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-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 -fschedule-insns -fschedule-insns2 @gol
330 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
331 -fsched-spec-load-dangerous @gol
332 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
333 -fsched2-use-superblocks @gol
334 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
335 -fsignaling-nans -fsingle-precision-constant @gol
336 -fstack-protector -fstack-protector-all @gol
337 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
338 -funroll-all-loops -funroll-loops -fpeel-loops @gol
339 -fsplit-ivs-in-unroller -funswitch-loops @gol
340 -fvariable-expansion-in-unroller @gol
341 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
342 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
343 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
344 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
345 -ftree-vect-loop-version -ftree-salias -fweb @gol
346 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
347 --param @var{name}=@var{value}
348 -O -O0 -O1 -O2 -O3 -Os}
350 @item Preprocessor Options
351 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
352 @gccoptlist{-A@var{question}=@var{answer} @gol
353 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
354 -C -dD -dI -dM -dN @gol
355 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
356 -idirafter @var{dir} @gol
357 -include @var{file} -imacros @var{file} @gol
358 -iprefix @var{file} -iwithprefix @var{dir} @gol
359 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
360 -isysroot @var{dir} @gol
361 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
362 -P -fworking-directory -remap @gol
363 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
364 -Xpreprocessor @var{option}}
366 @item Assembler Option
367 @xref{Assembler Options,,Passing Options to the Assembler}.
368 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
371 @xref{Link Options,,Options for Linking}.
372 @gccoptlist{@var{object-file-name} -l@var{library} @gol
373 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
374 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
375 -Wl,@var{option} -Xlinker @var{option} @gol
378 @item Directory Options
379 @xref{Directory Options,,Options for Directory Search}.
380 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
381 -specs=@var{file} -I- --sysroot=@var{dir}}
384 @c I wrote this xref this way to avoid overfull hbox. -- rms
385 @xref{Target Options}.
386 @gccoptlist{-V @var{version} -b @var{machine}}
388 @item Machine Dependent Options
389 @xref{Submodel Options,,Hardware Models and Configurations}.
390 @c This list is ordered alphanumerically by subsection name.
391 @c Try and put the significant identifier (CPU or system) first,
392 @c so users have a clue at guessing where the ones they want will be.
