c not currently implemented
7271 @item -mapcs-stack-check
7272 @opindex mapcs-stack-check
7273 Generate code to check the amount of stack space available upon entry to
7274 every function (that actually uses some stack space). If there is
7275 insufficient space available then either the function
7276 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7277 called, depending upon the amount of stack space required. The run time
7278 system is required to provide these functions. The default is
7279 @option{-mno-apcs-stack-check}, since this produces smaller code.
7281 @c not currently implemented
7283 @opindex mapcs-float
7284 Pass floating point arguments using the float point registers. This is
7285 one of the variants of the APCS@. This option is recommended if the
7286 target hardware has a floating point unit or if a lot of floating point
7287 arithmetic is going to be performed by the code. The default is
7288 @option{-mno-apcs-float}, since integer only code is slightly increased in
7289 size if @option{-mapcs-float} is used.
7291 @c not currently implemented
7292 @item -mapcs-reentrant
7293 @opindex mapcs-reentrant
7294 Generate reentrant, position independent code. The default is
7295 @option{-mno-apcs-reentrant}.
7298 @item -mthumb-interwork
7299 @opindex mthumb-interwork
7300 Generate code which supports calling between the ARM and Thumb
7301 instruction sets. Without this option the two instruction sets cannot
7302 be reliably used inside one program. The default is
7303 @option{-mno-thumb-interwork}, since slightly larger code is generated
7304 when @option{-mthumb-interwork} is specified.
7306 @item -mno-sched-prolog
7307 @opindex mno-sched-prolog
7308 Prevent the reordering of instructions in the function prolog, or the
7309 merging of those instruction with the instructions in the function's
7310 body. This means that all functions will start with a recognizable set
7311 of instructions (or in fact one of a choice from a small set of
7312 different function prologues), and this information can be used to
7313 locate the start if functions inside an executable piece of code. The
7314 default is @option{-msched-prolog}.
7317 @opindex mhard-float
7318 Generate output containing floating point instructions. This is the
7322 @opindex msoft-float
7323 Generate output containing library calls for floating point.
7324 @strong{Warning:} the requisite libraries are not available for all ARM
7325 targets. Normally the facilities of the machine's usual C compiler are
7326 used, but this cannot be done directly in cross-compilation. You must make
7327 your own arrangements to provide suitable library functions for
7330 @option{-msoft-float} changes the calling convention in the output file;
7331 therefore, it is only useful if you compile @emph{all} of a program with
7332 this option. In particular, you need to compile @file{libgcc.a}, the
7333 library that comes with GCC, with @option{-msoft-float} in order for
7336 @item -mfloat-abi=@var{name}
7338 Specifies which ABI to use for floating point values. Permissible values
7339 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7341 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7342 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7343 of floating point instructions, but still uses the soft-float calling
7346 @item -mlittle-endian
7347 @opindex mlittle-endian
7348 Generate code for a processor running in little-endian mode. This is
7349 the default for all standard configurations.
7352 @opindex mbig-endian
7353 Generate code for a processor running in big-endian mode; the default is
7354 to compile code for a little-endian processor.
7356 @item -mwords-little-endian
7357 @opindex mwords-little-endian
7358 This option only applies when generating code for big-endian processors.
7359 Generate code for a little-endian word order but a big-endian byte
7360 order. That is, a byte order of the form @samp{32107654}. Note: this
7361 option should only be used if you require compatibility with code for
7362 big-endian ARM processors generated by versions of the compiler prior to
7365 @item -mcpu=@var{name}
7367 This specifies the name of the target ARM processor. GCC uses this name
7368 to determine what kind of instructions it can emit when generating
7369 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7370 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7371 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7372 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7373 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7374 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7375 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7376 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7377 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7378 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7379 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7380 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7381 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7382 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7385 @itemx -mtune=@var{name}
7387 This option is very similar to the @option{-mcpu=} option, except that
7388 instead of specifying the actual target processor type, and hence
7389 restricting which instructions can be used, it specifies that GCC should
7390 tune the performance of the code as if the target were of the type
7391 specified in this option, but still choosing the instructions that it
7392 will generate based on the cpu specified by a @option{-mcpu=} option.
7393 For some ARM implementations better performance can be obtained by using
7396 @item -march=@var{name}
7398 This specifies the name of the target ARM architecture. GCC uses this
7399 name to determine what kind of instructions it can emit when generating
7400 assembly code. This option can be used in conjunction with or instead
7401 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7402 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7403 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7404 @samp{iwmmxt}, @samp{ep9312}.
7406 @item -mfpu=@var{name}
7407 @itemx -mfpe=@var{number}
7408 @itemx -mfp=@var{number}
7412 This specifies what floating point hardware (or hardware emulation) is
7413 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7414 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7415 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7416 with older versions of GCC@.
7418 If @option{-msoft-float} is specified this specifies the format of
7419 floating point values.
7421 @item -mstructure-size-boundary=@var{n}
7422 @opindex mstructure-size-boundary
7423 The size of all structures and unions will be rounded up to a multiple
7424 of the number of bits set by this option. Permissible values are 8, 32
7425 and 64. The default value varies for different toolchains. For the COFF
7426 targeted toolchain the default value is 8. A value of 64 is only allowed
7427 if the underlying ABI supports it.
7429 Specifying the larger number can produce faster, more efficient code, but
7430 can also increase the size of the program. Different values are potentially
7431 incompatible. Code compiled with one value cannot necessarily expect to
7432 work with code or libraries compiled with another value, if they exchange
7433 information using structures or unions.
7435 @item -mabort-on-noreturn
7436 @opindex mabort-on-noreturn
7437 Generate a call to the function @code{abort} at the end of a
7438 @code{noreturn} function. It will be executed if the function tries to
7442 @itemx -mno-long-calls
7443 @opindex mlong-calls
7444 @opindex mno-long-calls
7445 Tells the compiler to perform function calls by first loading the
7446 address of the function into a register and then performing a subroutine
7447 call on this register. This switch is needed if the target function
7448 will lie outside of the 64 megabyte addressing range of the offset based
7449 version of subroutine call instruction.
7451 Even if this switch is enabled, not all function calls will be turned
7452 into long calls. The heuristic is that static functions, functions
7453 which have the @samp{short-call} attribute, functions that are inside
7454 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7455 definitions have already been compiled within the current compilation
7456 unit, will not be turned into long calls. The exception to this rule is
7457 that weak function definitions, functions with the @samp{long-call}
7458 attribute or the @samp{section} attribute, and functions that are within
7459 the scope of a @samp{#pragma long_calls} directive, will always be
7460 turned into long calls.
7462 This feature is not enabled by default. Specifying
7463 @option{-mno-long-calls} will restore the default behavior, as will
7464 placing the function calls within the scope of a @samp{#pragma
7465 long_calls_off} directive. Note these switches have no effect on how
7466 the compiler generates code to handle function calls via function
7469 @item -mnop-fun-dllimport
7470 @opindex mnop-fun-dllimport
7471 Disable support for the @code{dllimport} attribute.
7473 @item -msingle-pic-base
7474 @opindex msingle-pic-base
7475 Treat the register used for PIC addressing as read-only, rather than
7476 loading it in the prologue for each function. The run-time system is
7477 responsible for initializing this register with an appropriate value
7478 before execution begins.
7480 @item -mpic-register=@var{reg}
7481 @opindex mpic-register
7482 Specify the register to be used for PIC addressing. The default is R10
7483 unless stack-checking is enabled, when R9 is used.
7485 @item -mcirrus-fix-invalid-insns
7486 @opindex mcirrus-fix-invalid-insns
7487 @opindex mno-cirrus-fix-invalid-insns
7488 Insert NOPs into the instruction stream to in order to work around
7489 problems with invalid Maverick instruction combinations. This option
7490 is only valid if the @option{-mcpu=ep9312} option has been used to
7491 enable generation of instructions for the Cirrus Maverick floating
7492 point co-processor. This option is not enabled by default, since the
7493 problem is only present in older Maverick implementations. The default
7494 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7497 @item -mpoke-function-name
7498 @opindex mpoke-function-name
7499 Write the name of each function into the text section, directly
7500 preceding the function prologue. The generated code is similar to this:
7504 .ascii "arm_poke_function_name", 0
7507 .word 0xff000000 + (t1 - t0)
7508 arm_poke_function_name
7510 stmfd sp!, @{fp, ip, lr, pc@}
7514 When performing a stack backtrace, code can inspect the value of
7515 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7516 location @code{pc - 12} and the top 8 bits are set, then we know that
7517 there is a function name embedded immediately preceding this location
7518 and has length @code{((pc[-3]) & 0xff000000)}.
7522 Generate code for the 16-bit Thumb instruction set. The default is to
7523 use the 32-bit ARM instruction set.
7526 @opindex mtpcs-frame
7527 Generate a stack frame that is compliant with the Thumb Procedure Call
7528 Standard for all non-leaf functions. (A leaf function is one that does
7529 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7531 @item -mtpcs-leaf-frame
7532 @opindex mtpcs-leaf-frame
7533 Generate a stack frame that is compliant with the Thumb Procedure Call
7534 Standard for all leaf functions. (A leaf function is one that does
7535 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7537 @item -mcallee-super-interworking
7538 @opindex mcallee-super-interworking
7539 Gives all externally visible functions in the file being compiled an ARM
7540 instruction set header which switches to Thumb mode before executing the
7541 rest of the function. This allows these functions to be called from
7542 non-interworking code.
7544 @item -mcaller-super-interworking
7545 @opindex mcaller-super-interworking
7546 Allows calls via function pointers (including virtual functions) to
7547 execute correctly regardless of whether the target code has been
7548 compiled for interworking or not. There is a small overhead in the cost
7549 of executing a function pointer if this option is enabled.
7551 @item -mtp=@var{name}
7553 Specify the access model for the thread local storage pointer. The valid
7554 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7555 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7556 (supported in the arm6k architecture), and @option{auto}, which uses the
7557 best available method for the selected processor. The default setting is
7563 @subsection AVR Options
7566 These options are defined for AVR implementations:
7569 @item -mmcu=@var{mcu}
7571 Specify ATMEL AVR instruction set or MCU type.
7573 Instruction set avr1 is for the minimal AVR core, not supported by the C
7574 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7575 attiny11, attiny12, attiny15, attiny28).
7577 Instruction set avr2 (default) is for the classic AVR core with up to
7578 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7579 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7580 at90c8534, at90s8535).
7582 Instruction set avr3 is for the classic AVR core with up to 128K program
7583 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7585 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7586 memory space (MCU types: atmega8, atmega83, atmega85).
7588 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7589 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7590 atmega64, atmega128, at43usb355, at94k).
7594 Output instruction sizes to the asm file.
7596 @item -minit-stack=@var{N}
7597 @opindex minit-stack
7598 Specify the initial stack address, which may be a symbol or numeric value,
7599 @samp{__stack} is the default.
7601 @item -mno-interrupts
7602 @opindex mno-interrupts
7603 Generated code is not compatible with hardware interrupts.
7604 Code size will be smaller.
7606 @item -mcall-prologues
7607 @opindex mcall-prologues
7608 Functions prologues/epilogues expanded as call to appropriate
7609 subroutines. Code size will be smaller.
7611 @item -mno-tablejump
7612 @opindex mno-tablejump
7613 Do not generate tablejump insns which sometimes increase code size.
7616 @opindex mtiny-stack
7617 Change only the low 8 bits of the stack pointer.
7621 Assume int to be 8 bit integer. This affects the sizes of all types: A
7622 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7623 and long long will be 4 bytes. Please note that this option does not
7624 comply to the C standards, but it will provide you with smaller code
7628 @node Blackfin Options
7629 @subsection Blackfin Options
7630 @cindex Blackfin Options
7633 @item -momit-leaf-frame-pointer
7634 @opindex momit-leaf-frame-pointer
7635 Don't keep the frame pointer in a register for leaf functions. This
7636 avoids the instructions to save, set up and restore frame pointers and
7637 makes an extra register available in leaf functions. The option
7638 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7639 which might make debugging harder.
7641 @item -mspecld-anomaly
7642 @opindex mspecld-anomaly
7643 When enabled, the compiler will ensure that the generated code does not
7644 contain speculative loads after jump instructions. This option is enabled
7647 @item -mno-specld-anomaly
7648 @opindex mno-specld-anomaly
7649 Don't generate extra code to prevent speculative loads from occurring.
7651 @item -mcsync-anomaly
7652 @opindex mcsync-anomaly
7653 When enabled, the compiler will ensure that the generated code does not
7654 contain CSYNC or SSYNC instructions too soon after conditional branches.
7655 This option is enabled by default.
7657 @item -mno-csync-anomaly
7658 @opindex mno-csync-anomaly
7659 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7660 occurring too soon after a conditional branch.
7664 When enabled, the compiler is free to take advantage of the knowledge that
7665 the entire program fits into the low 64k of memory.
7668 @opindex mno-low-64k
7669 Assume that the program is arbitrarily large. This is the default.
7671 @item -mid-shared-library
7672 @opindex mid-shared-library
7673 Generate code that supports shared libraries via the library ID method.
7674 This allows for execute in place and shared libraries in an environment
7675 without virtual memory management. This option implies @option{-fPIC}.
7677 @item -mno-id-shared-library
7678 @opindex mno-id-shared-library
7679 Generate code that doesn't assume ID based shared libraries are being used.
7680 This is the default.
7682 @item -mshared-library-id=n
7683 @opindex mshared-library-id
7684 Specified the identification number of the ID based shared library being
7685 compiled. Specifying a value of 0 will generate more compact code, specifying
7686 other values will force the allocation of that number to the current
7687 library but is no more space or time efficient than omitting this option.
7690 @itemx -mno-long-calls
7691 @opindex mlong-calls
7692 @opindex mno-long-calls
7693 Tells the compiler to perform function calls by first loading the
7694 address of the function into a register and then performing a subroutine
7695 call on this register. This switch is needed if the target function
7696 will lie outside of the 24 bit addressing range of the offset based
7697 version of subroutine call instruction.
7699 This feature is not enabled by default. Specifying
7700 @option{-mno-long-calls} will restore the default behavior. Note these
7701 switches have no effect on how the compiler generates code to handle
7702 function calls via function pointers.
7706 @subsection CRIS Options
7707 @cindex CRIS Options
7709 These options are defined specifically for the CRIS ports.
7712 @item -march=@var{architecture-type}
7713 @itemx -mcpu=@var{architecture-type}
7716 Generate code for the specified architecture. The choices for
7717 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7718 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7719 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7722 @item -mtune=@var{architecture-type}
7724 Tune to @var{architecture-type} everything applicable about the generated
7725 code, except for the ABI and the set of available instructions. The
7726 choices for @var{architecture-type} are the same as for
7727 @option{-march=@var{architecture-type}}.
7729 @item -mmax-stack-frame=@var{n}
7730 @opindex mmax-stack-frame
7731 Warn when the stack frame of a function exceeds @var{n} bytes.
7733 @item -melinux-stacksize=@var{n}
7734 @opindex melinux-stacksize
7735 Only available with the @samp{cris-axis-aout} target. Arranges for
7736 indications in the program to the kernel loader that the stack of the
7737 program should be set to @var{n} bytes.
7743 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7744 @option{-march=v3} and @option{-march=v8} respectively.
7746 @item -mmul-bug-workaround
7747 @itemx -mno-mul-bug-workaround
7748 @opindex mmul-bug-workaround
7749 @opindex mno-mul-bug-workaround
7750 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7751 models where it applies. This option is active by default.
7755 Enable CRIS-specific verbose debug-related information in the assembly
7756 code. This option also has the effect to turn off the @samp{#NO_APP}
7757 formatted-code indicator to the assembler at the beginning of the
7762 Do not use condition-code results from previous instruction; always emit
7763 compare and test instructions before use of condition codes.
7765 @item -mno-side-effects
7766 @opindex mno-side-effects
7767 Do not emit instructions with side-effects in addressing modes other than
7771 @itemx -mno-stack-align
7773 @itemx -mno-data-align
7774 @itemx -mconst-align
7775 @itemx -mno-const-align
7776 @opindex mstack-align
7777 @opindex mno-stack-align
7778 @opindex mdata-align
7779 @opindex mno-data-align
7780 @opindex mconst-align
7781 @opindex mno-const-align
7782 These options (no-options) arranges (eliminate arrangements) for the
7783 stack-frame, individual data and constants to be aligned for the maximum
7784 single data access size for the chosen CPU model. The default is to
7785 arrange for 32-bit alignment. ABI details such as structure layout are
7786 not affected by these options.
7794 Similar to the stack- data- and const-align options above, these options
7795 arrange for stack-frame, writable data and constants to all be 32-bit,
7796 16-bit or 8-bit aligned. The default is 32-bit alignment.
7798 @item -mno-prologue-epilogue
7799 @itemx -mprologue-epilogue
7800 @opindex mno-prologue-epilogue
7801 @opindex mprologue-epilogue
7802 With @option{-mno-prologue-epilogue}, the normal function prologue and
7803 epilogue that sets up the stack-frame are omitted and no return
7804 instructions or return sequences are generated in the code. Use this
7805 option only together with visual inspection of the compiled code: no
7806 warnings or errors are generated when call-saved registers must be saved,
7807 or storage for local variable needs to be allocated.
7813 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7814 instruction sequences that load addresses for functions from the PLT part
7815 of the GOT rather than (traditional on other architectures) calls to the
7816 PLT@. The default is @option{-mgotplt}.
7820 Legacy no-op option only recognized with the cris-axis-aout target.
7824 Legacy no-op option only recognized with the cris-axis-elf and
7825 cris-axis-linux-gnu targets.
7829 Only recognized with the cris-axis-aout target, where it selects a
7830 GNU/linux-like multilib, include files and instruction set for
7835 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7839 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7840 to link with input-output functions from a simulator library. Code,
7841 initialized data and zero-initialized data are allocated consecutively.
7845 Like @option{-sim}, but pass linker options to locate initialized data at
7846 0x40000000 and zero-initialized data at 0x80000000.
7850 @subsection CRX Options
7853 These options are defined specifically for the CRX ports.
7859 Enable the use of multiply-accumulate instructions. Disabled by default.
7863 Push instructions will be used to pass outgoing arguments when functions
7864 are called. Enabled by default.
7867 @node Darwin Options
7868 @subsection Darwin Options
7869 @cindex Darwin options
7871 These options are defined for all architectures running the Darwin operating
7874 FSF GCC on Darwin does not create ``fat'' object files; it will create
7875 an object file for the single architecture that it was built to
7876 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7877 @option{-arch} options are used; it does so by running the compiler or
7878 linker multiple times and joining the results together with
7881 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7882 @samp{i686}) is determined by the flags that specify the ISA
7883 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7884 @option{-force_cpusubtype_ALL} option can be used to override this.
7886 The Darwin tools vary in their behavior when presented with an ISA
7887 mismatch. The assembler, @file{as}, will only permit instructions to
7888 be used that are valid for the subtype of the file it is generating,
7889 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7890 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7891 and print an error if asked to create a shared library with a less
7892 restrictive subtype than its input files (for instance, trying to put
7893 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7894 for executables, @file{ld}, will quietly give the executable the most
7895 restrictive subtype of any of its input files.
7900 Add the framework directory @var{dir} to the head of the list of
7901 directories to be searched for header files. These directories are
7902 interleaved with those specified by @option{-I} options and are
7903 scanned in a left-to-right order.
7905 A framework directory is a directory with frameworks in it. A
7906 framework is a directory with a @samp{"Headers"} and/or
7907 @samp{"PrivateHeaders"} directory contained directly in it that ends
7908 in @samp{".framework"}. The name of a framework is the name of this
7909 directory excluding the @samp{".framework"}. Headers associated with
7910 the framework are found in one of those two directories, with
7911 @samp{"Headers"} being searched first. A subframework is a framework
7912 directory that is in a framework's @samp{"Frameworks"} directory.
7913 Includes of subframework headers can only appear in a header of a
7914 framework that contains the subframework, or in a sibling subframework
7915 header. Two subframeworks are siblings if they occur in the same
7916 framework. A subframework should not have the same name as a
7917 framework, a warning will be issued if this is violated. Currently a
7918 subframework cannot have subframeworks, in the future, the mechanism
7919 may be extended to support this. The standard frameworks can be found
7920 in @samp{"/System/Library/Frameworks"} and
7921 @samp{"/Library/Frameworks"}. An example include looks like
7922 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7923 the name of the framework and header.h is found in the
7924 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7928 Emit debugging information for symbols that are used. For STABS
7929 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7930 This is by default ON@.
7934 Emit debugging information for all symbols and types.
7936 @item -mmacosx-version-min=@var{version}
7937 The earliest version of MacOS X that this executable will run on
7938 is @var{version}. Typical values of @var{version} include @code{10.1},
7939 @code{10.2}, and @code{10.3.9}.
7941 The default for this option is to make choices that seem to be most
7944 @item -mone-byte-bool
7945 @opindex -mone-byte-bool
7946 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7947 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7948 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7949 option has no effect on x86.
7951 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7952 to generate code that is not binary compatible with code generated
7953 without that switch. Using this switch may require recompiling all
7954 other modules in a program, including system libraries. Use this
7955 switch to conform to a non-default data model.
7957 @item -mfix-and-continue
7958 @itemx -ffix-and-continue
7959 @itemx -findirect-data
7960 @opindex mfix-and-continue
7961 @opindex ffix-and-continue
7962 @opindex findirect-data
7963 Generate code suitable for fast turn around development. Needed to
7964 enable gdb to dynamically load @code{.o} files into already running
7965 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7966 are provided for backwards compatibility.
7970 Loads all members of static archive libraries.
7971 See man ld(1) for more information.
7973 @item -arch_errors_fatal
7974 @opindex arch_errors_fatal
7975 Cause the errors having to do with files that have the wrong architecture
7979 @opindex bind_at_load
7980 Causes the output file to be marked such that the dynamic linker will
7981 bind all undefined references when the file is loaded or launched.
7985 Produce a Mach-o bundle format file.
7986 See man ld(1) for more information.
7988 @item -bundle_loader @var{executable}
7989 @opindex bundle_loader
7990 This option specifies the @var{executable} that will be loading the build
7991 output file being linked. See man ld(1) for more information.
7994 @opindex -dynamiclib
7995 When passed this option, GCC will produce a dynamic library instead of
7996 an executable when linking, using the Darwin @file{libtool} command.
7998 @item -force_cpusubtype_ALL
7999 @opindex -force_cpusubtype_ALL
8000 This causes GCC's output file to have the @var{ALL} subtype, instead of
8001 one controlled by the @option{-mcpu} or @option{-march} option.
8003 @item -allowable_client @var{client_name}
8005 @itemx -compatibility_version
8006 @itemx -current_version
8008 @itemx -dependency-file
8010 @itemx -dylinker_install_name
8012 @itemx -exported_symbols_list
8014 @itemx -flat_namespace
8015 @itemx -force_flat_namespace
8016 @itemx -headerpad_max_install_names
8019 @itemx -install_name
8020 @itemx -keep_private_externs
8021 @itemx -multi_module
8022 @itemx -multiply_defined
8023 @itemx -multiply_defined_unused
8025 @itemx -no_dead_strip_inits_and_terms
8026 @itemx -nofixprebinding
8029 @itemx -noseglinkedit
8030 @itemx -pagezero_size
8032 @itemx -prebind_all_twolevel_modules
8033 @itemx -private_bundle
8034 @itemx -read_only_relocs
8036 @itemx -sectobjectsymbols
8040 @itemx -sectobjectsymbols
8043 @itemx -segs_read_only_addr
8044 @itemx -segs_read_write_addr
8045 @itemx -seg_addr_table
8046 @itemx -seg_addr_table_filename
8049 @itemx -segs_read_only_addr
8050 @itemx -segs_read_write_addr
8051 @itemx -single_module
8054 @itemx -sub_umbrella
8055 @itemx -twolevel_namespace
8058 @itemx -unexported_symbols_list
8059 @itemx -weak_reference_mismatches
8062 @opindex allowable_client
8063 @opindex client_name
8064 @opindex compatibility_version
8065 @opindex current_version
8067 @opindex dependency-file
8069 @opindex dylinker_install_name
8071 @opindex exported_symbols_list
8073 @opindex flat_namespace
8074 @opindex force_flat_namespace
8075 @opindex headerpad_max_install_names
8078 @opindex install_name
8079 @opindex keep_private_externs
8080 @opindex multi_module
8081 @opindex multiply_defined
8082 @opindex multiply_defined_unused
8084 @opindex no_dead_strip_inits_and_terms
8085 @opindex nofixprebinding
8086 @opindex nomultidefs
8088 @opindex noseglinkedit
8089 @opindex pagezero_size
8091 @opindex prebind_all_twolevel_modules
8092 @opindex private_bundle
8093 @opindex read_only_relocs
8095 @opindex sectobjectsymbols
8099 @opindex sectobjectsymbols
8102 @opindex segs_read_only_addr
8103 @opindex segs_read_write_addr
8104 @opindex seg_addr_table
8105 @opindex seg_addr_table_filename
8106 @opindex seglinkedit
8108 @opindex segs_read_only_addr
8109 @opindex segs_read_write_addr
8110 @opindex single_module
8112 @opindex sub_library
8113 @opindex sub_umbrella
8114 @opindex twolevel_namespace
8117 @opindex unexported_symbols_list
8118 @opindex weak_reference_mismatches
8119 @opindex whatsloaded
8121 These options are passed to the Darwin linker. The Darwin linker man page
8122 describes them in detail.
8125 @node DEC Alpha Options
8126 @subsection DEC Alpha Options
8128 These @samp{-m} options are defined for the DEC Alpha implementations:
8131 @item -mno-soft-float
8133 @opindex mno-soft-float
8134 @opindex msoft-float
8135 Use (do not use) the hardware floating-point instructions for
8136 floating-point operations. When @option{-msoft-float} is specified,
8137 functions in @file{libgcc.a} will be used to perform floating-point
8138 operations. Unless they are replaced by routines that emulate the
8139 floating-point operations, or compiled in such a way as to call such
8140 emulations routines, these routines will issue floating-point
8141 operations. If you are compiling for an Alpha without floating-point
8142 operations, you must ensure that the library is built so as not to call
8145 Note that Alpha implementations without floating-point operations are
8146 required to have floating-point registers.
8151 @opindex mno-fp-regs
8152 Generate code that uses (does not use) the floating-point register set.
8153 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8154 register set is not used, floating point operands are passed in integer
8155 registers as if they were integers and floating-point results are passed
8156 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8157 so any function with a floating-point argument or return value called by code
8158 compiled with @option{-mno-fp-regs} must also be compiled with that
8161 A typical use of this option is building a kernel that does not use,
8162 and hence need not save and restore, any floating-point registers.
8166 The Alpha architecture implements floating-point hardware optimized for
8167 maximum performance. It is mostly compliant with the IEEE floating
8168 point standard. However, for full compliance, software assistance is
8169 required. This option generates code fully IEEE compliant code
8170 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8171 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8172 defined during compilation. The resulting code is less efficient but is
8173 able to correctly support denormalized numbers and exceptional IEEE
8174 values such as not-a-number and plus/minus infinity. Other Alpha
8175 compilers call this option @option{-ieee_with_no_inexact}.
8177 @item -mieee-with-inexact
8178 @opindex mieee-with-inexact
8179 This is like @option{-mieee} except the generated code also maintains
8180 the IEEE @var{inexact-flag}. Turning on this option causes the
8181 generated code to implement fully-compliant IEEE math. In addition to
8182 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8183 macro. On some Alpha implementations the resulting code may execute
8184 significantly slower than the code generated by default. Since there is
8185 very little code that depends on the @var{inexact-flag}, you should
8186 normally not specify this option. Other Alpha compilers call this
8187 option @option{-ieee_with_inexact}.
8189 @item -mfp-trap-mode=@var{trap-mode}
8190 @opindex mfp-trap-mode
8191 This option controls what floating-point related traps are enabled.
8192 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8193 The trap mode can be set to one of four values:
8197 This is the default (normal) setting. The only traps that are enabled
8198 are the ones that cannot be disabled in software (e.g., division by zero
8202 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8206 Like @samp{su}, but the instructions are marked to be safe for software
8207 completion (see Alpha architecture manual for details).
8210 Like @samp{su}, but inexact traps are enabled as well.
8213 @item -mfp-rounding-mode=@var{rounding-mode}
8214 @opindex mfp-rounding-mode
8215 Selects the IEEE rounding mode. Other Alpha compilers call this option
8216 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8221 Normal IEEE rounding mode. Floating point numbers are rounded towards
8222 the nearest machine number or towards the even machine number in case
8226 Round towards minus infinity.
8229 Chopped rounding mode. Floating point numbers are rounded towards zero.
8232 Dynamic rounding mode. A field in the floating point control register
8233 (@var{fpcr}, see Alpha architecture reference manual) controls the
8234 rounding mode in effect. The C library initializes this register for
8235 rounding towards plus infinity. Thus, unless your program modifies the
8236 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8239 @item -mtrap-precision=@var{trap-precision}
8240 @opindex mtrap-precision
8241 In the Alpha architecture, floating point traps are imprecise. This
8242 means without software assistance it is impossible to recover from a
8243 floating trap and program execution normally needs to be terminated.
8244 GCC can generate code that can assist operating system trap handlers
8245 in determining the exact location that caused a floating point trap.
8246 Depending on the requirements of an application, different levels of
8247 precisions can be selected:
8251 Program precision. This option is the default and means a trap handler
8252 can only identify which program caused a floating point exception.
8255 Function precision. The trap handler can determine the function that
8256 caused a floating point exception.
8259 Instruction precision. The trap handler can determine the exact
8260 instruction that caused a floating point exception.
8263 Other Alpha compilers provide the equivalent options called
8264 @option{-scope_safe} and @option{-resumption_safe}.
8266 @item -mieee-conformant
8267 @opindex mieee-conformant
8268 This option marks the generated code as IEEE conformant. You must not
8269 use this option unless you also specify @option{-mtrap-precision=i} and either
8270 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8271 is to emit the line @samp{.eflag 48} in the function prologue of the
8272 generated assembly file. Under DEC Unix, this has the effect that
8273 IEEE-conformant math library routines will be linked in.
8275 @item -mbuild-constants
8276 @opindex mbuild-constants
8277 Normally GCC examines a 32- or 64-bit integer constant to
8278 see if it can construct it from smaller constants in two or three
8279 instructions. If it cannot, it will output the constant as a literal and
8280 generate code to load it from the data segment at runtime.
8282 Use this option to require GCC to construct @emph{all} integer constants
8283 using code, even if it takes more instructions (the maximum is six).
8285 You would typically use this option to build a shared library dynamic
8286 loader. Itself a shared library, it must relocate itself in memory
8287 before it can find the variables and constants in its own data segment.
8293 Select whether to generate code to be assembled by the vendor-supplied
8294 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8312 Indicate whether GCC should generate code to use the optional BWX,
8313 CIX, FIX and MAX instruction sets. The default is to use the instruction
8314 sets supported by the CPU type specified via @option{-mcpu=} option or that
8315 of the CPU on which GCC was built if none was specified.
8320 @opindex mfloat-ieee
8321 Generate code that uses (does not use) VAX F and G floating point
8322 arithmetic instead of IEEE single and double precision.
8324 @item -mexplicit-relocs
8325 @itemx -mno-explicit-relocs
8326 @opindex mexplicit-relocs
8327 @opindex mno-explicit-relocs
8328 Older Alpha assemblers provided no way to generate symbol relocations
8329 except via assembler macros. Use of these macros does not allow
8330 optimal instruction scheduling. GNU binutils as of version 2.12
8331 supports a new syntax that allows the compiler to explicitly mark
8332 which relocations should apply to which instructions. This option
8333 is mostly useful for debugging, as GCC detects the capabilities of
8334 the assembler when it is built and sets the default accordingly.
8338 @opindex msmall-data
8339 @opindex mlarge-data
8340 When @option{-mexplicit-relocs} is in effect, static data is
8341 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8342 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8343 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8344 16-bit relocations off of the @code{$gp} register. This limits the
8345 size of the small data area to 64KB, but allows the variables to be
8346 directly accessed via a single instruction.
8348 The default is @option{-mlarge-data}. With this option the data area
8349 is limited to just below 2GB@. Programs that require more than 2GB of
8350 data must use @code{malloc} or @code{mmap} to allocate the data in the
8351 heap instead of in the program's data segment.
8353 When generating code for shared libraries, @option{-fpic} implies
8354 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8358 @opindex msmall-text
8359 @opindex mlarge-text
8360 When @option{-msmall-text} is used, the compiler assumes that the
8361 code of the entire program (or shared library) fits in 4MB, and is
8362 thus reachable with a branch instruction. When @option{-msmall-data}
8363 is used, the compiler can assume that all local symbols share the
8364 same @code{$gp} value, and thus reduce the number of instructions
8365 required for a function call from 4 to 1.
8367 The default is @option{-mlarge-text}.
8369 @item -mcpu=@var{cpu_type}
8371 Set the instruction set and instruction scheduling parameters for
8372 machine type @var{cpu_type}. You can specify either the @samp{EV}
8373 style name or the corresponding chip number. GCC supports scheduling
8374 parameters for the EV4, EV5 and EV6 family of processors and will
8375 choose the default values for the instruction set from the processor
8376 you specify. If you do not specify a processor type, GCC will default
8377 to the processor on which the compiler was built.
8379 Supported values for @var{cpu_type} are
8385 Schedules as an EV4 and has no instruction set extensions.
8389 Schedules as an EV5 and has no instruction set extensions.
8393 Schedules as an EV5 and supports the BWX extension.
8398 Schedules as an EV5 and supports the BWX and MAX extensions.
8402 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8406 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8409 @item -mtune=@var{cpu_type}
8411 Set only the instruction scheduling parameters for machine type
8412 @var{cpu_type}. The instruction set is not changed.
8414 @item -mmemory-latency=@var{time}
8415 @opindex mmemory-latency
8416 Sets the latency the scheduler should assume for typical memory
8417 references as seen by the application. This number is highly
8418 dependent on the memory access patterns used by the application
8419 and the size of the external cache on the machine.
8421 Valid options for @var{time} are
8425 A decimal number representing clock cycles.
8431 The compiler contains estimates of the number of clock cycles for
8432 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8433 (also called Dcache, Scache, and Bcache), as well as to main memory.
8434 Note that L3 is only valid for EV5.
8439 @node DEC Alpha/VMS Options
8440 @subsection DEC Alpha/VMS Options
8442 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8445 @item -mvms-return-codes
8446 @opindex mvms-return-codes
8447 Return VMS condition codes from main. The default is to return POSIX
8448 style condition (e.g.@ error) codes.
8452 @subsection FRV Options
8459 Only use the first 32 general purpose registers.
8464 Use all 64 general purpose registers.
8469 Use only the first 32 floating point registers.
8474 Use all 64 floating point registers
8477 @opindex mhard-float
8479 Use hardware instructions for floating point operations.
8482 @opindex msoft-float
8484 Use library routines for floating point operations.
8489 Dynamically allocate condition code registers.
8494 Do not try to dynamically allocate condition code registers, only
8495 use @code{icc0} and @code{fcc0}.
8500 Change ABI to use double word insns.
8505 Do not use double word instructions.
8510 Use floating point double instructions.
8515 Do not use floating point double instructions.
8520 Use media instructions.
8525 Do not use media instructions.
8530 Use multiply and add/subtract instructions.
8535 Do not use multiply and add/subtract instructions.
8540 Select the FDPIC ABI, that uses function descriptors to represent
8541 pointers to functions. Without any PIC/PIE-related options, it
8542 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8543 assumes GOT entries and small data are within a 12-bit range from the
8544 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8545 are computed with 32 bits.
8548 @opindex minline-plt
8550 Enable inlining of PLT entries in function calls to functions that are
8551 not known to bind locally. It has no effect without @option{-mfdpic}.
8552 It's enabled by default if optimizing for speed and compiling for
8553 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8554 optimization option such as @option{-O3} or above is present in the
8560 Assume a large TLS segment when generating thread-local code.
8565 Do not assume a large TLS segment when generating thread-local code.
8570 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8571 that is known to be in read-only sections. It's enabled by default,
8572 except for @option{-fpic} or @option{-fpie}: even though it may help
8573 make the global offset table smaller, it trades 1 instruction for 4.
8574 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8575 one of which may be shared by multiple symbols, and it avoids the need
8576 for a GOT entry for the referenced symbol, so it's more likely to be a
8577 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8579 @item -multilib-library-pic
8580 @opindex multilib-library-pic
8582 Link with the (library, not FD) pic libraries. It's implied by
8583 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8584 @option{-fpic} without @option{-mfdpic}. You should never have to use
8590 Follow the EABI requirement of always creating a frame pointer whenever
8591 a stack frame is allocated. This option is enabled by default and can
8592 be disabled with @option{-mno-linked-fp}.
8595 @opindex mlong-calls
8597 Use indirect addressing to call functions outside the current
8598 compilation unit. This allows the functions to be placed anywhere
8599 within the 32-bit address space.
8601 @item -malign-labels
8602 @opindex malign-labels
8604 Try to align labels to an 8-byte boundary by inserting nops into the
8605 previous packet. This option only has an effect when VLIW packing
8606 is enabled. It doesn't create new packets; it merely adds nops to
8610 @opindex mlibrary-pic
8612 Generate position-independent EABI code.
8617 Use only the first four media accumulator registers.
8622 Use all eight media accumulator registers.
8627 Pack VLIW instructions.
8632 Do not pack VLIW instructions.
8637 Do not mark ABI switches in e_flags.
8642 Enable the use of conditional-move instructions (default).
8644 This switch is mainly for debugging the compiler and will likely be removed
8645 in a future version.
8647 @item -mno-cond-move
8648 @opindex mno-cond-move
8650 Disable the use of conditional-move instructions.
8652 This switch is mainly for debugging the compiler and will likely be removed
8653 in a future version.
8658 Enable the use of conditional set instructions (default).
8660 This switch is mainly for debugging the compiler and will likely be removed
8661 in a future version.
8666 Disable the use of conditional set instructions.
8668 This switch is mainly for debugging the compiler and will likely be removed
8669 in a future version.
8674 Enable the use of conditional execution (default).
8676 This switch is mainly for debugging the compiler and will likely be removed
8677 in a future version.
8679 @item -mno-cond-exec
8680 @opindex mno-cond-exec
8682 Disable the use of conditional execution.
8684 This switch is mainly for debugging the compiler and will likely be removed
8685 in a future version.
8688 @opindex mvliw-branch
8690 Run a pass to pack branches into VLIW instructions (default).
8692 This switch is mainly for debugging the compiler and will likely be removed
8693 in a future version.
8695 @item -mno-vliw-branch
8696 @opindex mno-vliw-branch
8698 Do not run a pass to pack branches into VLIW instructions.
8700 This switch is mainly for debugging the compiler and will likely be removed
8701 in a future version.
8703 @item -mmulti-cond-exec
8704 @opindex mmulti-cond-exec
8706 Enable optimization of @code{&&} and @code{||} in conditional execution
8709 This switch is mainly for debugging the compiler and will likely be removed
8710 in a future version.
8712 @item -mno-multi-cond-exec
8713 @opindex mno-multi-cond-exec
8715 Disable optimization of @code{&&} and @code{||} in conditional execution.
8717 This switch is mainly for debugging the compiler and will likely be removed
8718 in a future version.
8720 @item -mnested-cond-exec
8721 @opindex mnested-cond-exec
8723 Enable nested conditional execution optimizations (default).
8725 This switch is mainly for debugging the compiler and will likely be removed
8726 in a future version.
8728 @item -mno-nested-cond-exec
8729 @opindex mno-nested-cond-exec
8731 Disable nested conditional execution optimizations.
8733 This switch is mainly for debugging the compiler and will likely be removed
8734 in a future version.
8736 @item -moptimize-membar
8737 @opindex moptimize-membar
8739 This switch removes redundant @code{membar} instructions from the
8740 compiler generated code. It is enabled by default.
8742 @item -mno-optimize-membar
8743 @opindex mno-optimize-membar
8745 This switch disables the automatic removal of redundant @code{membar}
8746 instructions from the generated code.
8748 @item -mtomcat-stats
8749 @opindex mtomcat-stats
8751 Cause gas to print out tomcat statistics.
8753 @item -mcpu=@var{cpu}
8756 Select the processor type for which to generate code. Possible values are
8757 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8758 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8762 @node H8/300 Options
8763 @subsection H8/300 Options
8765 These @samp{-m} options are defined for the H8/300 implementations:
8770 Shorten some address references at link time, when possible; uses the
8771 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8772 ld, Using ld}, for a fuller description.
8776 Generate code for the H8/300H@.
8780 Generate code for the H8S@.
8784 Generate code for the H8S and H8/300H in the normal mode. This switch
8785 must be used either with @option{-mh} or @option{-ms}.
8789 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8793 Make @code{int} data 32 bits by default.
8797 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8798 The default for the H8/300H and H8S is to align longs and floats on 4
8800 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8801 This option has no effect on the H8/300.
8805 @subsection HPPA Options
8806 @cindex HPPA Options
8808 These @samp{-m} options are defined for the HPPA family of computers:
8811 @item -march=@var{architecture-type}
8813 Generate code for the specified architecture. The choices for
8814 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8815 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8816 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8817 architecture option for your machine. Code compiled for lower numbered
8818 architectures will run on higher numbered architectures, but not the
8822 @itemx -mpa-risc-1-1
8823 @itemx -mpa-risc-2-0
8824 @opindex mpa-risc-1-0
8825 @opindex mpa-risc-1-1
8826 @opindex mpa-risc-2-0
8827 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8830 @opindex mbig-switch
8831 Generate code suitable for big switch tables. Use this option only if
8832 the assembler/linker complain about out of range branches within a switch
8835 @item -mjump-in-delay
8836 @opindex mjump-in-delay
8837 Fill delay slots of function calls with unconditional jump instructions
8838 by modifying the return pointer for the function call to be the target
8839 of the conditional jump.
8841 @item -mdisable-fpregs
8842 @opindex mdisable-fpregs
8843 Prevent floating point registers from being used in any manner. This is
8844 necessary for compiling kernels which perform lazy context switching of
8845 floating point registers. If you use this option and attempt to perform
8846 floating point operations, the compiler will abort.
8848 @item -mdisable-indexing
8849 @opindex mdisable-indexing
8850 Prevent the compiler from using indexing address modes. This avoids some
8851 rather obscure problems when compiling MIG generated code under MACH@.
8853 @item -mno-space-regs
8854 @opindex mno-space-regs
8855 Generate code that assumes the target has no space registers. This allows
8856 GCC to generate faster indirect calls and use unscaled index address modes.
8858 Such code is suitable for level 0 PA systems and kernels.
8860 @item -mfast-indirect-calls
8861 @opindex mfast-indirect-calls
8862 Generate code that assumes calls never cross space boundaries. This
8863 allows GCC to emit code which performs faster indirect calls.
8865 This option will not work in the presence of shared libraries or nested
8868 @item -mfixed-range=@var{register-range}
8869 @opindex mfixed-range
8870 Generate code treating the given register range as fixed registers.
8871 A fixed register is one that the register allocator can not use. This is
8872 useful when compiling kernel code. A register range is specified as
8873 two registers separated by a dash. Multiple register ranges can be
8874 specified separated by a comma.
8876 @item -mlong-load-store
8877 @opindex mlong-load-store
8878 Generate 3-instruction load and store sequences as sometimes required by
8879 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8882 @item -mportable-runtime
8883 @opindex mportable-runtime
8884 Use the portable calling conventions proposed by HP for ELF systems.
8888 Enable the use of assembler directives only GAS understands.
8890 @item -mschedule=@var{cpu-type}
8892 Schedule code according to the constraints for the machine type
8893 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8894 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8895 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8896 proper scheduling option for your machine. The default scheduling is
8900 @opindex mlinker-opt
8901 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8902 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8903 linkers in which they give bogus error messages when linking some programs.
8906 @opindex msoft-float
8907 Generate output containing library calls for floating point.
8908 @strong{Warning:} the requisite libraries are not available for all HPPA
8909 targets. Normally the facilities of the machine's usual C compiler are
8910 used, but this cannot be done directly in cross-compilation. You must make
8911 your own arrangements to provide suitable library functions for
8912 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8913 does provide software floating point support.
8915 @option{-msoft-float} changes the calling convention in the output file;
8916 therefore, it is only useful if you compile @emph{all} of a program with
8917 this option. In particular, you need to compile @file{libgcc.a}, the
8918 library that comes with GCC, with @option{-msoft-float} in order for
8923 Generate the predefine, @code{_SIO}, for server IO@. The default is
8924 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8925 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8926 options are available under HP-UX and HI-UX@.
8930 Use GNU ld specific options. This passes @option{-shared} to ld when
8931 building a shared library. It is the default when GCC is configured,
8932 explicitly or implicitly, with the GNU linker. This option does not
8933 have any affect on which ld is called, it only changes what parameters
8934 are passed to that ld. The ld that is called is determined by the
8935 @option{--with-ld} configure option, GCC's program search path, and
8936 finally by the user's @env{PATH}. The linker used by GCC can be printed
8937 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
8938 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8942 Use HP ld specific options. This passes @option{-b} to ld when building
8943 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8944 links. It is the default when GCC is configured, explicitly or
8945 implicitly, with the HP linker. This option does not have any affect on
8946 which ld is called, it only changes what parameters are passed to that
8947 ld. The ld that is called is determined by the @option{--with-ld}
8948 configure option, GCC's program search path, and finally by the user's
8949 @env{PATH}. The linker used by GCC can be printed using @samp{which
8950 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
8951 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8954 @opindex mno-long-calls
8955 Generate code that uses long call sequences. This ensures that a call
8956 is always able to reach linker generated stubs. The default is to generate
8957 long calls only when the distance from the call site to the beginning
8958 of the function or translation unit, as the case may be, exceeds a
8959 predefined limit set by the branch type being used. The limits for
8960 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8961 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8964 Distances are measured from the beginning of functions when using the
8965 @option{-ffunction-sections} option, or when using the @option{-mgas}
8966 and @option{-mno-portable-runtime} options together under HP-UX with
8969 It is normally not desirable to use this option as it will degrade
8970 performance. However, it may be useful in large applications,
8971 particularly when partial linking is used to build the application.