395 @gccoptlist{-EB -EL @gol
396 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
397 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
400 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
401 -mabi=@var{name} @gol
402 -mapcs-stack-check -mno-apcs-stack-check @gol
403 -mapcs-float -mno-apcs-float @gol
404 -mapcs-reentrant -mno-apcs-reentrant @gol
405 -msched-prolog -mno-sched-prolog @gol
406 -mlittle-endian -mbig-endian -mwords-little-endian @gol
407 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
408 -mthumb-interwork -mno-thumb-interwork @gol
409 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
410 -mstructure-size-boundary=@var{n} @gol
411 -mabort-on-noreturn @gol
412 -mlong-calls -mno-long-calls @gol
413 -msingle-pic-base -mno-single-pic-base @gol
414 -mpic-register=@var{reg} @gol
415 -mnop-fun-dllimport @gol
416 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
417 -mpoke-function-name @gol
419 -mtpcs-frame -mtpcs-leaf-frame @gol
420 -mcaller-super-interworking -mcallee-super-interworking @gol
424 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
425 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
427 @emph{Blackfin Options}
428 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
429 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
430 -mlow-64k -mno-low64k -mid-shared-library @gol
431 -mno-id-shared-library -mshared-library-id=@var{n} @gol
432 -mlong-calls -mno-long-calls}
435 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
436 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
437 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
438 -mstack-align -mdata-align -mconst-align @gol
439 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
440 -melf -maout -melinux -mlinux -sim -sim2 @gol
441 -mmul-bug-workaround -mno-mul-bug-workaround}
444 @gccoptlist{-mmac -mpush-args}
446 @emph{Darwin Options}
447 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
448 -arch_only -bind_at_load -bundle -bundle_loader @gol
449 -client_name -compatibility_version -current_version @gol
451 -dependency-file -dylib_file -dylinker_install_name @gol
452 -dynamic -dynamiclib -exported_symbols_list @gol
453 -filelist -flat_namespace -force_cpusubtype_ALL @gol
454 -force_flat_namespace -headerpad_max_install_names @gol
455 -image_base -init -install_name -keep_private_externs @gol
456 -multi_module -multiply_defined -multiply_defined_unused @gol
457 -noall_load -no_dead_strip_inits_and_terms @gol
458 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
459 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
460 -private_bundle -read_only_relocs -sectalign @gol
461 -sectobjectsymbols -whyload -seg1addr @gol
462 -sectcreate -sectobjectsymbols -sectorder @gol
463 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
464 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
465 -segprot -segs_read_only_addr -segs_read_write_addr @gol
466 -single_module -static -sub_library -sub_umbrella @gol
467 -twolevel_namespace -umbrella -undefined @gol
468 -unexported_symbols_list -weak_reference_mismatches @gol
469 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
472 @emph{DEC Alpha Options}
473 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
474 -mieee -mieee-with-inexact -mieee-conformant @gol
475 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
476 -mtrap-precision=@var{mode} -mbuild-constants @gol
477 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
478 -mbwx -mmax -mfix -mcix @gol
479 -mfloat-vax -mfloat-ieee @gol
480 -mexplicit-relocs -msmall-data -mlarge-data @gol
481 -msmall-text -mlarge-text @gol
482 -mmemory-latency=@var{time}}
484 @emph{DEC Alpha/VMS Options}
485 @gccoptlist{-mvms-return-codes}
488 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
489 -mhard-float -msoft-float @gol
490 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
491 -mdouble -mno-double @gol
492 -mmedia -mno-media -mmuladd -mno-muladd @gol
493 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
494 -mlinked-fp -mlong-calls -malign-labels @gol
495 -mlibrary-pic -macc-4 -macc-8 @gol
496 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
497 -moptimize-membar -mno-optimize-membar @gol
498 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
499 -mvliw-branch -mno-vliw-branch @gol
500 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
501 -mno-nested-cond-exec -mtomcat-stats @gol
505 @emph{H8/300 Options}
506 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
509 @gccoptlist{-march=@var{architecture-type} @gol
510 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
511 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
512 -mfixed-range=@var{register-range} @gol
513 -mjump-in-delay -mlinker-opt -mlong-calls @gol
514 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
515 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
516 -mno-jump-in-delay -mno-long-load-store @gol
517 -mno-portable-runtime -mno-soft-float @gol
518 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
519 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
520 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
521 -munix=@var{unix-std} -nolibdld -static -threads}
523 @emph{i386 and x86-64 Options}
524 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
525 -mfpmath=@var{unit} @gol
526 -masm=@var{dialect} -mno-fancy-math-387 @gol
527 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
528 -mno-wide-multiply -mrtd -malign-double @gol
529 -mpreferred-stack-boundary=@var{num} @gol
530 -mmmx -msse -msse2 -msse3 -m3dnow @gol
531 -mthreads -mno-align-stringops -minline-all-stringops @gol
532 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
533 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
534 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
535 -mcmodel=@var{code-model} @gol
536 -m32 -m64 -mlarge-data-threshold=@var{num}}
539 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
540 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
541 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
542 -minline-float-divide-max-throughput @gol
543 -minline-int-divide-min-latency @gol
544 -minline-int-divide-max-throughput @gol
545 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
546 -mno-dwarf2-asm -mearly-stop-bits @gol
547 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
548 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
550 @emph{M32R/D Options}
551 @gccoptlist{-m32r2 -m32rx -m32r @gol
553 -malign-loops -mno-align-loops @gol
554 -missue-rate=@var{number} @gol
555 -mbranch-cost=@var{number} @gol
556 -mmodel=@var{code-size-model-type} @gol
557 -msdata=@var{sdata-type} @gol
558 -mno-flush-func -mflush-func=@var{name} @gol
559 -mno-flush-trap -mflush-trap=@var{number} @gol
563 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
565 @emph{M680x0 Options}
566 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
567 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
568 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
569 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
570 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
572 @emph{M68hc1x Options}
573 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
574 -mauto-incdec -minmax -mlong-calls -mshort @gol
575 -msoft-reg-count=@var{count}}
578 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
579 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
580 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
581 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
582 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
585 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
586 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
587 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
588 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
589 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
590 -mdsp -mpaired-single -mips3d @gol
591 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
592 -G@var{num} -membedded-data -mno-embedded-data @gol
593 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