8973 The types of long calls used depends on the capabilities of the
8974 assembler and linker, and the type of code being generated. The
8975 impact on systems that support long absolute calls, and long pic
8976 symbol-difference or pc-relative calls should be relatively small.
8977 However, an indirect call is used on 32-bit ELF systems in pic code
8978 and it is quite long.
8980 @item -munix=@var{unix-std}
8982 Generate compiler predefines and select a startfile for the specified
8983 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8984 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8985 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8986 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8987 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8990 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8991 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8992 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8993 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8994 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8995 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8997 It is @emph{important} to note that this option changes the interfaces
8998 for various library routines. It also affects the operational behavior
8999 of the C library. Thus, @emph{extreme} care is needed in using this
9002 Library code that is intended to operate with more than one UNIX
9003 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9004 as appropriate. Most GNU software doesn't provide this capability.
9008 Suppress the generation of link options to search libdld.sl when the
9009 @option{-static} option is specified on HP-UX 10 and later.
9013 The HP-UX implementation of setlocale in libc has a dependency on
9014 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9015 when the @option{-static} option is specified, special link options
9016 are needed to resolve this dependency.
9018 On HP-UX 10 and later, the GCC driver adds the necessary options to
9019 link with libdld.sl when the @option{-static} option is specified.
9020 This causes the resulting binary to be dynamic. On the 64-bit port,
9021 the linkers generate dynamic binaries by default in any case. The
9022 @option{-nolibdld} option can be used to prevent the GCC driver from
9023 adding these link options.
9027 Add support for multithreading with the @dfn{dce thread} library
9028 under HP-UX@. This option sets flags for both the preprocessor and
9032 @node i386 and x86-64 Options
9033 @subsection Intel 386 and AMD x86-64 Options
9034 @cindex i386 Options
9035 @cindex x86-64 Options
9036 @cindex Intel 386 Options
9037 @cindex AMD x86-64 Options
9039 These @samp{-m} options are defined for the i386 and x86-64 family of
9043 @item -mtune=@var{cpu-type}
9045 Tune to @var{cpu-type} everything applicable about the generated code, except
9046 for the ABI and the set of available instructions. The choices for
9050 Original Intel's i386 CPU@.
9052 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9054 Intel Pentium CPU with no MMX support.
9056 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9057 @item i686, pentiumpro
9058 Intel PentiumPro CPU@.
9060 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9061 @item pentium3, pentium3m
9062 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9065 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9066 support. Used by Centrino notebooks.
9067 @item pentium4, pentium4m
9068 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9070 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9073 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9074 SSE2 and SSE3 instruction set support.
9076 AMD K6 CPU with MMX instruction set support.
9078 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9079 @item athlon, athlon-tbird
9080 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9082 @item athlon-4, athlon-xp, athlon-mp
9083 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9084 instruction set support.
9085 @item k8, opteron, athlon64, athlon-fx
9086 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9087 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9089 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9092 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9093 instruction set support.
9095 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9096 implemented for this chip.)
9098 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9099 implemented for this chip.)
9102 While picking a specific @var{cpu-type} will schedule things appropriately
9103 for that particular chip, the compiler will not generate any code that
9104 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9107 @item -march=@var{cpu-type}
9109 Generate instructions for the machine type @var{cpu-type}. The choices
9110 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9111 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9113 @item -mcpu=@var{cpu-type}
9115 A deprecated synonym for @option{-mtune}.
9124 @opindex mpentiumpro
9125 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9126 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9127 These synonyms are deprecated.
9129 @item -mfpmath=@var{unit}
9131 Generate floating point arithmetics for selected unit @var{unit}. The choices
9136 Use the standard 387 floating point coprocessor present majority of chips and
9137 emulated otherwise. Code compiled with this option will run almost everywhere.
9138 The temporary results are computed in 80bit precision instead of precision
9139 specified by the type resulting in slightly different results compared to most
9140 of other chips. See @option{-ffloat-store} for more detailed description.
9142 This is the default choice for i386 compiler.
9145 Use scalar floating point instructions present in the SSE instruction set.
9146 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9147 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9148 instruction set supports only single precision arithmetics, thus the double and
9149 extended precision arithmetics is still done using 387. Later version, present
9150 only in Pentium4 and the future AMD x86-64 chips supports double precision
9153 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9154 or @option{-msse2} switches to enable SSE extensions and make this option
9155 effective. For the x86-64 compiler, these extensions are enabled by default.
9157 The resulting code should be considerably faster in the majority of cases and avoid
9158 the numerical instability problems of 387 code, but may break some existing
9159 code that expects temporaries to be 80bit.
9161 This is the default choice for the x86-64 compiler.
9164 Attempt to utilize both instruction sets at once. This effectively double the
9165 amount of available registers and on chips with separate execution units for
9166 387 and SSE the execution resources too. Use this option with care, as it is
9167 still experimental, because the GCC register allocator does not model separate
9168 functional units well resulting in instable performance.
9171 @item -masm=@var{dialect}
9172 @opindex masm=@var{dialect}
9173 Output asm instructions using selected @var{dialect}. Supported
9174 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9175 not support @samp{intel}.
9180 @opindex mno-ieee-fp
9181 Control whether or not the compiler uses IEEE floating point
9182 comparisons. These handle correctly the case where the result of a
9183 comparison is unordered.
9186 @opindex msoft-float
9187 Generate output containing library calls for floating point.
9188 @strong{Warning:} the requisite libraries are not part of GCC@.
9189 Normally the facilities of the machine's usual C compiler are used, but
9190 this can't be done directly in cross-compilation. You must make your
9191 own arrangements to provide suitable library functions for
9194 On machines where a function returns floating point results in the 80387
9195 register stack, some floating point opcodes may be emitted even if
9196 @option{-msoft-float} is used.
9198 @item -mno-fp-ret-in-387
9199 @opindex mno-fp-ret-in-387
9200 Do not use the FPU registers for return values of functions.
9202 The usual calling convention has functions return values of types
9203 @code{float} and @code{double} in an FPU register, even if there
9204 is no FPU@. The idea is that the operating system should emulate
9207 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9208 in ordinary CPU registers instead.
9210 @item -mno-fancy-math-387
9211 @opindex mno-fancy-math-387
9212 Some 387 emulators do not support the @code{sin}, @code{cos} and
9213 @code{sqrt} instructions for the 387. Specify this option to avoid
9214 generating those instructions. This option is the default on FreeBSD,
9215 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9216 indicates that the target cpu will always have an FPU and so the
9217 instruction will not need emulation. As of revision 2.6.1, these
9218 instructions are not generated unless you also use the
9219 @option{-funsafe-math-optimizations} switch.
9221 @item -malign-double
9222 @itemx -mno-align-double
9223 @opindex malign-double
9224 @opindex mno-align-double
9225 Control whether GCC aligns @code{double}, @code{long double}, and
9226 @code{long long} variables on a two word boundary or a one word
9227 boundary. Aligning @code{double} variables on a two word boundary will
9228 produce code that runs somewhat faster on a @samp{Pentium} at the
9229 expense of more memory.
9231 @strong{Warning:} if you use the @option{-malign-double} switch,
9232 structures containing the above types will be aligned differently than
9233 the published application binary interface specifications for the 386
9234 and will not be binary compatible with structures in code compiled
9235 without that switch.
9237 @item -m96bit-long-double
9238 @itemx -m128bit-long-double
9239 @opindex m96bit-long-double
9240 @opindex m128bit-long-double
9241 These switches control the size of @code{long double} type. The i386
9242 application binary interface specifies the size to be 96 bits,
9243 so @option{-m96bit-long-double} is the default in 32 bit mode.
9245 Modern architectures (Pentium and newer) would prefer @code{long double}
9246 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9247 conforming to the ABI, this would not be possible. So specifying a
9248 @option{-m128bit-long-double} will align @code{long double}
9249 to a 16 byte boundary by padding the @code{long double} with an additional
9252 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9253 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9255 Notice that neither of these options enable any extra precision over the x87
9256 standard of 80 bits for a @code{long double}.
9258 @strong{Warning:} if you override the default value for your target ABI, the
9259 structures and arrays containing @code{long double} variables will change
9260 their size as well as function calling convention for function taking
9261 @code{long double} will be modified. Hence they will not be binary
9262 compatible with arrays or structures in code compiled without that switch.
9264 @item -mmlarge-data-threshold=@var{number}
9265 @opindex mlarge-data-threshold=@var{number}
9266 When @option{-mcmodel=medium} is specified, the data greater than
9267 @var{threshold} are placed in large data section. This value must be the
9268 same across all object linked into the binary and defaults to 65535.
9271 @itemx -mno-svr3-shlib
9272 @opindex msvr3-shlib
9273 @opindex mno-svr3-shlib
9274 Control whether GCC places uninitialized local variables into the
9275 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9276 into @code{bss}. These options are meaningful only on System V Release 3.
9280 Use a different function-calling convention, in which functions that
9281 take a fixed number of arguments return with the @code{ret} @var{num}
9282 instruction, which pops their arguments while returning. This saves one
9283 instruction in the caller since there is no need to pop the arguments
9286 You can specify that an individual function is called with this calling
9287 sequence with the function attribute @samp{stdcall}. You can also
9288 override the @option{-mrtd} option by using the function attribute
9289 @samp{cdecl}. @xref{Function Attributes}.
9291 @strong{Warning:} this calling convention is incompatible with the one
9292 normally used on Unix, so you cannot use it if you need to call
9293 libraries compiled with the Unix compiler.
9295 Also, you must provide function prototypes for all functions that
9296 take variable numbers of arguments (including @code{printf});
9297 otherwise incorrect code will be generated for calls to those
9300 In addition, seriously incorrect code will result if you call a
9301 function with too many arguments. (Normally, extra arguments are
9302 harmlessly ignored.)
9304 @item -mregparm=@var{num}
9306 Control how many registers are used to pass integer arguments. By
9307 default, no registers are used to pass arguments, and at most 3
9308 registers can be used. You can control this behavior for a specific
9309 function by using the function attribute @samp{regparm}.
9310 @xref{Function Attributes}.
9312 @strong{Warning:} if you use this switch, and
9313 @var{num} is nonzero, then you must build all modules with the same
9314 value, including any libraries. This includes the system libraries and
9318 @opindex msseregparm
9319 Use SSE register passing conventions for float and double arguments
9320 and return values. You can control this behavior for a specific
9321 function by using the function attribute @samp{sseregparm}.
9322 @xref{Function Attributes}.
9324 @strong{Warning:} if you use this switch then you must build all
9325 modules with the same value, including any libraries. This includes
9326 the system libraries and startup modules.
9328 @item -mpreferred-stack-boundary=@var{num}
9329 @opindex mpreferred-stack-boundary
9330 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9331 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9332 the default is 4 (16 bytes or 128 bits), except when optimizing for code
9333 size (@option{-Os}), in which case the default is the minimum correct
9334 alignment (4 bytes for x86, and 8 bytes for x86-64).
9336 On Pentium and PentiumPro, @code{double} and @code{long double} values
9337 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9338 suffer significant run time performance penalties. On Pentium III, the
9339 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
9340 penalties if it is not 16 byte aligned.
9342 To ensure proper alignment of this values on the stack, the stack boundary
9343 must be as aligned as that required by any value stored on the stack.
9344 Further, every function must be generated such that it keeps the stack
9345 aligned. Thus calling a function compiled with a higher preferred
9346 stack boundary from a function compiled with a lower preferred stack
9347 boundary will most likely misalign the stack. It is recommended that
9348 libraries that use callbacks always use the default setting.
9350 This extra alignment does consume extra stack space, and generally
9351 increases code size. Code that is sensitive to stack space usage, such
9352 as embedded systems and operating system kernels, may want to reduce the
9353 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9371 These switches enable or disable the use of instructions in the MMX,
9372 SSE, SSE2 or 3DNow! extended instruction sets. These extensions are
9373 also available as built-in functions: see @ref{X86 Built-in Functions},
9374 for details of the functions enabled and disabled by these switches.
9376 To have SSE/SSE2 instructions generated automatically from floating-point
9377 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9379 These options will enable GCC to use these extended instructions in
9380 generated code, even without @option{-mfpmath=sse}. Applications which
9381 perform runtime CPU detection must compile separate files for each
9382 supported architecture, using the appropriate flags. In particular,
9383 the file containing the CPU detection code should be compiled without
9387 @itemx -mno-push-args
9389 @opindex mno-push-args
9390 Use PUSH operations to store outgoing parameters. This method is shorter
9391 and usually equally fast as method using SUB/MOV operations and is enabled
9392 by default. In some cases disabling it may improve performance because of
9393 improved scheduling and reduced dependencies.
9395 @item -maccumulate-outgoing-args
9396 @opindex maccumulate-outgoing-args
9397 If enabled, the maximum amount of space required for outgoing arguments will be
9398 computed in the function prologue. This is faster on most modern CPUs
9399 because of reduced dependencies, improved scheduling and reduced stack usage
9400 when preferred stack boundary is not equal to 2. The drawback is a notable
9401 increase in code size. This switch implies @option{-mno-push-args}.
9405 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9406 on thread-safe exception handling must compile and link all code with the
9407 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9408 @option{-D_MT}; when linking, it links in a special thread helper library
9409 @option{-lmingwthrd} which cleans up per thread exception handling data.
9411 @item -mno-align-stringops
9412 @opindex mno-align-stringops
9413 Do not align destination of inlined string operations. This switch reduces
9414 code size and improves performance in case the destination is already aligned,
9415 but GCC doesn't know about it.
9417 @item -minline-all-stringops
9418 @opindex minline-all-stringops
9419 By default GCC inlines string operations only when destination is known to be
9420 aligned at least to 4 byte boundary. This enables more inlining, increase code
9421 size, but may improve performance of code that depends on fast memcpy, strlen
9422 and memset for short lengths.
9424 @item -momit-leaf-frame-pointer
9425 @opindex momit-leaf-frame-pointer
9426 Don't keep the frame pointer in a register for leaf functions. This
9427 avoids the instructions to save, set up and restore frame pointers and
9428 makes an extra register available in leaf functions. The option
9429 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9430 which might make debugging harder.
9432 @item -mtls-direct-seg-refs
9433 @itemx -mno-tls-direct-seg-refs
9434 @opindex mtls-direct-seg-refs
9435 Controls whether TLS variables may be accessed with offsets from the
9436 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9437 or whether the thread base pointer must be added. Whether or not this
9438 is legal depends on the operating system, and whether it maps the
9439 segment to cover the entire TLS area.
9441 For systems that use GNU libc, the default is on.
9444 These @samp{-m} switches are supported in addition to the above
9445 on AMD x86-64 processors in 64-bit environments.
9452 Generate code for a 32-bit or 64-bit environment.
9453 The 32-bit environment sets int, long and pointer to 32 bits and
9454 generates code that runs on any i386 system.
9455 The 64-bit environment sets int to 32 bits and long and pointer
9456 to 64 bits and generates code for AMD's x86-64 architecture.
9459 @opindex no-red-zone
9460 Do not use a so called red zone for x86-64 code. The red zone is mandated
9461 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9462 stack pointer that will not be modified by signal or interrupt handlers
9463 and therefore can be used for temporary data without adjusting the stack
9464 pointer. The flag @option{-mno-red-zone} disables this red zone.
9466 @item -mcmodel=small
9467 @opindex mcmodel=small
9468 Generate code for the small code model: the program and its symbols must
9469 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9470 Programs can be statically or dynamically linked. This is the default
9473 @item -mcmodel=kernel
9474 @opindex mcmodel=kernel
9475 Generate code for the kernel code model. The kernel runs in the
9476 negative 2 GB of the address space.
9477 This model has to be used for Linux kernel code.
9479 @item -mcmodel=medium
9480 @opindex mcmodel=medium
9481 Generate code for the medium model: The program is linked in the lower 2
9482 GB of the address space but symbols can be located anywhere in the
9483 address space. Programs can be statically or dynamically linked, but
9484 building of shared libraries are not supported with the medium model.
9486 @item -mcmodel=large
9487 @opindex mcmodel=large
9488 Generate code for the large model: This model makes no assumptions
9489 about addresses and sizes of sections. Currently GCC does not implement
9494 @subsection IA-64 Options
9495 @cindex IA-64 Options
9497 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9501 @opindex mbig-endian
9502 Generate code for a big endian target. This is the default for HP-UX@.
9504 @item -mlittle-endian
9505 @opindex mlittle-endian
9506 Generate code for a little endian target. This is the default for AIX5
9513 Generate (or don't) code for the GNU assembler. This is the default.
9514 @c Also, this is the default if the configure option @option{--with-gnu-as}
9521 Generate (or don't) code for the GNU linker. This is the default.
9522 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9527 Generate code that does not use a global pointer register. The result
9528 is not position independent code, and violates the IA-64 ABI@.
9530 @item -mvolatile-asm-stop
9531 @itemx -mno-volatile-asm-stop
9532 @opindex mvolatile-asm-stop
9533 @opindex mno-volatile-asm-stop
9534 Generate (or don't) a stop bit immediately before and after volatile asm
9537 @item -mregister-names
9538 @itemx -mno-register-names
9539 @opindex mregister-names
9540 @opindex mno-register-names
9541 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9542 the stacked registers. This may make assembler output more readable.
9548 Disable (or enable) optimizations that use the small data section. This may
9549 be useful for working around optimizer bugs.
9552 @opindex mconstant-gp
9553 Generate code that uses a single constant global pointer value. This is
9554 useful when compiling kernel code.
9558 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9559 This is useful when compiling firmware code.
9561 @item -minline-float-divide-min-latency
9562 @opindex minline-float-divide-min-latency
9563 Generate code for inline divides of floating point values
9564 using the minimum latency algorithm.
9566 @item -minline-float-divide-max-throughput
9567 @opindex minline-float-divide-max-throughput
9568 Generate code for inline divides of floating point values
9569 using the maximum throughput algorithm.
9571 @item -minline-int-divide-min-latency
9572 @opindex minline-int-divide-min-latency
9573 Generate code for inline divides of integer values
9574 using the minimum latency algorithm.
9576 @item -minline-int-divide-max-throughput
9577 @opindex minline-int-divide-max-throughput
9578 Generate code for inline divides of integer values
9579 using the maximum throughput algorithm.
9581 @item -minline-sqrt-min-latency
9582 @opindex minline-sqrt-min-latency
9583 Generate code for inline square roots
9584 using the minimum latency algorithm.
9586 @item -minline-sqrt-max-throughput
9587 @opindex minline-sqrt-max-throughput
9588 Generate code for inline square roots
9589 using the maximum throughput algorithm.
9591 @item -mno-dwarf2-asm
9593 @opindex mno-dwarf2-asm
9594 @opindex mdwarf2-asm
9595 Don't (or do) generate assembler code for the DWARF2 line number debugging
9596 info. This may be useful when not using the GNU assembler.
9598 @item -mearly-stop-bits
9599 @itemx -mno-early-stop-bits
9600 @opindex mearly-stop-bits
9601 @opindex mno-early-stop-bits
9602 Allow stop bits to be placed earlier than immediately preceding the
9603 instruction that triggered the stop bit. This can improve instruction
9604 scheduling, but does not always do so.
9606 @item -mfixed-range=@var{register-range}
9607 @opindex mfixed-range
9608 Generate code treating the given register range as fixed registers.
9609 A fixed register is one that the register allocator can not use. This is
9610 useful when compiling kernel code. A register range is specified as
9611 two registers separated by a dash. Multiple register ranges can be
9612 specified separated by a comma.
9614 @item -mtls-size=@var{tls-size}
9616 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9619 @item -mtune=@var{cpu-type}
9621 Tune the instruction scheduling for a particular CPU, Valid values are
9622 itanium, itanium1, merced, itanium2, and mckinley.
9628 Add support for multithreading using the POSIX threads library. This
9629 option sets flags for both the preprocessor and linker. It does
9630 not affect the thread safety of object code produced by the compiler or
9631 that of libraries supplied with it. These are HP-UX specific flags.
9637 Generate code for a 32-bit or 64-bit environment.
9638 The 32-bit environment sets int, long and pointer to 32 bits.
9639 The 64-bit environment sets int to 32 bits and long and pointer
9640 to 64 bits. These are HP-UX specific flags.
9645 @subsection M32C Options
9646 @cindex M32C options
9649 @item -mcpu=@var{name}
9651 Select the CPU for which code is generated. @var{name} may be one of
9652 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9653 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9658 Specifies that the program will be run on the simulator. This causes
9659 an alternate runtime library to be linked in which supports, for
9660 example, file I/O. You must not use this option when generating
9661 programs that will run on real hardware; you must provide your own
9662 runtime library for whatever I/O functions are needed.
9664 @item -memregs=@var{number}
9666 Specifies the number of memory-based pseudo-registers GCC will use
9667 during code generation. These pseudo-registers will be used like real
9668 registers, so there is a tradeoff between GCC's ability to fit the
9669 code into available registers, and the performance penalty of using
9670 memory instead of registers. Note that all modules in a program must
9671 be compiled with the same value for this option. Because of that, you
9672 must not use this option with the default runtime libraries gcc
9677 @node M32R/D Options
9678 @subsection M32R/D Options
9679 @cindex M32R/D options
9681 These @option{-m} options are defined for Renesas M32R/D architectures:
9686 Generate code for the M32R/2@.
9690 Generate code for the M32R/X@.
9694 Generate code for the M32R@. This is the default.
9697 @opindex mmodel=small
9698 Assume all objects live in the lower 16MB of memory (so that their addresses
9699 can be loaded with the @code{ld24} instruction), and assume all subroutines
9700 are reachable with the @code{bl} instruction.
9701 This is the default.
9703 The addressability of a particular object can be set with the
9704 @code{model} attribute.
9706 @item -mmodel=medium
9707 @opindex mmodel=medium
9708 Assume objects may be anywhere in the 32-bit address space (the compiler
9709 will generate @code{seth/add3} instructions to load their addresses), and
9710 assume all subroutines are reachable with the @code{bl} instruction.
9713 @opindex mmodel=large
9714 Assume objects may be anywhere in the 32-bit address space (the compiler
9715 will generate @code{seth/add3} instructions to load their addresses), and
9716 assume subroutines may not be reachable with the @code{bl} instruction
9717 (the compiler will generate the much slower @code{seth/add3/jl}
9718 instruction sequence).
9721 @opindex msdata=none
9722 Disable use of the small data area. Variables will be put into
9723 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9724 @code{section} attribute has been specified).
9725 This is the default.
9727 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9728 Objects may be explicitly put in the small data area with the
9729 @code{section} attribute using one of these sections.
9732 @opindex msdata=sdata
9733 Put small global and static data in the small data area, but do not
9734 generate special code to reference them.
9738 Put small global and static data in the small data area, and generate
9739 special instructions to reference them.
9743 @cindex smaller data references
9744 Put global and static objects less than or equal to @var{num} bytes
9745 into the small data or bss sections instead of the normal data or bss
9746 sections. The default value of @var{num} is 8.
9747 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9748 for this option to have any effect.
9750 All modules should be compiled with the same @option{-G @var{num}} value.
9751 Compiling with different values of @var{num} may or may not work; if it
9752 doesn't the linker will give an error message---incorrect code will not be
9757 Makes the M32R specific code in the compiler display some statistics
9758 that might help in debugging programs.
9761 @opindex malign-loops
9762 Align all loops to a 32-byte boundary.
9764 @item -mno-align-loops
9765 @opindex mno-align-loops
9766 Do not enforce a 32-byte alignment for loops. This is the default.
9768 @item -missue-rate=@var{number}
9769 @opindex missue-rate=@var{number}
9770 Issue @var{number} instructions per cycle. @var{number} can only be 1
9773 @item -mbranch-cost=@var{number}
9774 @opindex mbranch-cost=@var{number}
9775 @var{number} can only be 1 or 2. If it is 1 then branches will be
9776 preferred over conditional code, if it is 2, then the opposite will
9779 @item -mflush-trap=@var{number}
9780 @opindex mflush-trap=@var{number}
9781 Specifies the trap number to use to flush the cache. The default is
9782 12. Valid numbers are between 0 and 15 inclusive.
9784 @item -mno-flush-trap
9785 @opindex mno-flush-trap
9786 Specifies that the cache cannot be flushed by using a trap.
9788 @item -mflush-func=@var{name}
9789 @opindex mflush-func=@var{name}
9790 Specifies the name of the operating system function to call to flush
9791 the cache. The default is @emph{_flush_cache}, but a function call
9792 will only be used if a trap is not available.
9794 @item -mno-flush-func
9795 @opindex mno-flush-func
9796 Indicates that there is no OS function for flushing the cache.
9800 @node M680x0 Options
9801 @subsection M680x0 Options
9802 @cindex M680x0 options
9804 These are the @samp{-m} options defined for the 68000 series. The default
9805 values for these options depends on which style of 68000 was selected when
9806 the compiler was configured; the defaults for the most common choices are
9814 Generate output for a 68000. This is the default
9815 when the compiler is configured for 68000-based systems.
9817 Use this option for microcontrollers with a 68000 or EC000 core,
9818 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9824 Generate output for a 68020. This is the default
9825 when the compiler is configured for 68020-based systems.
9829 Generate output containing 68881 instructions for floating point.
9830 This is the default for most 68020 systems unless @option{--nfp} was
9831 specified when the compiler was configured.
9835 Generate output for a 68030. This is the default when the compiler is
9836 configured for 68030-based systems.
9840 Generate output for a 68040. This is the default when the compiler is
9841 configured for 68040-based systems.
9843 This option inhibits the use of 68881/68882 instructions that have to be
9844 emulated by software on the 68040. Use this option if your 68040 does not
9845 have code to emulate those instructions.
9849 Generate output for a 68060. This is the default when the compiler is
9850 configured for 68060-based systems.
9852 This option inhibits the use of 68020 and 68881/68882 instructions that
9853 have to be emulated by software on the 68060. Use this option if your 68060
9854 does not have code to emulate those instructions.
9858 Generate output for a CPU32. This is the default
9859 when the compiler is configured for CPU32-based systems.
9861 Use this option for microcontrollers with a
9862 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9863 68336, 68340, 68341, 68349 and 68360.
9867 Generate output for a 520X ``coldfire'' family cpu. This is the default
9868 when the compiler is configured for 520X-based systems.
9870 Use this option for microcontroller with a 5200 core, including
9871 the MCF5202, MCF5203, MCF5204 and MCF5202.
9876 Generate output for a 68040, without using any of the new instructions.
9877 This results in code which can run relatively efficiently on either a
9878 68020/68881 or a 68030 or a 68040. The generated code does use the
9879 68881 instructions that are emulated on the 68040.
9883 Generate output for a 68060, without using any of the new instructions.
9884 This results in code which can run relatively efficiently on either a
9885 68020/68881 or a 68030 or a 68040. The generated code does use the
9886 68881 instructions that are emulated on the 68060.
9889 @opindex msoft-float
9890 Generate output containing library calls for floating point.
9891 @strong{Warning:} the requisite libraries are not available for all m68k
9892 targets. Normally the facilities of the machine's usual C compiler are
9893 used, but this can't be done directly in cross-compilation. You must
9894 make your own arrangements to provide suitable library functions for
9895 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9896 @samp{m68k-*-coff} do provide software floating point support.
9900 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9901 Additionally, parameters passed on the stack are also aligned to a
9902 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9905 @opindex mnobitfield
9906 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9907 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9911 Do use the bit-field instructions. The @option{-m68020} option implies
9912 @option{-mbitfield}. This is the default if you use a configuration
9913 designed for a 68020.
9917 Use a different function-calling convention, in which functions
9918 that take a fixed number of arguments return with the @code{rtd}
9919 instruction, which pops their arguments while returning. This
9920 saves one instruction in the caller since there is no need to pop
9921 the arguments there.
9923 This calling convention is incompatible with the one normally
9924 used on Unix, so you cannot use it if you need to call libraries
9925 compiled with the Unix compiler.
9927 Also, you must provide function prototypes for all functions that
9928 take variable numbers of arguments (including @code{printf});
9929 otherwise incorrect code will be generated for calls to those
9932 In addition, seriously incorrect code will result if you call a
9933 function with too many arguments. (Normally, extra arguments are
9934 harmlessly ignored.)
9936 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9937 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9940 @itemx -mno-align-int
9942 @opindex mno-align-int
9943 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9944 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9945 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9946 Aligning variables on 32-bit boundaries produces code that runs somewhat
9947 faster on processors with 32-bit busses at the expense of more memory.
9949 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9950 align structures containing the above types differently than
9951 most published application binary interface specifications for the m68k.
9955 Use the pc-relative addressing mode of the 68000 directly, instead of
9956 using a global offset table. At present, this option implies @option{-fpic},
9957 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9958 not presently supported with @option{-mpcrel}, though this could be supported for
9959 68020 and higher processors.
9961 @item -mno-strict-align
9962 @itemx -mstrict-align
9963 @opindex mno-strict-align
9964 @opindex mstrict-align
9965 Do not (do) assume that unaligned memory references will be handled by
9969 Generate code that allows the data segment to be located in a different
9970 area of memory from the text segment. This allows for execute in place in
9971 an environment without virtual memory management. This option implies
9975 Generate code that assumes that the data segment follows the text segment.
9976 This is the default.
9978 @item -mid-shared-library
9979 Generate code that supports shared libraries via the library ID method.
9980 This allows for execute in place and shared libraries in an environment
9981 without virtual memory management. This option implies @option{-fPIC}.
9983 @item -mno-id-shared-library
9984 Generate code that doesn't assume ID based shared libraries are being used.
9985 This is the default.
9987 @item -mshared-library-id=n
9988 Specified the identification number of the ID based shared library being
9989 compiled. Specifying a value of 0 will generate more compact code, specifying
9990 other values will force the allocation of that number to the current
9991 library but is no more space or time efficient than omitting this option.
9995 @node M68hc1x Options
9996 @subsection M68hc1x Options
9997 @cindex M68hc1x options
9999 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10000 microcontrollers. The default values for these options depends on
10001 which style of microcontroller was selected when the compiler was configured;
10002 the defaults for the most common choices are given below.
10009 Generate output for a 68HC11. This is the default
10010 when the compiler is configured for 68HC11-based systems.
10016 Generate output for a 68HC12. This is the default
10017 when the compiler is configured for 68HC12-based systems.
10023 Generate output for a 68HCS12.
10025 @item -mauto-incdec
10026 @opindex mauto-incdec
10027 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10034 Enable the use of 68HC12 min and max instructions.
10037 @itemx -mno-long-calls
10038 @opindex mlong-calls
10039 @opindex mno-long-calls
10040 Treat all calls as being far away (near). If calls are assumed to be
10041 far away, the compiler will use the @code{call} instruction to
10042 call a function and the @code{rtc} instruction for returning.
10046 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10048 @item -msoft-reg-count=@var{count}
10049 @opindex msoft-reg-count
10050 Specify the number of pseudo-soft registers which are used for the
10051 code generation. The maximum number is 32. Using more pseudo-soft
10052 register may or may not result in better code depending on the program.
10053 The default is 4 for 68HC11 and 2 for 68HC12.
10057 @node MCore Options
10058 @subsection MCore Options
10059 @cindex MCore options
10061 These are the @samp{-m} options defined for the Motorola M*Core
10067 @itemx -mno-hardlit
10069 @opindex mno-hardlit
10070 Inline constants into the code stream if it can be done in two
10071 instructions or less.
10077 Use the divide instruction. (Enabled by default).
10079 @item -mrelax-immediate
10080 @itemx -mno-relax-immediate
10081 @opindex mrelax-immediate
10082 @opindex mno-relax-immediate
10083 Allow arbitrary sized immediates in bit operations.
10085 @item -mwide-bitfields
10086 @itemx -mno-wide-bitfields
10087 @opindex mwide-bitfields
10088 @opindex mno-wide-bitfields
10089 Always treat bit-fields as int-sized.
10091 @item -m4byte-functions
10092 @itemx -mno-4byte-functions
10093 @opindex m4byte-functions
10094 @opindex mno-4byte-functions
10095 Force all functions to be aligned to a four byte boundary.
10097 @item -mcallgraph-data
10098 @itemx -mno-callgraph-data
10099 @opindex mcallgraph-data
10100 @opindex mno-callgraph-data
10101 Emit callgraph information.
10104 @itemx -mno-slow-bytes
10105 @opindex mslow-bytes
10106 @opindex mno-slow-bytes
10107 Prefer word access when reading byte quantities.
10109 @item -mlittle-endian
10110 @itemx -mbig-endian
10111 @opindex mlittle-endian
10112 @opindex mbig-endian
10113 Generate code for a little endian target.
10119 Generate code for the 210 processor.
10123 @subsection MIPS Options
10124 @cindex MIPS options
10130 Generate big-endian code.
10134 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10137 @item -march=@var{arch}
10139 Generate code that will run on @var{arch}, which can be the name of a
10140 generic MIPS ISA, or the name of a particular processor.
10142 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10143 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10144 The processor names are:
10145 @samp{4kc}, @samp{4km}, @samp{4kp},
10146 @samp{5kc}, @samp{5kf},
10148 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10151 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10152 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10153 @samp{rm7000}, @samp{rm9000},
10156 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10157 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10158 The special value @samp{from-abi} selects the
10159 most compatible architecture for the selected ABI (that is,
10160 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10162 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10163 (for example, @samp{-march=r2k}). Prefixes are optional, and
10164 @samp{vr} may be written @samp{r}.
10166 GCC defines two macros based on the value of this option. The first
10167 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10168 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10169 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10170 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10171 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10173 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10174 above. In other words, it will have the full prefix and will not
10175 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10176 the macro names the resolved architecture (either @samp{"mips1"} or
10177 @samp{"mips3"}). It names the default architecture when no
10178 @option{-march} option is given.
10180 @item -mtune=@var{arch}
10182 Optimize for @var{arch}. Among other things, this option controls
10183 the way instructions are scheduled, and the perceived cost of arithmetic
10184 operations. The list of @var{arch} values is the same as for
10187 When this option is not used, GCC will optimize for the processor
10188 specified by @option{-march}. By using @option{-march} and
10189 @option{-mtune} together, it is possible to generate code that will
10190 run on a family of processors, but optimize the code for one
10191 particular member of that family.
10193 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10194 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10195 @samp{-march} ones described above.
10199 Equivalent to @samp{-march=mips1}.
10203 Equivalent to @samp{-march=mips2}.
10207 Equivalent to @samp{-march=mips3}.
10211 Equivalent to @samp{-march=mips4}.
10215 Equivalent to @samp{-march=mips32}.
10219 Equivalent to @samp{-march=mips32r2}.
10223 Equivalent to @samp{-march=mips64}.
10228 @opindex mno-mips16
10229 Generate (do not generate) MIPS16 code. If GCC is targetting a
10230 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10242 Generate code for the given ABI@.
10244 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10245 generates 64-bit code when you select a 64-bit architecture, but you
10246 can use @option{-mgp32} to get 32-bit code instead.
10248 For information about the O64 ABI, see
10249 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10252 @itemx -mno-abicalls
10254 @opindex mno-abicalls
10255 Generate (do not generate) SVR4-style position-independent code.
10256 @option{-mabicalls} is the default for SVR4-based systems.
10262 Lift (do not lift) the usual restrictions on the size of the global
10265 GCC normally uses a single instruction to load values from the GOT@.
10266 While this is relatively efficient, it will only work if the GOT
10267 is smaller than about 64k. Anything larger will cause the linker
10268 to report an error such as:
10270 @cindex relocation truncated to fit (MIPS)
10272 relocation truncated to fit: R_MIPS_GOT16 foobar
10275 If this happens, you should recompile your code with @option{-mxgot}.
10276 It should then work with very large GOTs, although it will also be
10277 less efficient, since it will take three instructions to fetch the
10278 value of a global symbol.
10280 Note that some linkers can create multiple GOTs. If you have such a
10281 linker, you should only need to use @option{-mxgot} when a single object
10282 file accesses more than 64k's worth of GOT entries. Very few do.
10284 These options have no effect unless GCC is generating position
10289 Assume that general-purpose registers are 32 bits wide.
10293 Assume that general-purpose registers are 64 bits wide.
10297 Assume that floating-point registers are 32 bits wide.
10301 Assume that floating-point registers are 64 bits wide.
10304 @opindex mhard-float
10305 Use floating-point coprocessor instructions.
10308 @opindex msoft-float
10309 Do not use floating-point coprocessor instructions. Implement
10310 floating-point calculations using library calls instead.
10312 @item -msingle-float
10313 @opindex msingle-float
10314 Assume that the floating-point coprocessor only supports single-precision
10317 @itemx -mdouble-float
10318 @opindex mdouble-float
10319 Assume that the floating-point coprocessor supports double-precision
10320 operations. This is the default.
10326 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10328 @itemx -mpaired-single
10329 @itemx -mno-paired-single
10330 @opindex mpaired-single
10331 @opindex mno-paired-single
10332 Use (do not use) paired-single floating-point instructions.
10333 @xref{MIPS Paired-Single Support}. This option can only be used
10334 when generating 64-bit code and requires hardware floating-point
10335 support to be enabled.
10340 @opindex mno-mips3d
10341 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10342 The option @option{-mips3d} implies @option{-mpaired-single}.
10346 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10347 an explanation of the default and the way that the pointer size is
10352 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10354 The default size of @code{int}s, @code{long}s and pointers depends on
10355 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10356 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10357 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10358 or the same size as integer registers, whichever is smaller.
10364 Assume (do not assume) that all symbols have 32-bit values, regardless
10365 of the selected ABI@. This option is useful in combination with
10366 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10367 to generate shorter and faster references to symbolic addresses.
10371 @cindex smaller data references (MIPS)
10372 @cindex gp-relative references (MIPS)
10373 Put global and static items less than or equal to @var{num} bytes into
10374 the small data or bss section instead of the normal data or bss section.
10375 This allows the data to be accessed using a single instruction.
10377 All modules should be compiled with the same @option{-G @var{num}}
10380 @item -membedded-data
10381 @itemx -mno-embedded-data
10382 @opindex membedded-data
10383 @opindex mno-embedded-data
10384 Allocate variables to the read-only data section first if possible, then
10385 next in the small data section if possible, otherwise in data. This gives
10386 slightly slower code than the default, but reduces the amount of RAM required
10387 when executing, and thus may be preferred for some embedded systems.
10389 @item -muninit-const-in-rodata
10390 @itemx -mno-uninit-const-in-rodata
10391 @opindex muninit-const-in-rodata
10392 @opindex mno-uninit-const-in-rodata
10393 Put uninitialized @code{const} variables in the read-only data section.
10394 This option is only meaningful in conjunction with @option{-membedded-data}.
10396 @item -msplit-addresses
10397 @itemx -mno-split-addresses
10398 @opindex msplit-addresses
10399 @opindex mno-split-addresses
10400 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10401 relocation operators. This option has been superseded by
10402 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10404 @item -mexplicit-relocs
10405 @itemx -mno-explicit-relocs
10406 @opindex mexplicit-relocs
10407 @opindex mno-explicit-relocs
10408 Use (do not use) assembler relocation operators when dealing with symbolic
10409 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10410 is to use assembler macros instead.
10412 @option{-mexplicit-relocs} is the default if GCC was configured
10413 to use an assembler that supports relocation operators.
10415 @item -mcheck-zero-division
10416 @itemx -mno-check-zero-division
10417 @opindex mcheck-zero-division
10418 @opindex mno-check-zero-division
10419 Trap (do not trap) on integer division by zero. The default is
10420 @option{-mcheck-zero-division}.
10422 @item -mdivide-traps
10423 @itemx -mdivide-breaks
10424 @opindex mdivide-traps
10425 @opindex mdivide-breaks
10426 MIPS systems check for division by zero by generating either a
10427 conditional trap or a break instruction. Using traps results in
10428 smaller code, but is only supported on MIPS II and later. Also, some
10429 versions of the Linux kernel have a bug that prevents trap from
10430 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10431 allow conditional traps on architectures that support them and
10432 @option{-mdivide-breaks} to force the use of breaks.
10434 The default is usually @option{-mdivide-traps}, but this can be
10435 overridden at configure time using @option{--with-divide=breaks}.
10436 Divide-by-zero checks can be completely disabled using
10437 @option{-mno-check-zero-division}.
10442 @opindex mno-memcpy
10443 Force (do not force) the use of @code{memcpy()} for non-trivial block
10444 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10445 most constant-sized copies.
10448 @itemx -mno-long-calls
10449 @opindex mlong-calls
10450 @opindex mno-long-calls
10451 Disable (do not disable) use of the @code{jal} instruction. Calling
10452 functions using @code{jal} is more efficient but requires the caller
10453 and callee to be in the same 256 megabyte segment.
10455 This option has no effect on abicalls code. The default is
10456 @option{-mno-long-calls}.
10462 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10463 instructions, as provided by the R4650 ISA@.
10466 @itemx -mno-fused-madd
10467 @opindex mfused-madd
10468 @opindex mno-fused-madd
10469 Enable (disable) use of the floating point multiply-accumulate
10470 instructions, when they are available. The default is
10471 @option{-mfused-madd}.
10473 When multiply-accumulate instructions are used, the intermediate
10474 product is calculated to infinite precision and is not subject to
10475 the FCSR Flush to Zero bit. This may be undesirable in some
10480 Tell the MIPS assembler to not run its preprocessor over user
10481 assembler files (with a @samp{.s} suffix) when assembling them.
10484 @itemx -mno-fix-r4000
10485 @opindex mfix-r4000
10486 @opindex mno-fix-r4000
10487 Work around certain R4000 CPU errata:
10490 A double-word or a variable shift may give an incorrect result if executed
10491 immediately after starting an integer division.
10493 A double-word or a variable shift may give an incorrect result if executed
10494 while an integer multiplication is in progress.
10496 An integer division may give an incorrect result if started in a delay slot
10497 of a taken branch or a jump.
10501 @itemx -mno-fix-r4400
10502 @opindex mfix-r4400
10503 @opindex mno-fix-r4400
10504 Work around certain R4400 CPU errata:
10507 A double-word or a variable shift may give an incorrect result if executed
10508 immediately after starting an integer division.
10512 @itemx -mno-fix-vr4120
10513 @opindex mfix-vr4120
10514 Work around certain VR4120 errata:
10517 @code{dmultu} does not always produce the correct result.
10519 @code{div} and @code{ddiv} do not always produce the correct result if one
10520 of the operands is negative.
10522 The workarounds for the division errata rely on special functions in
10523 @file{libgcc.a}. At present, these functions are only provided by
10524 the @code{mips64vr*-elf} configurations.
10526 Other VR4120 errata require a nop to be inserted between certain pairs of
10527 instructions. These errata are handled by the assembler, not by GCC itself.
10530 @opindex mfix-vr4130
10531 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10532 workarounds are implemented by the assembler rather than by GCC,
10533 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10534 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10535 instructions are available instead.
10538 @itemx -mno-fix-sb1
10540 Work around certain SB-1 CPU core errata.
10541 (This flag currently works around the SB-1 revision 2
10542 ``F1'' and ``F2'' floating point errata.)
10544 @item -mflush-func=@var{func}
10545 @itemx -mno-flush-func
10546 @opindex mflush-func
10547 Specifies the function to call to flush the I and D caches, or to not
10548 call any such function. If called, the function must take the same
10549 arguments as the common @code{_flush_func()}, that is, the address of the
10550 memory range for which the cache is being flushed, the size of the
10551 memory range, and the number 3 (to flush both caches). The default
10552 depends on the target GCC was configured for, but commonly is either
10553 @samp{_flush_func} or @samp{__cpu_flush}.
10555 @item -mbranch-likely
10556 @itemx -mno-branch-likely
10557 @opindex mbranch-likely
10558 @opindex mno-branch-likely
10559 Enable or disable use of Branch Likely instructions, regardless of the
10560 default for the selected architecture. By default, Branch Likely
10561 instructions may be generated if they are supported by the selected
10562 architecture. An exception is for the MIPS32 and MIPS64 architectures
10563 and processors which implement those architectures; for those, Branch
10564 Likely instructions will not be generated by default because the MIPS32
10565 and MIPS64 architectures specifically deprecate their use.
10567 @item -mfp-exceptions
10568 @itemx -mno-fp-exceptions
10569 @opindex mfp-exceptions
10570 Specifies whether FP exceptions are enabled. This affects how we schedule
10571 FP instructions for some processors. The default is that FP exceptions are
10574 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10575 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10578 @item -mvr4130-align
10579 @itemx -mno-vr4130-align
10580 @opindex mvr4130-align
10581 The VR4130 pipeline is two-way superscalar, but can only issue two
10582 instructions together if the first one is 8-byte aligned. When this
10583 option is enabled, GCC will align pairs of instructions that it
10584 thinks should execute in parallel.
10586 This option only has an effect when optimizing for the VR4130.
10587 It normally makes code faster, but at the expense of making it bigger.
10588 It is enabled by default at optimization level @option{-O3}.
10592 @subsection MMIX Options
10593 @cindex MMIX Options
10595 These options are defined for the MMIX:
10599 @itemx -mno-libfuncs
10601 @opindex mno-libfuncs
10602 Specify that intrinsic library functions are being compiled, passing all
10603 values in registers, no matter the size.
10606 @itemx -mno-epsilon
10608 @opindex mno-epsilon
10609 Generate floating-point comparison instructions that compare with respect
10610 to the @code{rE} epsilon register.
10612 @item -mabi=mmixware
10614 @opindex mabi-mmixware
10616 Generate code that passes function parameters and return values that (in
10617 the called function) are seen as registers @code{$0} and up, as opposed to
10618 the GNU ABI which uses global registers @code{$231} and up.
10620 @item -mzero-extend
10621 @itemx -mno-zero-extend
10622 @opindex mzero-extend
10623 @opindex mno-zero-extend
10624 When reading data from memory in sizes shorter than 64 bits, use (do not
10625 use) zero-extending load instructions by default, rather than
10626 sign-extending ones.
10629 @itemx -mno-knuthdiv
10631 @opindex mno-knuthdiv
10632 Make the result of a division yielding a remainder have the same sign as
10633 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10634 remainder follows the sign of the dividend. Both methods are
10635 arithmetically valid, the latter being almost exclusively used.