594 -msplit-addresses -mno-split-addresses @gol
595 -mexplicit-relocs -mno-explicit-relocs @gol
596 -mcheck-zero-division -mno-check-zero-division @gol
597 -mdivide-traps -mdivide-breaks @gol
598 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
599 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
600 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
601 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
602 -mfix-sb1 -mno-fix-sb1 @gol
603 -mflush-func=@var{func} -mno-flush-func @gol
604 -mbranch-likely -mno-branch-likely @gol
605 -mfp-exceptions -mno-fp-exceptions @gol
606 -mvr4130-align -mno-vr4130-align}
609 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
610 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
611 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
612 -mno-base-addresses -msingle-exit -mno-single-exit}
614 @emph{MN10300 Options}
615 @gccoptlist{-mmult-bug -mno-mult-bug @gol
616 -mam33 -mno-am33 @gol
617 -mam33-2 -mno-am33-2 @gol
618 -mreturn-pointer-on-d0 @gol
622 @gccoptlist{-mno-crt0 -mbacc -msim @gol
623 -march=@var{cpu-type} }
625 @emph{PDP-11 Options}
626 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
627 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
628 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
629 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
630 -mbranch-expensive -mbranch-cheap @gol
631 -msplit -mno-split -munix-asm -mdec-asm}
633 @emph{PowerPC Options}
634 See RS/6000 and PowerPC Options.
636 @emph{RS/6000 and PowerPC Options}
637 @gccoptlist{-mcpu=@var{cpu-type} @gol
638 -mtune=@var{cpu-type} @gol
639 -mpower -mno-power -mpower2 -mno-power2 @gol
640 -mpowerpc -mpowerpc64 -mno-powerpc @gol
641 -maltivec -mno-altivec @gol
642 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
643 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
644 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
645 -mnew-mnemonics -mold-mnemonics @gol
646 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
647 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
648 -malign-power -malign-natural @gol
649 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
650 -mstring -mno-string -mupdate -mno-update @gol
651 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
652 -mstrict-align -mno-strict-align -mrelocatable @gol
653 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
654 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
655 -mdynamic-no-pic -maltivec -mswdiv @gol
656 -mprioritize-restricted-insns=@var{priority} @gol
657 -msched-costly-dep=@var{dependence_type} @gol
658 -minsert-sched-nops=@var{scheme} @gol
659 -mcall-sysv -mcall-netbsd @gol
660 -maix-struct-return -msvr4-struct-return @gol
661 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
662 -misel -mno-isel @gol
663 -misel=yes -misel=no @gol
665 -mspe=yes -mspe=no @gol
666 -mvrsave -mno-vrsave @gol
667 -mmulhw -mno-mulhw @gol
668 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
669 -mprototype -mno-prototype @gol
670 -msim -mmvme -mads -myellowknife -memb -msdata @gol
671 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
673 @emph{S/390 and zSeries Options}
674 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
675 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
676 -mpacked-stack -mno-packed-stack @gol
677 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
678 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
679 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
680 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
683 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
684 -m4-nofpu -m4-single-only -m4-single -m4 @gol
685 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
686 -m5-64media -m5-64media-nofpu @gol
687 -m5-32media -m5-32media-nofpu @gol
688 -m5-compact -m5-compact-nofpu @gol
689 -mb -ml -mdalign -mrelax @gol
690 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
691 -mieee -misize -mpadstruct -mspace @gol
692 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
693 -mdivsi3_libfunc=@var{name} @gol
694 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
698 @gccoptlist{-mcpu=@var{cpu-type} @gol
699 -mtune=@var{cpu-type} @gol
700 -mcmodel=@var{code-model} @gol
701 -m32 -m64 -mapp-regs -mno-app-regs @gol
702 -mfaster-structs -mno-faster-structs @gol
703 -mfpu -mno-fpu -mhard-float -msoft-float @gol
704 -mhard-quad-float -msoft-quad-float @gol
705 -mimpure-text -mno-impure-text -mlittle-endian @gol
706 -mstack-bias -mno-stack-bias @gol
707 -munaligned-doubles -mno-unaligned-doubles @gol
708 -mv8plus -mno-v8plus -mvis -mno-vis
711 @emph{System V Options}
712 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
714 @emph{TMS320C3x/C4x Options}
715 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
716 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
717 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
718 -mparallel-insns -mparallel-mpy -mpreserve-float}
721 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
722 -mprolog-function -mno-prolog-function -mspace @gol
723 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
724 -mapp-regs -mno-app-regs @gol
725 -mdisable-callt -mno-disable-callt @gol
731 @gccoptlist{-mg -mgnu -munix}
733 @emph{x86-64 Options}
734 See i386 and x86-64 Options.
736 @emph{Xstormy16 Options}
739 @emph{Xtensa Options}
740 @gccoptlist{-mconst16 -mno-const16 @gol
741 -mfused-madd -mno-fused-madd @gol
742 -mtext-section-literals -mno-text-section-literals @gol
743 -mtarget-align -mno-target-align @gol
744 -mlongcalls -mno-longcalls}
746 @emph{zSeries Options}
747 See S/390 and zSeries Options.
749 @item Code Generation Options
750 @xref{Code Gen Options,,Options for Code Generation Conventions}.
751 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
752 -ffixed-@var{reg} -fexceptions @gol
753 -fnon-call-exceptions -funwind-tables @gol
754 -fasynchronous-unwind-tables @gol
755 -finhibit-size-directive -finstrument-functions @gol
756 -fno-common -fno-ident @gol
757 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
758 -fno-jump-tables @gol
759 -freg-struct-return -fshared-data -fshort-enums @gol
760 -fshort-double -fshort-wchar @gol
761 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
762 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
763 -fargument-alias -fargument-noalias @gol
764 -fargument-noalias-global -fleading-underscore @gol
765 -ftls-model=@var{model} @gol
766 -ftrapv -fwrapv -fbounds-check @gol
771 * Overall Options:: Controlling the kind of output:
772 an executable, object files, assembler files,
773 or preprocessed source.
774 * C Dialect Options:: Controlling the variant of C language compiled.
775 * C++ Dialect Options:: Variations on C++.
776 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
778 * Language Independent Options:: Controlling how diagnostics should be
780 * Warning Options:: How picky should the compiler be?
781 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
782 * Optimize Options:: How much optimization?
783 * Preprocessor Options:: Controlling header files and macro definitions.
784 Also, getting dependency information for Make.
785 * Assembler Options:: Passing options to the assembler.
786 * Link Options:: Specifying libraries and so on.
787 * Directory Options:: Where to find header files and libraries.
788 Where to find the compiler executable files.
789 * Spec Files:: How to pass switches to sub-processes.
790 * Target Options:: Running a cross-compiler, or an old version of GCC.
793 @node Overall Options
794 @section Options Controlling the Kind of Output
796 Compilation can involve up to four stages: preprocessing, compilation
797 proper, assembly and linking, always in that order. GCC is capable of
798 preprocessing and compiling several files either into several
799 assembler input files, or into one assembler input file; then each
800 assembler input file produces an object file, and linking combines all
801 the object files (those newly compiled, and those specified as input)
802 into an executable file.
804 @cindex file name suffix
805 For any given input file, the file name suffix determines what kind of
810 C source code which must be preprocessed.
813 C source code which should not be preprocessed.
816 C++ source code which should not be preprocessed.
819 Objective-C source code. Note that you must link with the @file{libobjc}
820 library to make an Objective-C program work.
823 Objective-C source code which should not be preprocessed.
827 Objective-C++ source code. Note that you must link with the @file{libobjc}
828 library to make an Objective-C++ program work. Note that @samp{.M} refers
829 to a literal capital M@.
832 Objective-C++ source code which should not be preprocessed.
835 C, C++, Objective-C or Objective-C++ header file to be turned into a
840 @itemx @var{file}.cxx
841 @itemx @var{file}.cpp
842 @itemx @var{file}.CPP
843 @itemx @var{file}.c++
845 C++ source code which must be preprocessed. Note that in @samp{.cxx},
846 the last two letters must both be literally @samp{x}. Likewise,
847 @samp{.C} refers to a literal capital C@.
851 Objective-C++ source code which must be preprocessed.
854 Objective-C++ source code which should not be preprocessed.
858 C++ header file to be turned into a precompiled header.
861 @itemx @var{file}.for
862 @itemx @var{file}.FOR
863 Fixed form Fortran source code which should not be preprocessed.
866 @itemx @var{file}.fpp
867 @itemx @var{file}.FPP
868 Fixed form Fortran source code which must be preprocessed (with the traditional
872 @itemx @var{file}.f95
873 Free form Fortran source code which should not be preprocessed.
876 @itemx @var{file}.F95
877 Free form Fortran source code which must be preprocessed (with the
878 traditional preprocessor).
880 @c FIXME: Descriptions of Java file types.
887 Ada source code file which contains a library unit declaration (a
888 declaration of a package, subprogram, or generic, or a generic
889 instantiation), or a library unit renaming declaration (a package,
890 generic, or subprogram renaming declaration). Such files are also
893 @itemx @var{file}.adb
894 Ada source code file containing a library unit body (a subprogram or
895 package body). Such files are also called @dfn{bodies}.
897 @c GCC also knows about some suffixes for languages not yet included:
908 Assembler code which must be preprocessed.
911 An object file to be fed straight into linking.
912 Any file name with no recognized suffix is treated this way.
916 You can specify the input language explicitly with the @option{-x} option:
919 @item -x @var{language}
920 Specify explicitly the @var{language} for the following input files
921 (rather than letting the compiler choose a default based on the file
922 name suffix). This option applies to all following input files until
923 the next @option{-x} option. Possible values for @var{language} are:
925 c c-header c-cpp-output
926 c++ c++-header c++-cpp-output
927 objective-c objective-c-header objective-c-cpp-output
928 objective-c++ objective-c++-header objective-c++-cpp-output
929 assembler assembler-with-cpp
938 Turn off any specification of a language, so that subsequent files are
939 handled according to their file name suffixes (as they are if @option{-x}
940 has not been used at all).
942 @item -pass-exit-codes
943 @opindex pass-exit-codes
944 Normally the @command{gcc} program will exit with the code of 1 if any
945 phase of the compiler returns a non-success return code. If you specify
946 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
947 numerically highest error produced by any phase that returned an error
951 If you only want some of the stages of compilation, you can use
952 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
953 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
954 @command{gcc} is to stop. Note that some combinations (for example,
955 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
960 Compile or assemble the source files, but do not link. The linking
961 stage simply is not done. The ultimate output is in the form of an
962 object file for each source file.
964 By default, the object file name for a source file is made by replacing
965 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
967 Unrecognized input files, not requiring compilation or assembly, are
972 Stop after the stage of compilation proper; do not assemble. The output
973 is in the form of an assembler code file for each non-assembler input
976 By default, the assembler file name for a source file is made by
977 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
979 Input files that don't require compilation are ignored.
983 Stop after the preprocessing stage; do not run the compiler proper. The
984 output is in the form of preprocessed source code, which is sent to the
987 Input files which don't require preprocessing are ignored.
989 @cindex output file option
992 Place output in file @var{file}. This applies regardless to whatever
993 sort of output is being produced, whether it be an executable file,
994 an object file, an assembler file or preprocessed C code.
996 If @option{-o} is not specified, the default is to put an executable
997 file in @file{a.out}, the object file for
998 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
999 assembler file in @file{@var{source}.s}, a precompiled header file in
1000 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1005 Print (on standard error output) the commands executed to run the stages
1006 of compilation. Also print the version number of the compiler driver
1007 program and of the preprocessor and the compiler proper.
1011 Like @option{-v} except the commands are not executed and all command
1012 arguments are quoted. This is useful for shell scripts to capture the
1013 driver-generated command lines.