10637 @item -mtoplevel-symbols
10638 @itemx -mno-toplevel-symbols
10639 @opindex mtoplevel-symbols
10640 @opindex mno-toplevel-symbols
10641 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10642 code can be used with the @code{PREFIX} assembly directive.
10646 Generate an executable in the ELF format, rather than the default
10647 @samp{mmo} format used by the @command{mmix} simulator.
10649 @item -mbranch-predict
10650 @itemx -mno-branch-predict
10651 @opindex mbranch-predict
10652 @opindex mno-branch-predict
10653 Use (do not use) the probable-branch instructions, when static branch
10654 prediction indicates a probable branch.
10656 @item -mbase-addresses
10657 @itemx -mno-base-addresses
10658 @opindex mbase-addresses
10659 @opindex mno-base-addresses
10660 Generate (do not generate) code that uses @emph{base addresses}. Using a
10661 base address automatically generates a request (handled by the assembler
10662 and the linker) for a constant to be set up in a global register. The
10663 register is used for one or more base address requests within the range 0
10664 to 255 from the value held in the register. The generally leads to short
10665 and fast code, but the number of different data items that can be
10666 addressed is limited. This means that a program that uses lots of static
10667 data may require @option{-mno-base-addresses}.
10669 @item -msingle-exit
10670 @itemx -mno-single-exit
10671 @opindex msingle-exit
10672 @opindex mno-single-exit
10673 Force (do not force) generated code to have a single exit point in each
10677 @node MN10300 Options
10678 @subsection MN10300 Options
10679 @cindex MN10300 options
10681 These @option{-m} options are defined for Matsushita MN10300 architectures:
10686 Generate code to avoid bugs in the multiply instructions for the MN10300
10687 processors. This is the default.
10689 @item -mno-mult-bug
10690 @opindex mno-mult-bug
10691 Do not generate code to avoid bugs in the multiply instructions for the
10692 MN10300 processors.
10696 Generate code which uses features specific to the AM33 processor.
10700 Do not generate code which uses features specific to the AM33 processor. This
10703 @item -mreturn-pointer-on-d0
10704 @opindex mreturn-pointer-on-d0
10705 When generating a function which returns a pointer, return the pointer
10706 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
10707 only in a0, and attempts to call such functions without a prototype
10708 would result in errors. Note that this option is on by default; use
10709 @option{-mno-return-pointer-on-d0} to disable it.
10713 Do not link in the C run-time initialization object file.
10717 Indicate to the linker that it should perform a relaxation optimization pass
10718 to shorten branches, calls and absolute memory addresses. This option only
10719 has an effect when used on the command line for the final link step.
10721 This option makes symbolic debugging impossible.
10725 @subsection MT Options
10728 These @option{-m} options are defined for Morpho MT architectures:
10732 @item -march=@var{cpu-type}
10734 Generate code that will run on @var{cpu-type}, which is the name of a system
10735 representing a certain processor type. Possible values for
10736 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
10737 @samp{ms1-16-003} and @samp{ms2}.
10739 When this option is not used, the default is @option{-march=ms1-16-002}.
10743 Use byte loads and stores when generating code.
10747 Do not use byte loads and stores when generating code.
10751 Use simulator runtime
10755 Do not link in the C run-time initialization object file
10756 @file{crti.o}. Other run-time initialization and termination files
10757 such as @file{startup.o} and @file{exit.o} are still included on the
10758 linker command line.
10762 @node PDP-11 Options
10763 @subsection PDP-11 Options
10764 @cindex PDP-11 Options
10766 These options are defined for the PDP-11:
10771 Use hardware FPP floating point. This is the default. (FIS floating
10772 point on the PDP-11/40 is not supported.)
10775 @opindex msoft-float
10776 Do not use hardware floating point.
10780 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10784 Return floating-point results in memory. This is the default.
10788 Generate code for a PDP-11/40.
10792 Generate code for a PDP-11/45. This is the default.
10796 Generate code for a PDP-11/10.
10798 @item -mbcopy-builtin
10799 @opindex bcopy-builtin
10800 Use inline @code{movmemhi} patterns for copying memory. This is the
10805 Do not use inline @code{movmemhi} patterns for copying memory.
10811 Use 16-bit @code{int}. This is the default.
10817 Use 32-bit @code{int}.
10820 @itemx -mno-float32
10822 @opindex mno-float32
10823 Use 64-bit @code{float}. This is the default.
10826 @itemx -mno-float64
10828 @opindex mno-float64
10829 Use 32-bit @code{float}.
10833 Use @code{abshi2} pattern. This is the default.
10837 Do not use @code{abshi2} pattern.
10839 @item -mbranch-expensive
10840 @opindex mbranch-expensive
10841 Pretend that branches are expensive. This is for experimenting with
10842 code generation only.
10844 @item -mbranch-cheap
10845 @opindex mbranch-cheap
10846 Do not pretend that branches are expensive. This is the default.
10850 Generate code for a system with split I&D@.
10854 Generate code for a system without split I&D@. This is the default.
10858 Use Unix assembler syntax. This is the default when configured for
10859 @samp{pdp11-*-bsd}.
10863 Use DEC assembler syntax. This is the default when configured for any
10864 PDP-11 target other than @samp{pdp11-*-bsd}.
10867 @node PowerPC Options
10868 @subsection PowerPC Options
10869 @cindex PowerPC options
10871 These are listed under @xref{RS/6000 and PowerPC Options}.
10873 @node RS/6000 and PowerPC Options
10874 @subsection IBM RS/6000 and PowerPC Options
10875 @cindex RS/6000 and PowerPC Options
10876 @cindex IBM RS/6000 and PowerPC Options
10878 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10885 @itemx -mno-powerpc
10886 @itemx -mpowerpc-gpopt
10887 @itemx -mno-powerpc-gpopt
10888 @itemx -mpowerpc-gfxopt
10889 @itemx -mno-powerpc-gfxopt
10891 @itemx -mno-powerpc64
10895 @itemx -mno-popcntb
10901 @opindex mno-power2
10903 @opindex mno-powerpc
10904 @opindex mpowerpc-gpopt
10905 @opindex mno-powerpc-gpopt
10906 @opindex mpowerpc-gfxopt
10907 @opindex mno-powerpc-gfxopt
10908 @opindex mpowerpc64
10909 @opindex mno-powerpc64
10913 @opindex mno-popcntb
10916 GCC supports two related instruction set architectures for the
10917 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10918 instructions supported by the @samp{rios} chip set used in the original
10919 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10920 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
10921 the IBM 4xx, 6xx, and follow-on microprocessors.
10923 Neither architecture is a subset of the other. However there is a
10924 large common subset of instructions supported by both. An MQ
10925 register is included in processors supporting the POWER architecture.
10927 You use these options to specify which instructions are available on the
10928 processor you are using. The default value of these options is
10929 determined when configuring GCC@. Specifying the
10930 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10931 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10932 rather than the options listed above.
10934 The @option{-mpower} option allows GCC to generate instructions that
10935 are found only in the POWER architecture and to use the MQ register.
10936 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10937 to generate instructions that are present in the POWER2 architecture but
10938 not the original POWER architecture.
10940 The @option{-mpowerpc} option allows GCC to generate instructions that
10941 are found only in the 32-bit subset of the PowerPC architecture.
10942 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10943 GCC to use the optional PowerPC architecture instructions in the
10944 General Purpose group, including floating-point square root. Specifying
10945 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10946 use the optional PowerPC architecture instructions in the Graphics
10947 group, including floating-point select.
10949 The @option{-mmfcrf} option allows GCC to generate the move from
10950 condition register field instruction implemented on the POWER4
10951 processor and other processors that support the PowerPC V2.01
10953 The @option{-mpopcntb} option allows GCC to generate the popcount and
10954 double precision FP reciprocal estimate instruction implemented on the
10955 POWER5 processor and other processors that support the PowerPC V2.02
10957 The @option{-mfprnd} option allows GCC to generate the FP round to
10958 integer instructions implemented on the POWER5+ processor and other
10959 processors that support the PowerPC V2.03 architecture.
10961 The @option{-mpowerpc64} option allows GCC to generate the additional
10962 64-bit instructions that are found in the full PowerPC64 architecture
10963 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10964 @option{-mno-powerpc64}.
10966 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10967 will use only the instructions in the common subset of both
10968 architectures plus some special AIX common-mode calls, and will not use
10969 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10970 permits GCC to use any instruction from either architecture and to
10971 allow use of the MQ register; specify this for the Motorola MPC601.
10973 @item -mnew-mnemonics
10974 @itemx -mold-mnemonics
10975 @opindex mnew-mnemonics
10976 @opindex mold-mnemonics
10977 Select which mnemonics to use in the generated assembler code. With
10978 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10979 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10980 assembler mnemonics defined for the POWER architecture. Instructions
10981 defined in only one architecture have only one mnemonic; GCC uses that
10982 mnemonic irrespective of which of these options is specified.
10984 GCC defaults to the mnemonics appropriate for the architecture in
10985 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10986 value of these option. Unless you are building a cross-compiler, you
10987 should normally not specify either @option{-mnew-mnemonics} or
10988 @option{-mold-mnemonics}, but should instead accept the default.
10990 @item -mcpu=@var{cpu_type}
10992 Set architecture type, register usage, choice of mnemonics, and
10993 instruction scheduling parameters for machine type @var{cpu_type}.
10994 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10995 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10996 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10997 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10998 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10999 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11000 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11001 @samp{power4}, @samp{power5}, @samp{power5+},
11002 @samp{common}, @samp{powerpc}, @samp{powerpc64},
11003 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11005 @option{-mcpu=common} selects a completely generic processor. Code
11006 generated under this option will run on any POWER or PowerPC processor.
11007 GCC will use only the instructions in the common subset of both
11008 architectures, and will not use the MQ register. GCC assumes a generic
11009 processor model for scheduling purposes.
11011 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11012 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11013 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11014 types, with an appropriate, generic processor model assumed for
11015 scheduling purposes.
11017 The other options specify a specific processor. Code generated under
11018 those options will run best on that processor, and may not run at all on
11021 The @option{-mcpu} options automatically enable or disable the
11022 following options: @option{-maltivec}, @option{-mfprnd},
11023 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11024 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11025 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11026 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw}.
11027 The particular options
11028 set for any particular CPU will vary between compiler versions,
11029 depending on what setting seems to produce optimal code for that CPU;
11030 it doesn't necessarily reflect the actual hardware's capabilities. If
11031 you wish to set an individual option to a particular value, you may
11032 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
11035 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11036 not enabled or disabled by the @option{-mcpu} option at present because
11037 AIX does not have full support for these options. You may still
11038 enable or disable them individually if you're sure it'll work in your
11041 @item -mtune=@var{cpu_type}
11043 Set the instruction scheduling parameters for machine type
11044 @var{cpu_type}, but do not set the architecture type, register usage, or
11045 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11046 values for @var{cpu_type} are used for @option{-mtune} as for
11047 @option{-mcpu}. If both are specified, the code generated will use the
11048 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11049 scheduling parameters set by @option{-mtune}.
11055 Generate code to compute division as reciprocal estimate and iterative
11056 refinement, creating opportunities for increased throughput. This
11057 feature requires: optional PowerPC Graphics instruction set for single
11058 precision and FRE instruction for double precision, assuming divides
11059 cannot generate user-visible traps, and the domain values not include
11060 Infinities, denormals or zero denominator.
11063 @itemx -mno-altivec
11065 @opindex mno-altivec
11066 Generate code that uses (does not use) AltiVec instructions, and also
11067 enable the use of built-in functions that allow more direct access to
11068 the AltiVec instruction set. You may also need to set
11069 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11075 @opindex mno-vrsave
11076 Generate VRSAVE instructions when generating AltiVec code.
11080 Extend the current ABI with SPE ABI extensions. This does not change
11081 the default ABI, instead it adds the SPE ABI extensions to the current
11085 @opindex mabi=no-spe
11086 Disable Booke SPE ABI extensions for the current ABI@.
11089 @opindex msecure-plt
11090 Generate code that allows ld and ld.so to build executables and shared
11091 libraries with non-exec .plt and .got sections. This is a PowerPC
11092 32-bit SYSV ABI option.
11096 Generate code that uses a BSS .plt section that ld.so fills in, and
11097 requires .plt and .got sections that are both writable and executable.
11098 This is a PowerPC 32-bit SYSV ABI option.
11104 This switch enables or disables the generation of ISEL instructions.
11106 @item -misel=@var{yes/no}
11107 This switch has been deprecated. Use @option{-misel} and
11108 @option{-mno-isel} instead.
11114 This switch enables or disables the generation of SPE simd
11117 @item -mspe=@var{yes/no}
11118 This option has been deprecated. Use @option{-mspe} and
11119 @option{-mno-spe} instead.
11121 @item -mfloat-gprs=@var{yes/single/double/no}
11122 @itemx -mfloat-gprs
11123 @opindex mfloat-gprs
11124 This switch enables or disables the generation of floating point
11125 operations on the general purpose registers for architectures that
11128 The argument @var{yes} or @var{single} enables the use of
11129 single-precision floating point operations.
11131 The argument @var{double} enables the use of single and
11132 double-precision floating point operations.
11134 The argument @var{no} disables floating point operations on the
11135 general purpose registers.
11137 This option is currently only available on the MPC854x.
11143 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11144 targets (including GNU/Linux). The 32-bit environment sets int, long
11145 and pointer to 32 bits and generates code that runs on any PowerPC
11146 variant. The 64-bit environment sets int to 32 bits and long and
11147 pointer to 64 bits, and generates code for PowerPC64, as for
11148 @option{-mpowerpc64}.
11151 @itemx -mno-fp-in-toc
11152 @itemx -mno-sum-in-toc
11153 @itemx -mminimal-toc
11155 @opindex mno-fp-in-toc
11156 @opindex mno-sum-in-toc
11157 @opindex mminimal-toc
11158 Modify generation of the TOC (Table Of Contents), which is created for
11159 every executable file. The @option{-mfull-toc} option is selected by
11160 default. In that case, GCC will allocate at least one TOC entry for
11161 each unique non-automatic variable reference in your program. GCC
11162 will also place floating-point constants in the TOC@. However, only
11163 16,384 entries are available in the TOC@.
11165 If you receive a linker error message that saying you have overflowed
11166 the available TOC space, you can reduce the amount of TOC space used
11167 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11168 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11169 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11170 generate code to calculate the sum of an address and a constant at
11171 run-time instead of putting that sum into the TOC@. You may specify one
11172 or both of these options. Each causes GCC to produce very slightly
11173 slower and larger code at the expense of conserving TOC space.
11175 If you still run out of space in the TOC even when you specify both of
11176 these options, specify @option{-mminimal-toc} instead. This option causes
11177 GCC to make only one TOC entry for every file. When you specify this
11178 option, GCC will produce code that is slower and larger but which
11179 uses extremely little TOC space. You may wish to use this option
11180 only on files that contain less frequently executed code.
11186 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11187 @code{long} type, and the infrastructure needed to support them.
11188 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11189 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11190 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11193 @itemx -mno-xl-compat
11194 @opindex mxl-compat
11195 @opindex mno-xl-compat
11196 Produce code that conforms more closely to IBM XLC semantics when using
11197 AIX-compatible ABI. Pass floating-point arguments to prototyped
11198 functions beyond the register save area (RSA) on the stack in addition
11199 to argument FPRs. Do not assume that most significant double in 128
11200 bit long double value is properly rounded when comparing values.
11202 The AIX calling convention was extended but not initially documented to
11203 handle an obscure K&R C case of calling a function that takes the
11204 address of its arguments with fewer arguments than declared. AIX XL
11205 compilers access floating point arguments which do not fit in the
11206 RSA from the stack when a subroutine is compiled without
11207 optimization. Because always storing floating-point arguments on the
11208 stack is inefficient and rarely needed, this option is not enabled by
11209 default and only is necessary when calling subroutines compiled by AIX
11210 XL compilers without optimization.
11214 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11215 application written to use message passing with special startup code to
11216 enable the application to run. The system must have PE installed in the
11217 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11218 must be overridden with the @option{-specs=} option to specify the
11219 appropriate directory location. The Parallel Environment does not
11220 support threads, so the @option{-mpe} option and the @option{-pthread}
11221 option are incompatible.
11223 @item -malign-natural
11224 @itemx -malign-power
11225 @opindex malign-natural
11226 @opindex malign-power
11227 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11228 @option{-malign-natural} overrides the ABI-defined alignment of larger
11229 types, such as floating-point doubles, on their natural size-based boundary.
11230 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11231 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11233 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11237 @itemx -mhard-float
11238 @opindex msoft-float
11239 @opindex mhard-float
11240 Generate code that does not use (uses) the floating-point register set.
11241 Software floating point emulation is provided if you use the
11242 @option{-msoft-float} option, and pass the option to GCC when linking.
11245 @itemx -mno-multiple
11247 @opindex mno-multiple
11248 Generate code that uses (does not use) the load multiple word
11249 instructions and the store multiple word instructions. These
11250 instructions are generated by default on POWER systems, and not
11251 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11252 endian PowerPC systems, since those instructions do not work when the
11253 processor is in little endian mode. The exceptions are PPC740 and
11254 PPC750 which permit the instructions usage in little endian mode.
11259 @opindex mno-string
11260 Generate code that uses (does not use) the load string instructions
11261 and the store string word instructions to save multiple registers and
11262 do small block moves. These instructions are generated by default on
11263 POWER systems, and not generated on PowerPC systems. Do not use
11264 @option{-mstring} on little endian PowerPC systems, since those
11265 instructions do not work when the processor is in little endian mode.
11266 The exceptions are PPC740 and PPC750 which permit the instructions
11267 usage in little endian mode.
11272 @opindex mno-update
11273 Generate code that uses (does not use) the load or store instructions
11274 that update the base register to the address of the calculated memory
11275 location. These instructions are generated by default. If you use
11276 @option{-mno-update}, there is a small window between the time that the
11277 stack pointer is updated and the address of the previous frame is
11278 stored, which means code that walks the stack frame across interrupts or
11279 signals may get corrupted data.
11282 @itemx -mno-fused-madd
11283 @opindex mfused-madd
11284 @opindex mno-fused-madd
11285 Generate code that uses (does not use) the floating point multiply and
11286 accumulate instructions. These instructions are generated by default if
11287 hardware floating is used.
11293 Generate code that uses (does not use) the half-word multiply and
11294 multiply-accumulate instructions on the IBM 405 and 440 processors.
11295 These instructions are generated by default when targetting those
11298 @item -mno-bit-align
11300 @opindex mno-bit-align
11301 @opindex mbit-align
11302 On System V.4 and embedded PowerPC systems do not (do) force structures
11303 and unions that contain bit-fields to be aligned to the base type of the
11306 For example, by default a structure containing nothing but 8
11307 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11308 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11309 the structure would be aligned to a 1 byte boundary and be one byte in
11312 @item -mno-strict-align
11313 @itemx -mstrict-align
11314 @opindex mno-strict-align
11315 @opindex mstrict-align
11316 On System V.4 and embedded PowerPC systems do not (do) assume that
11317 unaligned memory references will be handled by the system.
11319 @item -mrelocatable
11320 @itemx -mno-relocatable
11321 @opindex mrelocatable
11322 @opindex mno-relocatable
11323 On embedded PowerPC systems generate code that allows (does not allow)
11324 the program to be relocated to a different address at runtime. If you
11325 use @option{-mrelocatable} on any module, all objects linked together must
11326 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11328 @item -mrelocatable-lib
11329 @itemx -mno-relocatable-lib
11330 @opindex mrelocatable-lib
11331 @opindex mno-relocatable-lib
11332 On embedded PowerPC systems generate code that allows (does not allow)
11333 the program to be relocated to a different address at runtime. Modules
11334 compiled with @option{-mrelocatable-lib} can be linked with either modules
11335 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11336 with modules compiled with the @option{-mrelocatable} options.
11342 On System V.4 and embedded PowerPC systems do not (do) assume that
11343 register 2 contains a pointer to a global area pointing to the addresses
11344 used in the program.
11347 @itemx -mlittle-endian
11349 @opindex mlittle-endian
11350 On System V.4 and embedded PowerPC systems compile code for the
11351 processor in little endian mode. The @option{-mlittle-endian} option is
11352 the same as @option{-mlittle}.
11355 @itemx -mbig-endian
11357 @opindex mbig-endian
11358 On System V.4 and embedded PowerPC systems compile code for the
11359 processor in big endian mode. The @option{-mbig-endian} option is
11360 the same as @option{-mbig}.
11362 @item -mdynamic-no-pic
11363 @opindex mdynamic-no-pic
11364 On Darwin and Mac OS X systems, compile code so that it is not
11365 relocatable, but that its external references are relocatable. The
11366 resulting code is suitable for applications, but not shared
11369 @item -mprioritize-restricted-insns=@var{priority}
11370 @opindex mprioritize-restricted-insns
11371 This option controls the priority that is assigned to
11372 dispatch-slot restricted instructions during the second scheduling
11373 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11374 @var{no/highest/second-highest} priority to dispatch slot restricted
11377 @item -msched-costly-dep=@var{dependence_type}
11378 @opindex msched-costly-dep
11379 This option controls which dependences are considered costly
11380 by the target during instruction scheduling. The argument
11381 @var{dependence_type} takes one of the following values:
11382 @var{no}: no dependence is costly,
11383 @var{all}: all dependences are costly,
11384 @var{true_store_to_load}: a true dependence from store to load is costly,
11385 @var{store_to_load}: any dependence from store to load is costly,
11386 @var{number}: any dependence which latency >= @var{number} is costly.
11388 @item -minsert-sched-nops=@var{scheme}
11389 @opindex minsert-sched-nops
11390 This option controls which nop insertion scheme will be used during
11391 the second scheduling pass. The argument @var{scheme} takes one of the
11393 @var{no}: Don't insert nops.
11394 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11395 according to the scheduler's grouping.
11396 @var{regroup_exact}: Insert nops to force costly dependent insns into
11397 separate groups. Insert exactly as many nops as needed to force an insn
11398 to a new group, according to the estimated processor grouping.
11399 @var{number}: Insert nops to force costly dependent insns into
11400 separate groups. Insert @var{number} nops to force an insn to a new group.
11403 @opindex mcall-sysv
11404 On System V.4 and embedded PowerPC systems compile code using calling
11405 conventions that adheres to the March 1995 draft of the System V
11406 Application Binary Interface, PowerPC processor supplement. This is the
11407 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11409 @item -mcall-sysv-eabi
11410 @opindex mcall-sysv-eabi
11411 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11413 @item -mcall-sysv-noeabi
11414 @opindex mcall-sysv-noeabi
11415 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11417 @item -mcall-solaris
11418 @opindex mcall-solaris
11419 On System V.4 and embedded PowerPC systems compile code for the Solaris
11423 @opindex mcall-linux
11424 On System V.4 and embedded PowerPC systems compile code for the
11425 Linux-based GNU system.
11429 On System V.4 and embedded PowerPC systems compile code for the
11430 Hurd-based GNU system.
11432 @item -mcall-netbsd
11433 @opindex mcall-netbsd
11434 On System V.4 and embedded PowerPC systems compile code for the
11435 NetBSD operating system.
11437 @item -maix-struct-return
11438 @opindex maix-struct-return
11439 Return all structures in memory (as specified by the AIX ABI)@.
11441 @item -msvr4-struct-return
11442 @opindex msvr4-struct-return
11443 Return structures smaller than 8 bytes in registers (as specified by the
11446 @item -mabi=@var{abi-type}
11448 Extend the current ABI with a particular extension, or remove such extension.
11449 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11453 @itemx -mno-prototype
11454 @opindex mprototype
11455 @opindex mno-prototype
11456 On System V.4 and embedded PowerPC systems assume that all calls to
11457 variable argument functions are properly prototyped. Otherwise, the
11458 compiler must insert an instruction before every non prototyped call to
11459 set or clear bit 6 of the condition code register (@var{CR}) to
11460 indicate whether floating point values were passed in the floating point
11461 registers in case the function takes a variable arguments. With
11462 @option{-mprototype}, only calls to prototyped variable argument functions
11463 will set or clear the bit.