1017 Use pipes rather than temporary files for communication between the
1018 various stages of compilation. This fails to work on some systems where
1019 the assembler is unable to read from a pipe; but the GNU assembler has
1024 If you are compiling multiple source files, this option tells the driver
1025 to pass all the source files to the compiler at once (for those
1026 languages for which the compiler can handle this). This will allow
1027 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1028 language for which this is supported is C@. If you pass source files for
1029 multiple languages to the driver, using this option, the driver will invoke
1030 the compiler(s) that support IMA once each, passing each compiler all the
1031 source files appropriate for it. For those languages that do not support
1032 IMA this option will be ignored, and the compiler will be invoked once for
1033 each source file in that language. If you use this option in conjunction
1034 with @option{-save-temps}, the compiler will generate multiple
1036 (one for each source file), but only one (combined) @file{.o} or
1041 Print (on the standard output) a description of the command line options
1042 understood by @command{gcc}. If the @option{-v} option is also specified
1043 then @option{--help} will also be passed on to the various processes
1044 invoked by @command{gcc}, so that they can display the command line options
1045 they accept. If the @option{-Wextra} option is also specified then command
1046 line options which have no documentation associated with them will also
1050 @opindex target-help
1051 Print (on the standard output) a description of target specific command
1052 line options for each tool.
1056 Display the version number and copyrights of the invoked GCC@.
1058 @include @value{srcdir}/../libiberty/at-file.texi
1062 @section Compiling C++ Programs
1064 @cindex suffixes for C++ source
1065 @cindex C++ source file suffixes
1066 C++ source files conventionally use one of the suffixes @samp{.C},
1067 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1068 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1069 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1070 files with these names and compiles them as C++ programs even if you
1071 call the compiler the same way as for compiling C programs (usually
1072 with the name @command{gcc}).
1076 However, C++ programs often require class libraries as well as a
1077 compiler that understands the C++ language---and under some
1078 circumstances, you might want to compile programs or header files from
1079 standard input, or otherwise without a suffix that flags them as C++
1080 programs. You might also like to precompile a C header file with a
1081 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1082 program that calls GCC with the default language set to C++, and
1083 automatically specifies linking against the C++ library. On many
1084 systems, @command{g++} is also installed with the name @command{c++}.
1086 @cindex invoking @command{g++}
1087 When you compile C++ programs, you may specify many of the same
1088 command-line options that you use for compiling programs in any
1089 language; or command-line options meaningful for C and related
1090 languages; or options that are meaningful only for C++ programs.
1091 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1092 explanations of options for languages related to C@.
1093 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1094 explanations of options that are meaningful only for C++ programs.
1096 @node C Dialect Options
1097 @section Options Controlling C Dialect
1098 @cindex dialect options
1099 @cindex language dialect options
1100 @cindex options, dialect
1102 The following options control the dialect of C (or languages derived
1103 from C, such as C++, Objective-C and Objective-C++) that the compiler
1107 @cindex ANSI support
1111 In C mode, support all ISO C90 programs. In C++ mode,
1112 remove GNU extensions that conflict with ISO C++.
1114 This turns off certain features of GCC that are incompatible with ISO
1115 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1116 such as the @code{asm} and @code{typeof} keywords, and
1117 predefined macros such as @code{unix} and @code{vax} that identify the
1118 type of system you are using. It also enables the undesirable and
1119 rarely used ISO trigraph feature. For the C compiler,
1120 it disables recognition of C++ style @samp{//} comments as well as
1121 the @code{inline} keyword.
1123 The alternate keywords @code{__asm__}, @code{__extension__},
1124 @code{__inline__} and @code{__typeof__} continue to work despite
1125 @option{-ansi}. You would not want to use them in an ISO C program, of
1126 course, but it is useful to put them in header files that might be included
1127 in compilations done with @option{-ansi}. Alternate predefined macros
1128 such as @code{__unix__} and @code{__vax__} are also available, with or
1129 without @option{-ansi}.
1131 The @option{-ansi} option does not cause non-ISO programs to be
1132 rejected gratuitously. For that, @option{-pedantic} is required in
1133 addition to @option{-ansi}. @xref{Warning Options}.
1135 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1136 option is used. Some header files may notice this macro and refrain
1137 from declaring certain functions or defining certain macros that the
1138 ISO standard doesn't call for; this is to avoid interfering with any
1139 programs that might use these names for other things.
1141 Functions which would normally be built in but do not have semantics
1142 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1143 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1144 built-in functions provided by GCC}, for details of the functions
1149 Determine the language standard. This option is currently only
1150 supported when compiling C or C++. A value for this option must be
1151 provided; possible values are
1156 ISO C90 (same as @option{-ansi}).
1158 @item iso9899:199409
1159 ISO C90 as modified in amendment 1.
1165 ISO C99. Note that this standard is not yet fully supported; see
1166 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1167 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1170 Default, ISO C90 plus GNU extensions (including some C99 features).
1174 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1175 this will become the default. The name @samp{gnu9x} is deprecated.
1178 The 1998 ISO C++ standard plus amendments.
1181 The same as @option{-std=c++98} plus GNU extensions. This is the
1182 default for C++ code.
1185 Even when this option is not specified, you can still use some of the
1186 features of newer standards in so far as they do not conflict with
1187 previous C standards. For example, you may use @code{__restrict__} even
1188 when @option{-std=c99} is not specified.
1190 The @option{-std} options specifying some version of ISO C have the same
1191 effects as @option{-ansi}, except that features that were not in ISO C90
1192 but are in the specified version (for example, @samp{//} comments and
1193 the @code{inline} keyword in ISO C99) are not disabled.
1195 @xref{Standards,,Language Standards Supported by GCC}, for details of
1196 these standard versions.