11467 On embedded PowerPC systems, assume that the startup module is called
11468 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11469 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11474 On embedded PowerPC systems, assume that the startup module is called
11475 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11480 On embedded PowerPC systems, assume that the startup module is called
11481 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11484 @item -myellowknife
11485 @opindex myellowknife
11486 On embedded PowerPC systems, assume that the startup module is called
11487 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11492 On System V.4 and embedded PowerPC systems, specify that you are
11493 compiling for a VxWorks system.
11497 Specify that you are compiling for the WindISS simulation environment.
11501 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11502 header to indicate that @samp{eabi} extended relocations are used.
11508 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11509 Embedded Applications Binary Interface (eabi) which is a set of
11510 modifications to the System V.4 specifications. Selecting @option{-meabi}
11511 means that the stack is aligned to an 8 byte boundary, a function
11512 @code{__eabi} is called to from @code{main} to set up the eabi
11513 environment, and the @option{-msdata} option can use both @code{r2} and
11514 @code{r13} to point to two separate small data areas. Selecting
11515 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11516 do not call an initialization function from @code{main}, and the
11517 @option{-msdata} option will only use @code{r13} to point to a single
11518 small data area. The @option{-meabi} option is on by default if you
11519 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11522 @opindex msdata=eabi
11523 On System V.4 and embedded PowerPC systems, put small initialized
11524 @code{const} global and static data in the @samp{.sdata2} section, which
11525 is pointed to by register @code{r2}. Put small initialized
11526 non-@code{const} global and static data in the @samp{.sdata} section,
11527 which is pointed to by register @code{r13}. Put small uninitialized
11528 global and static data in the @samp{.sbss} section, which is adjacent to
11529 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11530 incompatible with the @option{-mrelocatable} option. The
11531 @option{-msdata=eabi} option also sets the @option{-memb} option.
11534 @opindex msdata=sysv
11535 On System V.4 and embedded PowerPC systems, put small global and static
11536 data in the @samp{.sdata} section, which is pointed to by register
11537 @code{r13}. Put small uninitialized global and static data in the
11538 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11539 The @option{-msdata=sysv} option is incompatible with the
11540 @option{-mrelocatable} option.
11542 @item -msdata=default
11544 @opindex msdata=default
11546 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11547 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11548 same as @option{-msdata=sysv}.
11551 @opindex msdata-data
11552 On System V.4 and embedded PowerPC systems, put small global
11553 data in the @samp{.sdata} section. Put small uninitialized global
11554 data in the @samp{.sbss} section. Do not use register @code{r13}
11555 to address small data however. This is the default behavior unless
11556 other @option{-msdata} options are used.
11560 @opindex msdata=none
11562 On embedded PowerPC systems, put all initialized global and static data
11563 in the @samp{.data} section, and all uninitialized data in the
11564 @samp{.bss} section.
11568 @cindex smaller data references (PowerPC)
11569 @cindex .sdata/.sdata2 references (PowerPC)
11570 On embedded PowerPC systems, put global and static items less than or
11571 equal to @var{num} bytes into the small data or bss sections instead of
11572 the normal data or bss section. By default, @var{num} is 8. The
11573 @option{-G @var{num}} switch is also passed to the linker.
11574 All modules should be compiled with the same @option{-G @var{num}} value.
11577 @itemx -mno-regnames
11579 @opindex mno-regnames
11580 On System V.4 and embedded PowerPC systems do (do not) emit register
11581 names in the assembly language output using symbolic forms.
11584 @itemx -mno-longcall
11586 @opindex mno-longcall
11587 Default to making all function calls indirectly, using a register, so
11588 that functions which reside further than 32 megabytes (33,554,432
11589 bytes) from the current location can be called. This setting can be
11590 overridden by the @code{shortcall} function attribute, or by
11591 @code{#pragma longcall(0)}.
11593 Some linkers are capable of detecting out-of-range calls and generating
11594 glue code on the fly. On these systems, long calls are unnecessary and
11595 generate slower code. As of this writing, the AIX linker can do this,
11596 as can the GNU linker for PowerPC/64. It is planned to add this feature
11597 to the GNU linker for 32-bit PowerPC systems as well.
11599 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11600 callee, L42'', plus a ``branch island'' (glue code). The two target
11601 addresses represent the callee and the ``branch island''. The
11602 Darwin/PPC linker will prefer the first address and generate a ``bl
11603 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11604 otherwise, the linker will generate ``bl L42'' to call the ``branch
11605 island''. The ``branch island'' is appended to the body of the
11606 calling function; it computes the full 32-bit address of the callee
11609 On Mach-O (Darwin) systems, this option directs the compiler emit to
11610 the glue for every direct call, and the Darwin linker decides whether
11611 to use or discard it.
11613 In the future, we may cause GCC to ignore all longcall specifications
11614 when the linker is known to generate glue.
11618 Adds support for multithreading with the @dfn{pthreads} library.
11619 This option sets flags for both the preprocessor and linker.
11623 @node S/390 and zSeries Options
11624 @subsection S/390 and zSeries Options
11625 @cindex S/390 and zSeries Options
11627 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11631 @itemx -msoft-float
11632 @opindex mhard-float
11633 @opindex msoft-float
11634 Use (do not use) the hardware floating-point instructions and registers
11635 for floating-point operations. When @option{-msoft-float} is specified,
11636 functions in @file{libgcc.a} will be used to perform floating-point
11637 operations. When @option{-mhard-float} is specified, the compiler
11638 generates IEEE floating-point instructions. This is the default.
11641 @itemx -mno-backchain
11642 @opindex mbackchain
11643 @opindex mno-backchain
11644 Store (do not store) the address of the caller's frame as backchain pointer
11645 into the callee's stack frame.
11646 A backchain may be needed to allow debugging using tools that do not understand
11647 DWARF-2 call frame information.
11648 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11649 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11650 the backchain is placed into the topmost word of the 96/160 byte register
11653 In general, code compiled with @option{-mbackchain} is call-compatible with
11654 code compiled with @option{-mmo-backchain}; however, use of the backchain
11655 for debugging purposes usually requires that the whole binary is built with
11656 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11657 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11658 to build a linux kernel use @option{-msoft-float}.
11660 The default is to not maintain the backchain.
11662 @item -mpacked-stack
11663 @item -mno-packed-stack
11664 @opindex mpacked-stack
11665 @opindex mno-packed-stack
11666 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11667 specified, the compiler uses the all fields of the 96/160 byte register save
11668 area only for their default purpose; unused fields still take up stack space.
11669 When @option{-mpacked-stack} is specified, register save slots are densely
11670 packed at the top of the register save area; unused space is reused for other
11671 purposes, allowing for more efficient use of the available stack space.
11672 However, when @option{-mbackchain} is also in effect, the topmost word of
11673 the save area is always used to store the backchain, and the return address
11674 register is always saved two words below the backchain.
11676 As long as the stack frame backchain is not used, code generated with
11677 @option{-mpacked-stack} is call-compatible with code generated with
11678 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11679 S/390 or zSeries generated code that uses the stack frame backchain at run
11680 time, not just for debugging purposes. Such code is not call-compatible
11681 with code compiled with @option{-mpacked-stack}. Also, note that the
11682 combination of @option{-mbackchain},
11683 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11684 to build a linux kernel use @option{-msoft-float}.
11686 The default is to not use the packed stack layout.
11689 @itemx -mno-small-exec
11690 @opindex msmall-exec
11691 @opindex mno-small-exec
11692 Generate (or do not generate) code using the @code{bras} instruction
11693 to do subroutine calls.
11694 This only works reliably if the total executable size does not
11695 exceed 64k. The default is to use the @code{basr} instruction instead,
11696 which does not have this limitation.
11702 When @option{-m31} is specified, generate code compliant to the
11703 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11704 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11705 particular to generate 64-bit instructions. For the @samp{s390}
11706 targets, the default is @option{-m31}, while the @samp{s390x}
11707 targets default to @option{-m64}.
11713 When @option{-mzarch} is specified, generate code using the
11714 instructions available on z/Architecture.
11715 When @option{-mesa} is specified, generate code using the
11716 instructions available on ESA/390. Note that @option{-mesa} is
11717 not possible with @option{-m64}.
11718 When generating code compliant to the GNU/Linux for S/390 ABI,
11719 the default is @option{-mesa}. When generating code compliant
11720 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11726 Generate (or do not generate) code using the @code{mvcle} instruction
11727 to perform block moves. When @option{-mno-mvcle} is specified,
11728 use a @code{mvc} loop instead. This is the default unless optimizing for
11735 Print (or do not print) additional debug information when compiling.
11736 The default is to not print debug information.
11738 @item -march=@var{cpu-type}
11740 Generate code that will run on @var{cpu-type}, which is the name of a system
11741 representing a certain processor type. Possible values for
11742 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11743 When generating code using the instructions available on z/Architecture,
11744 the default is @option{-march=z900}. Otherwise, the default is
11745 @option{-march=g5}.
11747 @item -mtune=@var{cpu-type}
11749 Tune to @var{cpu-type} everything applicable about the generated code,
11750 except for the ABI and the set of available instructions.
11751 The list of @var{cpu-type} values is the same as for @option{-march}.
11752 The default is the value used for @option{-march}.
11755 @itemx -mno-tpf-trace
11756 @opindex mtpf-trace
11757 @opindex mno-tpf-trace
11758 Generate code that adds (does not add) in TPF OS specific branches to trace
11759 routines in the operating system. This option is off by default, even
11760 when compiling for the TPF OS@.
11763 @itemx -mno-fused-madd
11764 @opindex mfused-madd
11765 @opindex mno-fused-madd
11766 Generate code that uses (does not use) the floating point multiply and
11767 accumulate instructions. These instructions are generated by default if
11768 hardware floating point is used.
11770 @item -mwarn-framesize=@var{framesize}
11771 @opindex mwarn-framesize
11772 Emit a warning if the current function exceeds the given frame size. Because
11773 this is a compile time check it doesn't need to be a real problem when the program
11774 runs. It is intended to identify functions which most probably cause
11775 a stack overflow. It is useful to be used in an environment with limited stack
11776 size e.g.@: the linux kernel.
11778 @item -mwarn-dynamicstack
11779 @opindex mwarn-dynamicstack
11780 Emit a warning if the function calls alloca or uses dynamically
11781 sized arrays. This is generally a bad idea with a limited stack size.
11783 @item -mstack-guard=@var{stack-guard}
11784 @item -mstack-size=@var{stack-size}
11785 @opindex mstack-guard
11786 @opindex mstack-size
11787 These arguments always have to be used in conjunction. If they are present the s390
11788 back end emits additional instructions in the function prologue which trigger a trap
11789 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11790 (remember that the stack on s390 grows downward). These options are intended to
11791 be used to help debugging stack overflow problems. The additionally emitted code
11792 causes only little overhead and hence can also be used in production like systems
11793 without greater performance degradation. The given values have to be exact
11794 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
11796 In order to be efficient the extra code makes the assumption that the stack starts
11797 at an address aligned to the value given by @var{stack-size}.
11801 @subsection SH Options
11803 These @samp{-m} options are defined for the SH implementations:
11808 Generate code for the SH1.
11812 Generate code for the SH2.
11815 Generate code for the SH2e.
11819 Generate code for the SH3.
11823 Generate code for the SH3e.
11827 Generate code for the SH4 without a floating-point unit.
11829 @item -m4-single-only
11830 @opindex m4-single-only
11831 Generate code for the SH4 with a floating-point unit that only
11832 supports single-precision arithmetic.
11836 Generate code for the SH4 assuming the floating-point unit is in
11837 single-precision mode by default.
11841 Generate code for the SH4.
11845 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11846 floating-point unit is not used.
11848 @item -m4a-single-only
11849 @opindex m4a-single-only
11850 Generate code for the SH4a, in such a way that no double-precision
11851 floating point operations are used.
11854 @opindex m4a-single
11855 Generate code for the SH4a assuming the floating-point unit is in
11856 single-precision mode by default.
11860 Generate code for the SH4a.
11864 Same as @option{-m4a-nofpu}, except that it implicitly passes
11865 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11866 instructions at the moment.
11870 Compile code for the processor in big endian mode.
11874 Compile code for the processor in little endian mode.
11878 Align doubles at 64-bit boundaries. Note that this changes the calling
11879 conventions, and thus some functions from the standard C library will
11880 not work unless you recompile it first with @option{-mdalign}.
11884 Shorten some address references at link time, when possible; uses the
11885 linker option @option{-relax}.
11889 Use 32-bit offsets in @code{switch} tables. The default is to use
11894 Enable the use of the instruction @code{fmovd}.
11898 Comply with the calling conventions defined by Renesas.
11902 Comply with the calling conventions defined by Renesas.
11906 Comply with the calling conventions defined for GCC before the Renesas
11907 conventions were available. This option is the default for all
11908 targets of the SH toolchain except for @samp{sh-symbianelf}.
11911 @opindex mnomacsave
11912 Mark the @code{MAC} register as call-clobbered, even if
11913 @option{-mhitachi} is given.
11917 Increase IEEE-compliance of floating-point code.
11918 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11919 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11920 comparisons of NANs / infinities incurs extra overhead in every
11921 floating point comparison, therefore the default is set to
11922 @option{-ffinite-math-only}.
11926 Dump instruction size and location in the assembly code.
11929 @opindex mpadstruct
11930 This option is deprecated. It pads structures to multiple of 4 bytes,
11931 which is incompatible with the SH ABI@.
11935 Optimize for space instead of speed. Implied by @option{-Os}.
11938 @opindex mprefergot
11939 When generating position-independent code, emit function calls using
11940 the Global Offset Table instead of the Procedure Linkage Table.
11944 Generate a library function call to invalidate instruction cache
11945 entries, after fixing up a trampoline. This library function call
11946 doesn't assume it can write to the whole memory address space. This
11947 is the default when the target is @code{sh-*-linux*}.
11949 @item -multcost=@var{number}
11950 @opindex multcost=@var{number}
11951 Set the cost to assume for a multiply insn.
11953 @item -mdiv=@var{strategy}
11954 @opindex mdiv=@var{strategy}
11955 Set the division strategy to use for SHmedia code. @var{strategy} must be
11956 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11957 inv:call2, inv:fp .
11958 "fp" performs the operation in floating point. This has a very high latency,
11959 but needs only a few instructions, so it might be a good choice if
11960 your code has enough easily exploitable ILP to allow the compiler to
11961 schedule the floating point instructions together with other instructions.
11962 Division by zero causes a floating point exception.
11963 "inv" uses integer operations to calculate the inverse of the divisor,
11964 and then multiplies the dividend with the inverse. This strategy allows
11965 cse and hoisting of the inverse calculation. Division by zero calculates
11966 an unspecified result, but does not trap.
11967 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11968 have been found, or if the entire operation has been hoisted to the same
11969 place, the last stages of the inverse calculation are intertwined with the
11970 final multiply to reduce the overall latency, at the expense of using a few
11971 more instructions, and thus offering fewer scheduling opportunities with
11973 "call" calls a library function that usually implements the inv:minlat
11975 This gives high code density for m5-*media-nofpu compilations.
11976 "call2" uses a different entry point of the same library function, where it
11977 assumes that a pointer to a lookup table has already been set up, which
11978 exposes the pointer load to cse / code hoisting optimizations.
11979 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11980 code generation, but if the code stays unoptimized, revert to the "call",
11981 "call2", or "fp" strategies, respectively. Note that the
11982 potentially-trapping side effect of division by zero is carried by a
11983 separate instruction, so it is possible that all the integer instructions
11984 are hoisted out, but the marker for the side effect stays where it is.
11985 A recombination to fp operations or a call is not possible in that case.
11986 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11987 that the inverse calculation was nor separated from the multiply, they speed
11988 up division where the dividend fits into 20 bits (plus sign where applicable),
11989 by inserting a test to skip a number of operations in this case; this test
11990 slows down the case of larger dividends. inv20u assumes the case of a such
11991 a small dividend to be unlikely, and inv20l assumes it to be likely.
11993 @item -mdivsi3_libfunc=@var{name}
11994 @opindex mdivsi3_libfunc=@var{name}
11995 Set the name of the library function used for 32 bit signed division to
11996 @var{name}. This only affect the name used in the call and inv:call
11997 division strategies, and the compiler will still expect the same
11998 sets of input/output/clobbered registers as if this option was not present.
12000 @item -madjust-unroll
12001 @opindex madjust-unroll
12002 Throttle unrolling to avoid thrashing target registers.
12003 This option only has an effect if the gcc code base supports the
12004 TARGET_ADJUST_UNROLL_MAX target hook.
12006 @item -mindexed-addressing
12007 @opindex mindexed-addressing
12008 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12009 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12010 semantics for the indexed addressing mode. The architecture allows the
12011 implementation of processors with 64 bit MMU, which the OS could use to
12012 get 32 bit addressing, but since no current hardware implementation supports
12013 this or any other way to make the indexed addressing mode safe to use in
12014 the 32 bit ABI, the default is -mno-indexed-addressing.
12016 @item -mgettrcost=@var{number}
12017 @opindex mgettrcost=@var{number}
12018 Set the cost assumed for the gettr instruction to @var{number}.
12019 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12023 Assume pt* instructions won't trap. This will generally generate better
12024 scheduled code, but is unsafe on current hardware. The current architecture
12025 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12026 This has the unintentional effect of making it unsafe to schedule ptabs /
12027 ptrel before a branch, or hoist it out of a loop. For example,
12028 __do_global_ctors, a part of libgcc that runs constructors at program
12029 startup, calls functions in a list which is delimited by -1. With the
12030 -mpt-fixed option, the ptabs will be done before testing against -1.
12031 That means that all the constructors will be run a bit quicker, but when
12032 the loop comes to the end of the list, the program crashes because ptabs
12033 loads -1 into a target register. Since this option is unsafe for any
12034 hardware implementing the current architecture specification, the default
12035 is -mno-pt-fixed. Unless the user specifies a specific cost with
12036 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12037 this deters register allocation using target registers for storing
12040 @item -minvalid-symbols
12041 @opindex minvalid-symbols
12042 Assume symbols might be invalid. Ordinary function symbols generated by
12043 the compiler will always be valid to load with movi/shori/ptabs or
12044 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12045 to generate symbols that will cause ptabs / ptrel to trap.
12046 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12047 It will then prevent cross-basic-block cse, hoisting and most scheduling
12048 of symbol loads. The default is @option{-mno-invalid-symbols}.
12051 @node SPARC Options
12052 @subsection SPARC Options
12053 @cindex SPARC options
12055 These @samp{-m} options are supported on the SPARC:
12058 @item -mno-app-regs
12060 @opindex mno-app-regs
12062 Specify @option{-mapp-regs} to generate output using the global registers
12063 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12066 To be fully SVR4 ABI compliant at the cost of some performance loss,
12067 specify @option{-mno-app-regs}. You should compile libraries and system
12068 software with this option.
12071 @itemx -mhard-float
12073 @opindex mhard-float
12074 Generate output containing floating point instructions. This is the
12078 @itemx -msoft-float
12080 @opindex msoft-float
12081 Generate output containing library calls for floating point.
12082 @strong{Warning:} the requisite libraries are not available for all SPARC
12083 targets. Normally the facilities of the machine's usual C compiler are
12084 used, but this cannot be done directly in cross-compilation. You must make
12085 your own arrangements to provide suitable library functions for
12086 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12087 @samp{sparclite-*-*} do provide software floating point support.
12089 @option{-msoft-float} changes the calling convention in the output file;
12090 therefore, it is only useful if you compile @emph{all} of a program with
12091 this option. In particular, you need to compile @file{libgcc.a}, the
12092 library that comes with GCC, with @option{-msoft-float} in order for
12095 @item -mhard-quad-float
12096 @opindex mhard-quad-float
12097 Generate output containing quad-word (long double) floating point
12100 @item -msoft-quad-float
12101 @opindex msoft-quad-float
12102 Generate output containing library calls for quad-word (long double)
12103 floating point instructions. The functions called are those specified
12104 in the SPARC ABI@. This is the default.
12106 As of this writing, there are no SPARC implementations that have hardware
12107 support for the quad-word floating point instructions. They all invoke
12108 a trap handler for one of these instructions, and then the trap handler
12109 emulates the effect of the instruction. Because of the trap handler overhead,
12110 this is much slower than calling the ABI library routines. Thus the
12111 @option{-msoft-quad-float} option is the default.
12113 @item -mno-unaligned-doubles
12114 @itemx -munaligned-doubles
12115 @opindex mno-unaligned-doubles
12116 @opindex munaligned-doubles
12117 Assume that doubles have 8 byte alignment. This is the default.
12119 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12120 alignment only if they are contained in another type, or if they have an
12121 absolute address. Otherwise, it assumes they have 4 byte alignment.
12122 Specifying this option avoids some rare compatibility problems with code
12123 generated by other compilers. It is not the default because it results
12124 in a performance loss, especially for floating point code.
12126 @item -mno-faster-structs
12127 @itemx -mfaster-structs
12128 @opindex mno-faster-structs
12129 @opindex mfaster-structs
12130 With @option{-mfaster-structs}, the compiler assumes that structures
12131 should have 8 byte alignment. This enables the use of pairs of
12132 @code{ldd} and @code{std} instructions for copies in structure
12133 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12134 However, the use of this changed alignment directly violates the SPARC
12135 ABI@. Thus, it's intended only for use on targets where the developer
12136 acknowledges that their resulting code will not be directly in line with
12137 the rules of the ABI@.
12139 @item -mimpure-text
12140 @opindex mimpure-text
12141 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12142 the compiler to not pass @option{-z text} to the linker when linking a
12143 shared object. Using this option, you can link position-dependent
12144 code into a shared object.
12146 @option{-mimpure-text} suppresses the ``relocations remain against
12147 allocatable but non-writable sections'' linker error message.
12148 However, the necessary relocations will trigger copy-on-write, and the
12149 shared object is not actually shared across processes. Instead of
12150 using @option{-mimpure-text}, you should compile all source code with
12151 @option{-fpic} or @option{-fPIC}.
12153 This option is only available on SunOS and Solaris.
12155 @item -mcpu=@var{cpu_type}
12157 Set the instruction set, register set, and instruction scheduling parameters
12158 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12159 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12160 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12161 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
12162 @samp{ultrasparc3}.