1198 @item -aux-info @var{filename}
1200 Output to the given filename prototyped declarations for all functions
1201 declared and/or defined in a translation unit, including those in header
1202 files. This option is silently ignored in any language other than C@.
1204 Besides declarations, the file indicates, in comments, the origin of
1205 each declaration (source file and line), whether the declaration was
1206 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1207 @samp{O} for old, respectively, in the first character after the line
1208 number and the colon), and whether it came from a declaration or a
1209 definition (@samp{C} or @samp{F}, respectively, in the following
1210 character). In the case of function definitions, a K&R-style list of
1211 arguments followed by their declarations is also provided, inside
1212 comments, after the declaration.
1216 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1217 keyword, so that code can use these words as identifiers. You can use
1218 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1219 instead. @option{-ansi} implies @option{-fno-asm}.
1221 In C++, this switch only affects the @code{typeof} keyword, since
1222 @code{asm} and @code{inline} are standard keywords. You may want to
1223 use the @option{-fno-gnu-keywords} flag instead, which has the same
1224 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1225 switch only affects the @code{asm} and @code{typeof} keywords, since
1226 @code{inline} is a standard keyword in ISO C99.
1229 @itemx -fno-builtin-@var{function}
1230 @opindex fno-builtin
1231 @cindex built-in functions
1232 Don't recognize built-in functions that do not begin with
1233 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1234 functions provided by GCC}, for details of the functions affected,
1235 including those which are not built-in functions when @option{-ansi} or
1236 @option{-std} options for strict ISO C conformance are used because they
1237 do not have an ISO standard meaning.
1239 GCC normally generates special code to handle certain built-in functions
1240 more efficiently; for instance, calls to @code{alloca} may become single
1241 instructions that adjust the stack directly, and calls to @code{memcpy}
1242 may become inline copy loops. The resulting code is often both smaller
1243 and faster, but since the function calls no longer appear as such, you
1244 cannot set a breakpoint on those calls, nor can you change the behavior
1245 of the functions by linking with a different library. In addition,
1246 when a function is recognized as a built-in function, GCC may use
1247 information about that function to warn about problems with calls to
1248 that function, or to generate more efficient code, even if the
1249 resulting code still contains calls to that function. For example,
1250 warnings are given with @option{-Wformat} for bad calls to
1251 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1252 known not to modify global memory.
1254 With the @option{-fno-builtin-@var{function}} option
1255 only the built-in function @var{function} is
1256 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1257 function is named this is not built-in in this version of GCC, this
1258 option is ignored. There is no corresponding
1259 @option{-fbuiltin-@var{function}} option; if you wish to enable
1260 built-in functions selectively when using @option{-fno-builtin} or
1261 @option{-ffreestanding}, you may define macros such as:
1264 #define abs(n) __builtin_abs ((n))
1265 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1270 @cindex hosted environment
1272 Assert that compilation takes place in a hosted environment. This implies
1273 @option{-fbuiltin}. A hosted environment is one in which the
1274 entire standard library is available, and in which @code{main} has a return
1275 type of @code{int}. Examples are nearly everything except a kernel.
1276 This is equivalent to @option{-fno-freestanding}.
1278 @item -ffreestanding
1279 @opindex ffreestanding
1280 @cindex hosted environment
1282 Assert that compilation takes place in a freestanding environment. This
1283 implies @option{-fno-builtin}. A freestanding environment
1284 is one in which the standard library may not exist, and program startup may
1285 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1286 This is equivalent to @option{-fno-hosted}.
1288 @xref{Standards,,Language Standards Supported by GCC}, for details of
1289 freestanding and hosted environments.
1291 @item -fms-extensions
1292 @opindex fms-extensions
1293 Accept some non-standard constructs used in Microsoft header files.
1295 Some cases of unnamed fields in structures and unions are only
1296 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1297 fields within structs/unions}, for details.
1301 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1302 options for strict ISO C conformance) implies @option{-trigraphs}.
1304 @item -no-integrated-cpp
1305 @opindex no-integrated-cpp
1306 Performs a compilation in two passes: preprocessing and compiling. This
1307 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1308 @option{-B} option. The user supplied compilation step can then add in
1309 an additional preprocessing step after normal preprocessing but before
1310 compiling. The default is to use the integrated cpp (internal cpp)
1312 The semantics of this option will change if "cc1", "cc1plus", and
1313 "cc1obj" are merged.
1315 @cindex traditional C language
1316 @cindex C language, traditional
1318 @itemx -traditional-cpp
1319 @opindex traditional-cpp
1320 @opindex traditional
1321 Formerly, these options caused GCC to attempt to emulate a pre-standard
1322 C compiler. They are now only supported with the @option{-E} switch.
1323 The preprocessor continues to support a pre-standard mode. See the GNU
1324 CPP manual for details.
1326 @item -fcond-mismatch
1327 @opindex fcond-mismatch
1328 Allow conditional expressions with mismatched types in the second and
1329 third arguments. The value of such an expression is void. This option
1330 is not supported for C++.
1332 @item -funsigned-char
1333 @opindex funsigned-char
1334 Let the type @code{char} be unsigned, like @code{unsigned char}.
1336 Each kind of machine has a default for what @code{char} should
1337 be. It is either like @code{unsigned char} by default or like
1338 @code{signed char} by default.