12164 Default instruction scheduling parameters are used for values that select
12165 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12166 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12168 Here is a list of each supported architecture and their supported
12173 v8: supersparc, hypersparc
12174 sparclite: f930, f934, sparclite86x
12176 v9: ultrasparc, ultrasparc3
12179 By default (unless configured otherwise), GCC generates code for the V7
12180 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12181 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12182 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12183 SPARCStation 1, 2, IPX etc.
12185 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12186 architecture. The only difference from V7 code is that the compiler emits
12187 the integer multiply and integer divide instructions which exist in SPARC-V8
12188 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12189 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12192 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12193 the SPARC architecture. This adds the integer multiply, integer divide step
12194 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12195 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12196 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12197 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12198 MB86934 chip, which is the more recent SPARClite with FPU@.
12200 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12201 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12202 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12203 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12204 optimizes it for the TEMIC SPARClet chip.
12206 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12207 architecture. This adds 64-bit integer and floating-point move instructions,
12208 3 additional floating-point condition code registers and conditional move
12209 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12210 optimizes it for the Sun UltraSPARC I/II chips. With
12211 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12212 Sun UltraSPARC III chip.
12214 @item -mtune=@var{cpu_type}
12216 Set the instruction scheduling parameters for machine type
12217 @var{cpu_type}, but do not set the instruction set or register set that the
12218 option @option{-mcpu=@var{cpu_type}} would.
12220 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12221 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12222 that select a particular cpu implementation. Those are @samp{cypress},
12223 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12224 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
12225 @samp{ultrasparc3}.
12230 @opindex mno-v8plus
12231 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12232 difference from the V8 ABI is that the global and out registers are
12233 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12234 mode for all SPARC-V9 processors.
12240 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12241 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12244 These @samp{-m} options are supported in addition to the above
12245 on SPARC-V9 processors in 64-bit environments:
12248 @item -mlittle-endian
12249 @opindex mlittle-endian
12250 Generate code for a processor running in little-endian mode. It is only
12251 available for a few configurations and most notably not on Solaris and Linux.
12257 Generate code for a 32-bit or 64-bit environment.
12258 The 32-bit environment sets int, long and pointer to 32 bits.
12259 The 64-bit environment sets int to 32 bits and long and pointer
12262 @item -mcmodel=medlow
12263 @opindex mcmodel=medlow
12264 Generate code for the Medium/Low code model: 64-bit addresses, programs
12265 must be linked in the low 32 bits of memory. Programs can be statically
12266 or dynamically linked.
12268 @item -mcmodel=medmid
12269 @opindex mcmodel=medmid
12270 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12271 must be linked in the low 44 bits of memory, the text and data segments must
12272 be less than 2GB in size and the data segment must be located within 2GB of
12275 @item -mcmodel=medany
12276 @opindex mcmodel=medany
12277 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12278 may be linked anywhere in memory, the text and data segments must be less
12279 than 2GB in size and the data segment must be located within 2GB of the
12282 @item -mcmodel=embmedany
12283 @opindex mcmodel=embmedany
12284 Generate code for the Medium/Anywhere code model for embedded systems:
12285 64-bit addresses, the text and data segments must be less than 2GB in
12286 size, both starting anywhere in memory (determined at link time). The
12287 global register %g4 points to the base of the data segment. Programs
12288 are statically linked and PIC is not supported.
12291 @itemx -mno-stack-bias
12292 @opindex mstack-bias
12293 @opindex mno-stack-bias
12294 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12295 frame pointer if present, are offset by @minus{}2047 which must be added back
12296 when making stack frame references. This is the default in 64-bit mode.
12297 Otherwise, assume no such offset is present.
12300 These switches are supported in addition to the above on Solaris:
12305 Add support for multithreading using the Solaris threads library. This
12306 option sets flags for both the preprocessor and linker. This option does
12307 not affect the thread safety of object code produced by the compiler or
12308 that of libraries supplied with it.
12312 Add support for multithreading using the POSIX threads library. This
12313 option sets flags for both the preprocessor and linker. This option does
12314 not affect the thread safety of object code produced by the compiler or
12315 that of libraries supplied with it.
12318 @node System V Options
12319 @subsection Options for System V
12321 These additional options are available on System V Release 4 for
12322 compatibility with other compilers on those systems:
12327 Create a shared object.
12328 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12332 Identify the versions of each tool used by the compiler, in a
12333 @code{.ident} assembler directive in the output.
12337 Refrain from adding @code{.ident} directives to the output file (this is
12340 @item -YP,@var{dirs}
12342 Search the directories @var{dirs}, and no others, for libraries
12343 specified with @option{-l}.
12345 @item -Ym,@var{dir}
12347 Look in the directory @var{dir} to find the M4 preprocessor.
12348 The assembler uses this option.
12349 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12350 @c the generic assembler that comes with Solaris takes just -Ym.
12353 @node TMS320C3x/C4x Options
12354 @subsection TMS320C3x/C4x Options
12355 @cindex TMS320C3x/C4x Options
12357 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12361 @item -mcpu=@var{cpu_type}
12363 Set the instruction set, register set, and instruction scheduling
12364 parameters for machine type @var{cpu_type}. Supported values for
12365 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12366 @samp{c44}. The default is @samp{c40} to generate code for the
12371 @itemx -msmall-memory
12373 @opindex mbig-memory
12375 @opindex msmall-memory
12377 Generates code for the big or small memory model. The small memory
12378 model assumed that all data fits into one 64K word page. At run-time
12379 the data page (DP) register must be set to point to the 64K page
12380 containing the .bss and .data program sections. The big memory model is
12381 the default and requires reloading of the DP register for every direct
12388 Allow (disallow) allocation of general integer operands into the block
12389 count register BK@.
12395 Enable (disable) generation of code using decrement and branch,
12396 DBcond(D), instructions. This is enabled by default for the C4x. To be
12397 on the safe side, this is disabled for the C3x, since the maximum
12398 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12399 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12400 that it can utilize the decrement and branch instruction, but will give
12401 up if there is more than one memory reference in the loop. Thus a loop
12402 where the loop counter is decremented can generate slightly more
12403 efficient code, in cases where the RPTB instruction cannot be utilized.
12405 @item -mdp-isr-reload
12407 @opindex mdp-isr-reload
12409 Force the DP register to be saved on entry to an interrupt service
12410 routine (ISR), reloaded to point to the data section, and restored on
12411 exit from the ISR@. This should not be required unless someone has
12412 violated the small memory model by modifying the DP register, say within
12419 For the C3x use the 24-bit MPYI instruction for integer multiplies
12420 instead of a library call to guarantee 32-bit results. Note that if one
12421 of the operands is a constant, then the multiplication will be performed
12422 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12423 then squaring operations are performed inline instead of a library call.
12426 @itemx -mno-fast-fix
12428 @opindex mno-fast-fix
12429 The C3x/C4x FIX instruction to convert a floating point value to an
12430 integer value chooses the nearest integer less than or equal to the
12431 floating point value rather than to the nearest integer. Thus if the
12432 floating point number is negative, the result will be incorrectly
12433 truncated an additional code is necessary to detect and correct this
12434 case. This option can be used to disable generation of the additional
12435 code required to correct the result.
12441 Enable (disable) generation of repeat block sequences using the RPTB
12442 instruction for zero overhead looping. The RPTB construct is only used
12443 for innermost loops that do not call functions or jump across the loop
12444 boundaries. There is no advantage having nested RPTB loops due to the
12445 overhead required to save and restore the RC, RS, and RE registers.
12446 This is enabled by default with @option{-O2}.
12448 @item -mrpts=@var{count}
12452 Enable (disable) the use of the single instruction repeat instruction
12453 RPTS@. If a repeat block contains a single instruction, and the loop
12454 count can be guaranteed to be less than the value @var{count}, GCC will
12455 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12456 then a RPTS will be emitted even if the loop count cannot be determined
12457 at compile time. Note that the repeated instruction following RPTS does
12458 not have to be reloaded from memory each iteration, thus freeing up the
12459 CPU buses for operands. However, since interrupts are blocked by this
12460 instruction, it is disabled by default.
12462 @item -mloop-unsigned
12463 @itemx -mno-loop-unsigned
12464 @opindex mloop-unsigned
12465 @opindex mno-loop-unsigned
12466 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12467 is @math{2^{31} + 1} since these instructions test if the iteration count is
12468 negative to terminate the loop. If the iteration count is unsigned
12469 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12470 exceeded. This switch allows an unsigned iteration count.
12474 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12475 with. This also enforces compatibility with the API employed by the TI
12476 C3x C compiler. For example, long doubles are passed as structures
12477 rather than in floating point registers.
12483 Generate code that uses registers (stack) for passing arguments to functions.
12484 By default, arguments are passed in registers where possible rather
12485 than by pushing arguments on to the stack.
12487 @item -mparallel-insns
12488 @itemx -mno-parallel-insns
12489 @opindex mparallel-insns
12490 @opindex mno-parallel-insns
12491 Allow the generation of parallel instructions. This is enabled by
12492 default with @option{-O2}.
12494 @item -mparallel-mpy
12495 @itemx -mno-parallel-mpy
12496 @opindex mparallel-mpy
12497 @opindex mno-parallel-mpy
12498 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12499 provided @option{-mparallel-insns} is also specified. These instructions have
12500 tight register constraints which can pessimize the code generation
12501 of large functions.
12506 @subsection V850 Options
12507 @cindex V850 Options
12509 These @samp{-m} options are defined for V850 implementations:
12513 @itemx -mno-long-calls
12514 @opindex mlong-calls
12515 @opindex mno-long-calls
12516 Treat all calls as being far away (near). If calls are assumed to be
12517 far away, the compiler will always load the functions address up into a
12518 register, and call indirect through the pointer.
12524 Do not optimize (do optimize) basic blocks that use the same index
12525 pointer 4 or more times to copy pointer into the @code{ep} register, and
12526 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12527 option is on by default if you optimize.
12529 @item -mno-prolog-function
12530 @itemx -mprolog-function
12531 @opindex mno-prolog-function
12532 @opindex mprolog-function
12533 Do not use (do use) external functions to save and restore registers
12534 at the prologue and epilogue of a function. The external functions
12535 are slower, but use less code space if more than one function saves
12536 the same number of registers. The @option{-mprolog-function} option
12537 is on by default if you optimize.
12541 Try to make the code as small as possible. At present, this just turns
12542 on the @option{-mep} and @option{-mprolog-function} options.
12544 @item -mtda=@var{n}
12546 Put static or global variables whose size is @var{n} bytes or less into
12547 the tiny data area that register @code{ep} points to. The tiny data
12548 area can hold up to 256 bytes in total (128 bytes for byte references).
12550 @item -msda=@var{n}
12552 Put static or global variables whose size is @var{n} bytes or less into
12553 the small data area that register @code{gp} points to. The small data
12554 area can hold up to 64 kilobytes.
12556 @item -mzda=@var{n}
12558 Put static or global variables whose size is @var{n} bytes or less into
12559 the first 32 kilobytes of memory.
12563 Specify that the target processor is the V850.
12566 @opindex mbig-switch
12567 Generate code suitable for big switch tables. Use this option only if
12568 the assembler/linker complain about out of range branches within a switch
12573 This option will cause r2 and r5 to be used in the code generated by
12574 the compiler. This setting is the default.
12576 @item -mno-app-regs
12577 @opindex mno-app-regs
12578 This option will cause r2 and r5 to be treated as fixed registers.
12582 Specify that the target processor is the V850E1. The preprocessor
12583 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12584 this option is used.
12588 Specify that the target processor is the V850E@. The preprocessor
12589 constant @samp{__v850e__} will be defined if this option is used.
12591 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12592 are defined then a default target processor will be chosen and the
12593 relevant @samp{__v850*__} preprocessor constant will be defined.
12595 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12596 defined, regardless of which processor variant is the target.
12598 @item -mdisable-callt
12599 @opindex mdisable-callt
12600 This option will suppress generation of the CALLT instruction for the
12601 v850e and v850e1 flavors of the v850 architecture. The default is
12602 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12607 @subsection VAX Options
12608 @cindex VAX options
12610 These @samp{-m} options are defined for the VAX:
12615 Do not output certain jump instructions (@code{aobleq} and so on)
12616 that the Unix assembler for the VAX cannot handle across long
12621 Do output those jump instructions, on the assumption that you
12622 will assemble with the GNU assembler.
12626 Output code for g-format floating point numbers instead of d-format.
12629 @node x86-64 Options
12630 @subsection x86-64 Options
12631 @cindex x86-64 options
12633 These are listed under @xref{i386 and x86-64 Options}.
12635 @node Xstormy16 Options
12636 @subsection Xstormy16 Options
12637 @cindex Xstormy16 Options
12639 These options are defined for Xstormy16:
12644 Choose startup files and linker script suitable for the simulator.
12647 @node Xtensa Options
12648 @subsection Xtensa Options
12649 @cindex Xtensa Options
12651 These options are supported for Xtensa targets:
12655 @itemx -mno-const16
12657 @opindex mno-const16
12658 Enable or disable use of @code{CONST16} instructions for loading
12659 constant values. The @code{CONST16} instruction is currently not a
12660 standard option from Tensilica. When enabled, @code{CONST16}
12661 instructions are always used in place of the standard @code{L32R}
12662 instructions. The use of @code{CONST16} is enabled by default only if
12663 the @code{L32R} instruction is not available.
12666 @itemx -mno-fused-madd
12667 @opindex mfused-madd
12668 @opindex mno-fused-madd
12669 Enable or disable use of fused multiply/add and multiply/subtract
12670 instructions in the floating-point option. This has no effect if the
12671 floating-point option is not also enabled. Disabling fused multiply/add
12672 and multiply/subtract instructions forces the compiler to use separate
12673 instructions for the multiply and add/subtract operations. This may be
12674 desirable in some cases where strict IEEE 754-compliant results are
12675 required: the fused multiply add/subtract instructions do not round the
12676 intermediate result, thereby producing results with @emph{more} bits of
12677 precision than specified by the IEEE standard. Disabling fused multiply
12678 add/subtract instructions also ensures that the program output is not
12679 sensitive to the compiler's ability to combine multiply and add/subtract
12682 @item -mtext-section-literals
12683 @itemx -mno-text-section-literals
12684 @opindex mtext-section-literals
12685 @opindex mno-text-section-literals
12686 Control the treatment of literal pools. The default is
12687 @option{-mno-text-section-literals}, which places literals in a separate
12688 section in the output file. This allows the literal pool to be placed
12689 in a data RAM/ROM, and it also allows the linker to combine literal
12690 pools from separate object files to remove redundant literals and
12691 improve code size. With @option{-mtext-section-literals}, the literals
12692 are interspersed in the text section in order to keep them as close as
12693 possible to their references. This may be necessary for large assembly
12696 @item -mtarget-align
12697 @itemx -mno-target-align
12698 @opindex mtarget-align
12699 @opindex mno-target-align
12700 When this option is enabled, GCC instructs the assembler to
12701 automatically align instructions to reduce branch penalties at the
12702 expense of some code density. The assembler attempts to widen density
12703 instructions to align branch targets and the instructions following call
12704 instructions. If there are not enough preceding safe density
12705 instructions to align a target, no widening will be performed. The
12706 default is @option{-mtarget-align}. These options do not affect the
12707 treatment of auto-aligned instructions like @code{LOOP}, which the
12708 assembler will always align, either by widening density instructions or
12709 by inserting no-op instructions.
12712 @itemx -mno-longcalls
12713 @opindex mlongcalls
12714 @opindex mno-longcalls
12715 When this option is enabled, GCC instructs the assembler to translate
12716 direct calls to indirect calls unless it can determine that the target
12717 of a direct call is in the range allowed by the call instruction. This
12718 translation typically occurs for calls to functions in other source
12719 files. Specifically, the assembler translates a direct @code{CALL}
12720 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12721 The default is @option{-mno-longcalls}. This option should be used in
12722 programs where the call target can potentially be out of range. This
12723 option is implemented in the assembler, not the compiler, so the
12724 assembly code generated by GCC will still show direct call
12725 instructions---look at the disassembled object code to see the actual
12726 instructions. Note that the assembler will use an indirect call for
12727 every cross-file call, not just those that really will be out of range.
12730 @node zSeries Options
12731 @subsection zSeries Options
12732 @cindex zSeries options
12734 These are listed under @xref{S/390 and zSeries Options}.
12736 @node Code Gen Options
12737 @section Options for Code Generation Conventions
12738 @cindex code generation conventions
12739 @cindex options, code generation
12740 @cindex run-time options
12742 These machine-independent options control the interface conventions
12743 used in code generation.
12745 Most of them have both positive and negative forms; the negative form
12746 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12747 one of the forms is listed---the one which is not the default. You
12748 can figure out the other form by either removing @samp{no-} or adding
12752 @item -fbounds-check
12753 @opindex fbounds-check
12754 For front-ends that support it, generate additional code to check that
12755 indices used to access arrays are within the declared range. This is
12756 currently only supported by the Java and Fortran 77 front-ends, where
12757 this option defaults to true and false respectively.
12761 This option generates traps for signed overflow on addition, subtraction,
12762 multiplication operations.
12766 This option instructs the compiler to assume that signed arithmetic
12767 overflow of addition, subtraction and multiplication wraps around
12768 using twos-complement representation. This flag enables some optimizations
12769 and disables others. This option is enabled by default for the Java
12770 front-end, as required by the Java language specification.
12773 @opindex fexceptions
12774 Enable exception handling. Generates extra code needed to propagate
12775 exceptions. For some targets, this implies GCC will generate frame
12776 unwind information for all functions, which can produce significant data
12777 size overhead, although it does not affect execution. If you do not
12778 specify this option, GCC will enable it by default for languages like
12779 C++ which normally require exception handling, and disable it for
12780 languages like C that do not normally require it. However, you may need
12781 to enable this option when compiling C code that needs to interoperate
12782 properly with exception handlers written in C++. You may also wish to
12783 disable this option if you are compiling older C++ programs that don't
12784 use exception handling.
12786 @item -fnon-call-exceptions
12787 @opindex fnon-call-exceptions
12788 Generate code that allows trapping instructions to throw exceptions.
12789 Note that this requires platform-specific runtime support that does
12790 not exist everywhere. Moreover, it only allows @emph{trapping}
12791 instructions to throw exceptions, i.e.@: memory references or floating
12792 point instructions. It does not allow exceptions to be thrown from
12793 arbitrary signal handlers such as @code{SIGALRM}.
12795 @item -funwind-tables
12796 @opindex funwind-tables
12797 Similar to @option{-fexceptions}, except that it will just generate any needed
12798 static data, but will not affect the generated code in any other way.
12799 You will normally not enable this option; instead, a language processor
12800 that needs this handling would enable it on your behalf.
12802 @item -fasynchronous-unwind-tables
12803 @opindex fasynchronous-unwind-tables
12804 Generate unwind table in dwarf2 format, if supported by target machine. The
12805 table is exact at each instruction boundary, so it can be used for stack
12806 unwinding from asynchronous events (such as debugger or garbage collector).
12808 @item -fpcc-struct-return
12809 @opindex fpcc-struct-return
12810 Return ``short'' @code{struct} and @code{union} values in memory like
12811 longer ones, rather than in registers. This convention is less
12812 efficient, but it has the advantage of allowing intercallability between
12813 GCC-compiled files and files compiled with other compilers, particularly
12814 the Portable C Compiler (pcc).
12816 The precise convention for returning structures in memory depends
12817 on the target configuration macros.
12819 Short structures and unions are those whose size and alignment match
12820 that of some integer type.
12822 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12823 switch is not binary compatible with code compiled with the
12824 @option{-freg-struct-return} switch.
12825 Use it to conform to a non-default application binary interface.
12827 @item -freg-struct-return
12828 @opindex freg-struct-return
12829 Return @code{struct} and @code{union} values in registers when possible.
12830 This is more efficient for small structures than
12831 @option{-fpcc-struct-return}.
12833 If you specify neither @option{-fpcc-struct-return} nor
12834 @option{-freg-struct-return}, GCC defaults to whichever convention is
12835 standard for the target. If there is no standard convention, GCC
12836 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12837 the principal compiler. In those cases, we can choose the standard, and
12838 we chose the more efficient register return alternative.
12840 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12841 switch is not binary compatible with code compiled with the
12842 @option{-fpcc-struct-return} switch.
12843 Use it to conform to a non-default application binary interface.
12845 @item -fshort-enums
12846 @opindex fshort-enums
12847 Allocate to an @code{enum} type only as many bytes as it needs for the
12848 declared range of possible values. Specifically, the @code{enum} type
12849 will be equivalent to the smallest integer type which has enough room.
12851 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12852 code that is not binary compatible with code generated without that switch.
12853 Use it to conform to a non-default application binary interface.
12855 @item -fshort-double
12856 @opindex fshort-double
12857 Use the same size for @code{double} as for @code{float}.
12859 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12860 code that is not binary compatible with code generated without that switch.
12861 Use it to conform to a non-default application binary interface.
12863 @item -fshort-wchar
12864 @opindex fshort-wchar
12865 Override the underlying type for @samp{wchar_t} to be @samp{short
12866 unsigned int} instead of the default for the target. This option is
12867 useful for building programs to run under WINE@.
12869 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12870 code that is not binary compatible with code generated without that switch.
12871 Use it to conform to a non-default application binary interface.
12873 @item -fshared-data
12874 @opindex fshared-data
12875 Requests that the data and non-@code{const} variables of this
12876 compilation be shared data rather than private data. The distinction
12877 makes sense only on certain operating systems, where shared data is
12878 shared between processes running the same program, while private data
12879 exists in one copy per process.
12882 @opindex fno-common
12883 In C, allocate even uninitialized global variables in the data section of the
12884 object file, rather than generating them as common blocks. This has the
12885 effect that if the same variable is declared (without @code{extern}) in
12886 two different compilations, you will get an error when you link them.
12887 The only reason this might be useful is if you wish to verify that the
12888 program will work on other systems which always work this way.
12892 Ignore the @samp{#ident} directive.
12894 @item -finhibit-size-directive
12895 @opindex finhibit-size-directive
12896 Don't output a @code{.size} assembler directive, or anything else that
12897 would cause trouble if the function is split in the middle, and the
12898 two halves are placed at locations far apart in memory. This option is
12899 used when compiling @file{crtstuff.c}; you should not need to use it
12902 @item -fverbose-asm
12903 @opindex fverbose-asm
12904 Put extra commentary information in the generated assembly code to
12905 make it more readable. This option is generally only of use to those
12906 who actually need to read the generated assembly code (perhaps while
12907 debugging the compiler itself).
12909 @option{-fno-verbose-asm}, the default, causes the
12910 extra information to be omitted and is useful when comparing two assembler
12915 @cindex global offset table
12917 Generate position-independent code (PIC) suitable for use in a shared
12918 library, if supported for the target machine. Such code accesses all
12919 constant addresses through a global offset table (GOT)@. The dynamic
12920 loader resolves the GOT entries when the program starts (the dynamic
12921 loader is not part of GCC; it is part of the operating system). If
12922 the GOT size for the linked executable exceeds a machine-specific
12923 maximum size, you get an error message from the linker indicating that
12924 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12925 instead. (These maximums are 8k on the SPARC and 32k
12926 on the m68k and RS/6000. The 386 has no such limit.)
12928 Position-independent code requires special support, and therefore works
12929 only on certain machines. For the 386, GCC supports PIC for System V
12930 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12931 position-independent.