1340 Ideally, a portable program should always use @code{signed char} or
1341 @code{unsigned char} when it depends on the signedness of an object.
1342 But many programs have been written to use plain @code{char} and
1343 expect it to be signed, or expect it to be unsigned, depending on the
1344 machines they were written for. This option, and its inverse, let you
1345 make such a program work with the opposite default.
1347 The type @code{char} is always a distinct type from each of
1348 @code{signed char} or @code{unsigned char}, even though its behavior
1349 is always just like one of those two.
1352 @opindex fsigned-char
1353 Let the type @code{char} be signed, like @code{signed char}.
1355 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1356 the negative form of @option{-funsigned-char}. Likewise, the option
1357 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1359 @item -fsigned-bitfields
1360 @itemx -funsigned-bitfields
1361 @itemx -fno-signed-bitfields
1362 @itemx -fno-unsigned-bitfields
1363 @opindex fsigned-bitfields
1364 @opindex funsigned-bitfields
1365 @opindex fno-signed-bitfields
1366 @opindex fno-unsigned-bitfields
1367 These options control whether a bit-field is signed or unsigned, when the
1368 declaration does not use either @code{signed} or @code{unsigned}. By
1369 default, such a bit-field is signed, because this is consistent: the
1370 basic integer types such as @code{int} are signed types.
1373 @node C++ Dialect Options
1374 @section Options Controlling C++ Dialect
1376 @cindex compiler options, C++
1377 @cindex C++ options, command line
1378 @cindex options, C++
1379 This section describes the command-line options that are only meaningful
1380 for C++ programs; but you can also use most of the GNU compiler options
1381 regardless of what language your program is in. For example, you
1382 might compile a file @code{firstClass.C} like this:
1385 g++ -g -frepo -O -c firstClass.C
1389 In this example, only @option{-frepo} is an option meant
1390 only for C++ programs; you can use the other options with any
1391 language supported by GCC@.
1393 Here is a list of options that are @emph{only} for compiling C++ programs:
1397 @item -fabi-version=@var{n}
1398 @opindex fabi-version
1399 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1400 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1401 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1402 the version that conforms most closely to the C++ ABI specification.
1403 Therefore, the ABI obtained using version 0 will change as ABI bugs
1406 The default is version 2.
1408 @item -fno-access-control
1409 @opindex fno-access-control
1410 Turn off all access checking. This switch is mainly useful for working
1411 around bugs in the access control code.
1415 Check that the pointer returned by @code{operator new} is non-null
1416 before attempting to modify the storage allocated. This check is
1417 normally unnecessary because the C++ standard specifies that
1418 @code{operator new} will only return @code{0} if it is declared
1419 @samp{throw()}, in which case the compiler will always check the
1420 return value even without this option. In all other cases, when
1421 @code{operator new} has a non-empty exception specification, memory
1422 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1423 @samp{new (nothrow)}.
1425 @item -fconserve-space
1426 @opindex fconserve-space
1427 Put uninitialized or runtime-initialized global variables into the
1428 common segment, as C does. This saves space in the executable at the
1429 cost of not diagnosing duplicate definitions. If you compile with this
1430 flag and your program mysteriously crashes after @code{main()} has
1431 completed, you may have an object that is being destroyed twice because
1432 two definitions were merged.
1434 This option is no longer useful on most targets, now that support has
1435 been added for putting variables into BSS without making them common.
1437 @item -ffriend-injection
1438 @opindex ffriend-injection
1439 Inject friend functions into the enclosing namespace, so that they are
1440 visible outside the scope of the class in which they are declared.
1441 Friend functions were documented to work this way in the old Annotated
1442 C++ Reference Manual, and versions of G++ before 4.1 always worked
1443 that way. However, in ISO C++ a friend function which is not declared
1444 in an enclosing scope can only be found using argument dependent
1445 lookup. This option causes friends to be injected as they were in
1448 This option is for compatibility, and may be removed in a future
1451 @item -fno-const-strings
1452 @opindex fno-const-strings
1453 Give string constants type @code{char *} instead of type @code{const
1454 char *}. By default, G++ uses type @code{const char *} as required by
1455 the standard. Even if you use @option{-fno-const-strings}, you cannot
1456 actually modify the value of a string constant.
1458 This option might be removed in a future release of G++. For maximum
1459 portability, you should structure your code so that it works with
1460 string constants that have type @code{const char *}.
1462 @item -fno-elide-constructors
1463 @opindex fno-elide-constructors
1464 The C++ standard allows an implementation to omit creating a temporary
1465 which is only used to initialize another object of the same type.
1466 Specifying this option disables that optimization, and forces G++ to
1467 call the copy constructor in all cases.
1469 @item -fno-enforce-eh-specs
1470 @opindex fno-enforce-eh-specs
1471 Don't generate code to check for violation of exception specifications
1472 at runtime. This option violates the C++ standard, but may be useful
1473 for reducing code size in production builds, much like defining
1474 @samp{NDEBUG}. This does not give user code permission to throw
1475 exceptions in violation of the exception specifications; the compiler
1476 will still optimize based on the specifications, so throwing an
1477 unexpected exception will result in undefined behavior.