12933 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12938 If supported for the target machine, emit position-independent code,
12939 suitable for dynamic linking and avoiding any limit on the size of the
12940 global offset table. This option makes a difference on the m68k,
12941 PowerPC and SPARC@.
12943 Position-independent code requires special support, and therefore works
12944 only on certain machines.
12946 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12953 These options are similar to @option{-fpic} and @option{-fPIC}, but
12954 generated position independent code can be only linked into executables.
12955 Usually these options are used when @option{-pie} GCC option will be
12956 used during linking.
12958 @item -fno-jump-tables
12959 @opindex fno-jump-tables
12960 Do not use jump tables for switch statements even where it would be
12961 more efficient than other code generation strategies. This option is
12962 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12963 building code which forms part of a dynamic linker and cannot
12964 reference the address of a jump table. On some targets, jump tables
12965 do not require a GOT and this option is not needed.
12967 @item -ffixed-@var{reg}
12969 Treat the register named @var{reg} as a fixed register; generated code
12970 should never refer to it (except perhaps as a stack pointer, frame
12971 pointer or in some other fixed role).
12973 @var{reg} must be the name of a register. The register names accepted
12974 are machine-specific and are defined in the @code{REGISTER_NAMES}
12975 macro in the machine description macro file.
12977 This flag does not have a negative form, because it specifies a
12980 @item -fcall-used-@var{reg}
12981 @opindex fcall-used
12982 Treat the register named @var{reg} as an allocable register that is
12983 clobbered by function calls. It may be allocated for temporaries or
12984 variables that do not live across a call. Functions compiled this way
12985 will not save and restore the register @var{reg}.
12987 It is an error to used this flag with the frame pointer or stack pointer.
12988 Use of this flag for other registers that have fixed pervasive roles in
12989 the machine's execution model will produce disastrous results.
12991 This flag does not have a negative form, because it specifies a
12994 @item -fcall-saved-@var{reg}
12995 @opindex fcall-saved
12996 Treat the register named @var{reg} as an allocable register saved by
12997 functions. It may be allocated even for temporaries or variables that
12998 live across a call. Functions compiled this way will save and restore
12999 the register @var{reg} if they use it.
13001 It is an error to used this flag with the frame pointer or stack pointer.
13002 Use of this flag for other registers that have fixed pervasive roles in
13003 the machine's execution model will produce disastrous results.
13005 A different sort of disaster will result from the use of this flag for
13006 a register in which function values may be returned.
13008 This flag does not have a negative form, because it specifies a
13011 @item -fpack-struct[=@var{n}]
13012 @opindex fpack-struct
13013 Without a value specified, pack all structure members together without
13014 holes. When a value is specified (which must be a small power of two), pack
13015 structure members according to this value, representing the maximum
13016 alignment (that is, objects with default alignment requirements larger than
13017 this will be output potentially unaligned at the next fitting location.
13019 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13020 code that is not binary compatible with code generated without that switch.
13021 Additionally, it makes the code suboptimal.
13022 Use it to conform to a non-default application binary interface.
13024 @item -finstrument-functions
13025 @opindex finstrument-functions
13026 Generate instrumentation calls for entry and exit to functions. Just
13027 after function entry and just before function exit, the following
13028 profiling functions will be called with the address of the current
13029 function and its call site. (On some platforms,
13030 @code{__builtin_return_address} does not work beyond the current
13031 function, so the call site information may not be available to the
13032 profiling functions otherwise.)
13035 void __cyg_profile_func_enter (void *this_fn,
13037 void __cyg_profile_func_exit (void *this_fn,
13041 The first argument is the address of the start of the current function,
13042 which may be looked up exactly in the symbol table.
13044 This instrumentation is also done for functions expanded inline in other
13045 functions. The profiling calls will indicate where, conceptually, the
13046 inline function is entered and exited. This means that addressable
13047 versions of such functions must be available. If all your uses of a
13048 function are expanded inline, this may mean an additional expansion of
13049 code size. If you use @samp{extern inline} in your C code, an
13050 addressable version of such functions must be provided. (This is
13051 normally the case anyways, but if you get lucky and the optimizer always
13052 expands the functions inline, you might have gotten away without
13053 providing static copies.)
13055 A function may be given the attribute @code{no_instrument_function}, in
13056 which case this instrumentation will not be done. This can be used, for
13057 example, for the profiling functions listed above, high-priority
13058 interrupt routines, and any functions from which the profiling functions
13059 cannot safely be called (perhaps signal handlers, if the profiling
13060 routines generate output or allocate memory).
13062 @item -fstack-check
13063 @opindex fstack-check
13064 Generate code to verify that you do not go beyond the boundary of the
13065 stack. You should specify this flag if you are running in an
13066 environment with multiple threads, but only rarely need to specify it in
13067 a single-threaded environment since stack overflow is automatically
13068 detected on nearly all systems if there is only one stack.
13070 Note that this switch does not actually cause checking to be done; the
13071 operating system must do that. The switch causes generation of code
13072 to ensure that the operating system sees the stack being extended.
13074 @item -fstack-limit-register=@var{reg}
13075 @itemx -fstack-limit-symbol=@var{sym}
13076 @itemx -fno-stack-limit
13077 @opindex fstack-limit-register
13078 @opindex fstack-limit-symbol
13079 @opindex fno-stack-limit
13080 Generate code to ensure that the stack does not grow beyond a certain value,
13081 either the value of a register or the address of a symbol. If the stack
13082 would grow beyond the value, a signal is raised. For most targets,
13083 the signal is raised before the stack overruns the boundary, so
13084 it is possible to catch the signal without taking special precautions.
13086 For instance, if the stack starts at absolute address @samp{0x80000000}
13087 and grows downwards, you can use the flags
13088 @option{-fstack-limit-symbol=__stack_limit} and
13089 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13090 of 128KB@. Note that this may only work with the GNU linker.
13092 @cindex aliasing of parameters
13093 @cindex parameters, aliased
13094 @item -fargument-alias
13095 @itemx -fargument-noalias
13096 @itemx -fargument-noalias-global
13097 @opindex fargument-alias
13098 @opindex fargument-noalias
13099 @opindex fargument-noalias-global
13100 Specify the possible relationships among parameters and between
13101 parameters and global data.
13103 @option{-fargument-alias} specifies that arguments (parameters) may
13104 alias each other and may alias global storage.@*
13105 @option{-fargument-noalias} specifies that arguments do not alias
13106 each other, but may alias global storage.@*
13107 @option{-fargument-noalias-global} specifies that arguments do not
13108 alias each other and do not alias global storage.
13110 Each language will automatically use whatever option is required by
13111 the language standard. You should not need to use these options yourself.
13113 @item -fleading-underscore
13114 @opindex fleading-underscore
13115 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13116 change the way C symbols are represented in the object file. One use
13117 is to help link with legacy assembly code.
13119 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13120 generate code that is not binary compatible with code generated without that
13121 switch. Use it to conform to a non-default application binary interface.
13122 Not all targets provide complete support for this switch.
13124 @item -ftls-model=@var{model}
13125 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13126 The @var{model} argument should be one of @code{global-dynamic},
13127 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13129 The default without @option{-fpic} is @code{initial-exec}; with
13130 @option{-fpic} the default is @code{global-dynamic}.
13132 @item -fvisibility=@var{default|internal|hidden|protected}
13133 @opindex fvisibility
13134 Set the default ELF image symbol visibility to the specified option---all
13135 symbols will be marked with this unless overridden within the code.
13136 Using this feature can very substantially improve linking and
13137 load times of shared object libraries, produce more optimized
13138 code, provide near-perfect API export and prevent symbol clashes.
13139 It is @strong{strongly} recommended that you use this in any shared objects
13142 Despite the nomenclature, @code{default} always means public ie;
13143 available to be linked against from outside the shared object.
13144 @code{protected} and @code{internal} are pretty useless in real-world
13145 usage so the only other commonly used option will be @code{hidden}.
13146 The default if @option{-fvisibility} isn't specified is
13147 @code{default}, i.e., make every
13148 symbol public---this causes the same behavior as previous versions of
13151 A good explanation of the benefits offered by ensuring ELF
13152 symbols have the correct visibility is given by ``How To Write
13153 Shared Libraries'' by Ulrich Drepper (which can be found at
13154 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13155 solution made possible by this option to marking things hidden when
13156 the default is public is to make the default hidden and mark things
13157 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13158 and @code{__attribute__ ((visibility("default")))} instead of
13159 @code{__declspec(dllexport)} you get almost identical semantics with
13160 identical syntax. This is a great boon to those working with
13161 cross-platform projects.
13163 For those adding visibility support to existing code, you may find
13164 @samp{#pragma GCC visibility} of use. This works by you enclosing
13165 the declarations you wish to set visibility for with (for example)
13166 @samp{#pragma GCC visibility push(hidden)} and
13167 @samp{#pragma GCC visibility pop}.
13168 Bear in mind that symbol visibility should be viewed @strong{as
13169 part of the API interface contract} and thus all new code should
13170 always specify visibility when it is not the default ie; declarations
13171 only for use within the local DSO should @strong{always} be marked explicitly
13172 as hidden as so to avoid PLT indirection overheads---making this
13173 abundantly clear also aids readability and self-documentation of the code.
13174 Note that due to ISO C++ specification requirements, operator new and
13175 operator delete must always be of default visibility.
13177 An overview of these techniques, their benefits and how to use them
13178 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13184 @node Environment Variables
13185 @section Environment Variables Affecting GCC
13186 @cindex environment variables
13188 @c man begin ENVIRONMENT
13189 This section describes several environment variables that affect how GCC
13190 operates. Some of them work by specifying directories or prefixes to use
13191 when searching for various kinds of files. Some are used to specify other
13192 aspects of the compilation environment.
13194 Note that you can also specify places to search using options such as
13195 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13196 take precedence over places specified using environment variables, which
13197 in turn take precedence over those specified by the configuration of GCC@.
13198 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13199 GNU Compiler Collection (GCC) Internals}.
13204 @c @itemx LC_COLLATE
13206 @c @itemx LC_MONETARY
13207 @c @itemx LC_NUMERIC
13212 @c @findex LC_COLLATE
13213 @findex LC_MESSAGES
13214 @c @findex LC_MONETARY
13215 @c @findex LC_NUMERIC
13219 These environment variables control the way that GCC uses
13220 localization information that allow GCC to work with different
13221 national conventions. GCC inspects the locale categories
13222 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13223 so. These locale categories can be set to any value supported by your
13224 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13225 Kingdom encoded in UTF-8.
13227 The @env{LC_CTYPE} environment variable specifies character
13228 classification. GCC uses it to determine the character boundaries in
13229 a string; this is needed for some multibyte encodings that contain quote
13230 and escape characters that would otherwise be interpreted as a string
13233 The @env{LC_MESSAGES} environment variable specifies the language to
13234 use in diagnostic messages.
13236 If the @env{LC_ALL} environment variable is set, it overrides the value
13237 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13238 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13239 environment variable. If none of these variables are set, GCC
13240 defaults to traditional C English behavior.
13244 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13245 files. GCC uses temporary files to hold the output of one stage of
13246 compilation which is to be used as input to the next stage: for example,
13247 the output of the preprocessor, which is the input to the compiler
13250 @item GCC_EXEC_PREFIX
13251 @findex GCC_EXEC_PREFIX
13252 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13253 names of the subprograms executed by the compiler. No slash is added
13254 when this prefix is combined with the name of a subprogram, but you can
13255 specify a prefix that ends with a slash if you wish.
13257 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13258 an appropriate prefix to use based on the pathname it was invoked with.
13260 If GCC cannot find the subprogram using the specified prefix, it
13261 tries looking in the usual places for the subprogram.
13263 The default value of @env{GCC_EXEC_PREFIX} is
13264 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13265 of @code{prefix} when you ran the @file{configure} script.
13267 Other prefixes specified with @option{-B} take precedence over this prefix.
13269 This prefix is also used for finding files such as @file{crt0.o} that are
13272 In addition, the prefix is used in an unusual way in finding the
13273 directories to search for header files. For each of the standard
13274 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13275 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13276 replacing that beginning with the specified prefix to produce an
13277 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13278 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13279 These alternate directories are searched first; the standard directories
13282 @item COMPILER_PATH
13283 @findex COMPILER_PATH
13284 The value of @env{COMPILER_PATH} is a colon-separated list of
13285 directories, much like @env{PATH}. GCC tries the directories thus
13286 specified when searching for subprograms, if it can't find the
13287 subprograms using @env{GCC_EXEC_PREFIX}.
13290 @findex LIBRARY_PATH
13291 The value of @env{LIBRARY_PATH} is a colon-separated list of
13292 directories, much like @env{PATH}. When configured as a native compiler,
13293 GCC tries the directories thus specified when searching for special
13294 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13295 using GCC also uses these directories when searching for ordinary
13296 libraries for the @option{-l} option (but directories specified with
13297 @option{-L} come first).
13301 @cindex locale definition
13302 This variable is used to pass locale information to the compiler. One way in
13303 which this information is used is to determine the character set to be used
13304 when character literals, string literals and comments are parsed in C and C++.
13305 When the compiler is configured to allow multibyte characters,
13306 the following values for @env{LANG} are recognized:
13310 Recognize JIS characters.
13312 Recognize SJIS characters.
13314 Recognize EUCJP characters.
13317 If @env{LANG} is not defined, or if it has some other value, then the
13318 compiler will use mblen and mbtowc as defined by the default locale to
13319 recognize and translate multibyte characters.
13323 Some additional environments variables affect the behavior of the
13326 @include cppenv.texi
13330 @node Precompiled Headers
13331 @section Using Precompiled Headers
13332 @cindex precompiled headers
13333 @cindex speed of compilation
13335 Often large projects have many header files that are included in every
13336 source file. The time the compiler takes to process these header files
13337 over and over again can account for nearly all of the time required to
13338 build the project. To make builds faster, GCC allows users to
13339 `precompile' a header file; then, if builds can use the precompiled
13340 header file they will be much faster.
13342 To create a precompiled header file, simply compile it as you would any
13343 other file, if necessary using the @option{-x} option to make the driver
13344 treat it as a C or C++ header file. You will probably want to use a
13345 tool like @command{make} to keep the precompiled header up-to-date when
13346 the headers it contains change.
13348 A precompiled header file will be searched for when @code{#include} is
13349 seen in the compilation. As it searches for the included file
13350 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13351 compiler looks for a precompiled header in each directory just before it
13352 looks for the include file in that directory. The name searched for is
13353 the name specified in the @code{#include} with @samp{.gch} appended. If
13354 the precompiled header file can't be used, it is ignored.
13356 For instance, if you have @code{#include "all.h"}, and you have
13357 @file{all.h.gch} in the same directory as @file{all.h}, then the
13358 precompiled header file will be used if possible, and the original
13359 header will be used otherwise.
13361 Alternatively, you might decide to put the precompiled header file in a
13362 directory and use @option{-I} to ensure that directory is searched
13363 before (or instead of) the directory containing the original header.
13364 Then, if you want to check that the precompiled header file is always
13365 used, you can put a file of the same name as the original header in this
13366 directory containing an @code{#error} command.
13368 This also works with @option{-include}. So yet another way to use
13369 precompiled headers, good for projects not designed with precompiled
13370 header files in mind, is to simply take most of the header files used by
13371 a project, include them from another header file, precompile that header
13372 file, and @option{-include} the precompiled header. If the header files
13373 have guards against multiple inclusion, they will be skipped because
13374 they've already been included (in the precompiled header).
13376 If you need to precompile the same header file for different
13377 languages, targets, or compiler options, you can instead make a
13378 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13379 header in the directory, perhaps using @option{-o}. It doesn't matter
13380 what you call the files in the directory, every precompiled header in
13381 the directory will be considered. The first precompiled header
13382 encountered in the directory that is valid for this compilation will
13383 be used; they're searched in no particular order.
13385 There are many other possibilities, limited only by your imagination,
13386 good sense, and the constraints of your build system.
13388 A precompiled header file can be used only when these conditions apply:
13392 Only one precompiled header can be used in a particular compilation.
13395 A precompiled header can't be used once the first C token is seen. You
13396 can have preprocessor directives before a precompiled header; you can
13397 even include a precompiled header from inside another header, so long as
13398 there are no C tokens before the @code{#include}.
13401 The precompiled header file must be produced for the same language as
13402 the current compilation. You can't use a C precompiled header for a C++
13406 The precompiled header file must have been produced by the same compiler
13407 binary as the current compilation is using.
13410 Any macros defined before the precompiled header is included must
13411 either be defined in the same way as when the precompiled header was
13412 generated, or must not affect the precompiled header, which usually
13413 means that they don't appear in the precompiled header at all.
13415 The @option{-D} option is one way to define a macro before a
13416 precompiled header is included; using a @code{#define} can also do it.
13417 There are also some options that define macros implicitly, like
13418 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13421 @item If debugging information is output when using the precompiled
13422 header, using @option{-g} or similar, the same kind of debugging information
13423 must have been output when building the precompiled header. However,
13424 a precompiled header built using @option{-g} can be used in a compilation
13425 when no debugging information is being output.
13427 @item The same @option{-m} options must generally be used when building
13428 and using the precompiled header. @xref{Submodel Options},
13429 for any cases where this rule is relaxed.
13431 @item Each of the following options must be the same when building and using
13432 the precompiled header:
13434 @gccoptlist{-fexceptions -funit-at-a-time}
13437 Some other command-line options starting with @option{-f},
13438 @option{-p}, or @option{-O} must be defined in the same way as when
13439 the precompiled header was generated. At present, it's not clear
13440 which options are safe to change and which are not; the safest choice
13441 is to use exactly the same options when generating and using the
13442 precompiled header. The following are known to be safe:
13444 @gccoptlist{-fmessage-length= -fpreprocessed
13445 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13446 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13451 For all of these except the last, the compiler will automatically
13452 ignore the precompiled header if the conditions aren't met. If you
13453 find an option combination that doesn't work and doesn't cause the
13454 precompiled header to be ignored, please consider filing a bug report,
13457 If you do use differing options when generating and using the
13458 precompiled header, the actual behavior will be a mixture of the
13459 behavior for the options. For instance, if you use @option{-g} to
13460 generate the precompiled header but not when using it, you may or may
13461 not get debugging information for routines in the precompiled header.
13463 @node Running Protoize
13464 @section Running Protoize
13466 The program @code{protoize} is an optional part of GCC@. You can use
13467 it to add prototypes to a program, thus converting the program to ISO
13468 C in one respect. The companion program @code{unprotoize} does the
13469 reverse: it removes argument types from any prototypes that are found.
13471 When you run these programs, you must specify a set of source files as
13472 command line arguments. The conversion programs start out by compiling
13473 these files to see what functions they define. The information gathered
13474 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13476 After scanning comes actual conversion. The specified files are all
13477 eligible to be converted; any files they include (whether sources or
13478 just headers) are eligible as well.
13480 But not all the eligible files are converted. By default,
13481 @code{protoize} and @code{unprotoize} convert only source and header
13482 files in the current directory. You can specify additional directories
13483 whose files should be converted with the @option{-d @var{directory}}
13484 option. You can also specify particular files to exclude with the
13485 @option{-x @var{file}} option. A file is converted if it is eligible, its
13486 directory name matches one of the specified directory names, and its
13487 name within the directory has not been excluded.
13489 Basic conversion with @code{protoize} consists of rewriting most
13490 function definitions and function declarations to specify the types of
13491 the arguments. The only ones not rewritten are those for varargs
13494 @code{protoize} optionally inserts prototype declarations at the
13495 beginning of the source file, to make them available for any calls that
13496 precede the function's definition. Or it can insert prototype
13497 declarations with block scope in the blocks where undeclared functions
13500 Basic conversion with @code{unprotoize} consists of rewriting most
13501 function declarations to remove any argument types, and rewriting
13502 function definitions to the old-style pre-ISO form.
13504 Both conversion programs print a warning for any function declaration or
13505 definition that they can't convert. You can suppress these warnings
13508 The output from @code{protoize} or @code{unprotoize} replaces the
13509 original source file. The original file is renamed to a name ending
13510 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13511 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13512 for DOS) file already exists, then the source file is simply discarded.
13514 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13515 scan the program and collect information about the functions it uses.
13516 So neither of these programs will work until GCC is installed.
13518 Here is a table of the options you can use with @code{protoize} and
13519 @code{unprotoize}. Each option works with both programs unless
13523 @item -B @var{directory}
13524 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13525 usual directory (normally @file{/usr/local/lib}). This file contains
13526 prototype information about standard system functions. This option
13527 applies only to @code{protoize}.
13529 @item -c @var{compilation-options}
13530 Use @var{compilation-options} as the options when running @command{gcc} to
13531 produce the @samp{.X} files. The special option @option{-aux-info} is
13532 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13534 Note that the compilation options must be given as a single argument to
13535 @code{protoize} or @code{unprotoize}. If you want to specify several
13536 @command{gcc} options, you must quote the entire set of compilation options
13537 to make them a single word in the shell.
13539 There are certain @command{gcc} arguments that you cannot use, because they
13540 would produce the wrong kind of output. These include @option{-g},
13541 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13542 the @var{compilation-options}, they are ignored.
13545 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13546 systems) instead of @samp{.c}. This is convenient if you are converting
13547 a C program to C++. This option applies only to @code{protoize}.
13550 Add explicit global declarations. This means inserting explicit
13551 declarations at the beginning of each source file for each function
13552 that is called in the file and was not declared. These declarations
13553 precede the first function definition that contains a call to an
13554 undeclared function. This option applies only to @code{protoize}.
13556 @item -i @var{string}
13557 Indent old-style parameter declarations with the string @var{string}.
13558 This option applies only to @code{protoize}.
13560 @code{unprotoize} converts prototyped function definitions to old-style
13561 function definitions, where the arguments are declared between the
13562 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13563 uses five spaces as the indentation. If you want to indent with just
13564 one space instead, use @option{-i " "}.
13567 Keep the @samp{.X} files. Normally, they are deleted after conversion
13571 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13572 a prototype declaration for each function in each block which calls the
13573 function without any declaration. This option applies only to
13577 Make no real changes. This mode just prints information about the conversions
13578 that would have been done without @option{-n}.
13581 Make no @samp{.save} files. The original files are simply deleted.
13582 Use this option with caution.
13584 @item -p @var{program}
13585 Use the program @var{program} as the compiler. Normally, the name
13586 @file{gcc} is used.
13589 Work quietly. Most warnings are suppressed.
13592 Print the version number, just like @option{-v} for @command{gcc}.
13595 If you need special compiler options to compile one of your program's
13596 source files, then you should generate that file's @samp{.X} file
13597 specially, by running @command{gcc} on that source file with the
13598 appropriate options and the option @option{-aux-info}. Then run
13599 @code{protoize} on the entire set of files. @code{protoize} will use
13600 the existing @samp{.X} file because it is newer than the source file.
13604 gcc -Dfoo=bar file1.c -aux-info file1.X
13609 You need to include the special files along with the rest in the
13610 @code{protoize} command, even though their @samp{.X} files already
13611 exist, because otherwise they won't get converted.
13613 @xref{Protoize Caveats}, for more information on how to use
13614 @code{protoize} successfully.