1480 @itemx -fno-for-scope
1482 @opindex fno-for-scope
1483 If @option{-ffor-scope} is specified, the scope of variables declared in
1484 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1485 as specified by the C++ standard.
1486 If @option{-fno-for-scope} is specified, the scope of variables declared in
1487 a @i{for-init-statement} extends to the end of the enclosing scope,
1488 as was the case in old versions of G++, and other (traditional)
1489 implementations of C++.
1491 The default if neither flag is given to follow the standard,
1492 but to allow and give a warning for old-style code that would
1493 otherwise be invalid, or have different behavior.
1495 @item -fno-gnu-keywords
1496 @opindex fno-gnu-keywords
1497 Do not recognize @code{typeof} as a keyword, so that code can use this
1498 word as an identifier. You can use the keyword @code{__typeof__} instead.
1499 @option{-ansi} implies @option{-fno-gnu-keywords}.
1501 @item -fno-implicit-templates
1502 @opindex fno-implicit-templates
1503 Never emit code for non-inline templates which are instantiated
1504 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1505 @xref{Template Instantiation}, for more information.
1507 @item -fno-implicit-inline-templates
1508 @opindex fno-implicit-inline-templates
1509 Don't emit code for implicit instantiations of inline templates, either.
1510 The default is to handle inlines differently so that compiles with and
1511 without optimization will need the same set of explicit instantiations.
1513 @item -fno-implement-inlines
1514 @opindex fno-implement-inlines
1515 To save space, do not emit out-of-line copies of inline functions
1516 controlled by @samp{#pragma implementation}. This will cause linker
1517 errors if these functions are not inlined everywhere they are called.
1519 @item -fms-extensions
1520 @opindex fms-extensions
1521 Disable pedantic warnings about constructs used in MFC, such as implicit
1522 int and getting a pointer to member function via non-standard syntax.
1524 @item -fno-nonansi-builtins
1525 @opindex fno-nonansi-builtins
1526 Disable built-in declarations of functions that are not mandated by
1527 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1528 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1530 @item -fno-operator-names
1531 @opindex fno-operator-names
1532 Do not treat the operator name keywords @code{and}, @code{bitand},
1533 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1534 synonyms as keywords.
1536 @item -fno-optional-diags
1537 @opindex fno-optional-diags
1538 Disable diagnostics that the standard says a compiler does not need to
1539 issue. Currently, the only such diagnostic issued by G++ is the one for
1540 a name having multiple meanings within a class.
1543 @opindex fpermissive
1544 Downgrade some diagnostics about nonconformant code from errors to
1545 warnings. Thus, using @option{-fpermissive} will allow some
1546 nonconforming code to compile.
1550 Enable automatic template instantiation at link time. This option also
1551 implies @option{-fno-implicit-templates}. @xref{Template
1552 Instantiation}, for more information.
1556 Disable generation of information about every class with virtual
1557 functions for use by the C++ runtime type identification features
1558 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1559 of the language, you can save some space by using this flag. Note that
1560 exception handling uses the same information, but it will generate it as
1565 Emit statistics about front-end processing at the end of the compilation.
1566 This information is generally only useful to the G++ development team.
1568 @item -ftemplate-depth-@var{n}
1569 @opindex ftemplate-depth
1570 Set the maximum instantiation depth for template classes to @var{n}.
1571 A limit on the template instantiation depth is needed to detect
1572 endless recursions during template class instantiation. ANSI/ISO C++
1573 conforming programs must not rely on a maximum depth greater than 17.
1575 @item -fno-threadsafe-statics
1576 @opindex fno-threadsafe-statics
1577 Do not emit the extra code to use the routines specified in the C++
1578 ABI for thread-safe initialization of local statics. You can use this
1579 option to reduce code size slightly in code that doesn't need to be
1582 @item -fuse-cxa-atexit
1583 @opindex fuse-cxa-atexit
1584 Register destructors for objects with static storage duration with the
1585 @code{__cxa_atexit} function rather than the @code{atexit} function.
1586 This option is required for fully standards-compliant handling of static
1587 destructors, but will only work if your C library supports
1588 @code{__cxa_atexit}.
1590 @item -fvisibility-inlines-hidden
1591 @opindex fvisibility-inlines-hidden
1592 Causes all inlined methods to be marked with
1593 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1594 appear in the export table of a DSO and do not require a PLT indirection
1595 when used within the DSO@. Enabling this option can have a dramatic effect
1596 on load and link times of a DSO as it massively reduces the size of the
1597 dynamic export table when the library makes heavy use of templates. While
1598 it can cause bloating through duplication of code within each DSO where
1599 it is used, often the wastage is less than the considerable space occupied
1600 by a long symbol name in the export table which is typical when using
1601 templates and namespaces. For even more savings, combine with the
1602 @option{-fvisibility=hidden} switch.
1606 Do not use weak symbol support, even if it is provided by the linker.
1607 By default, G++ will use weak symbols if they are available. This
1608 option exists only for testing, and should not be used by end-users;
1609 it will result in inferior code and has no benefits. This option may
1610 be removed in a future release of G++.
1614 Do not search for header files in the standard directories specific to
1615 C++, but do still search the other standard directories. (This option
1616 is used when building the C++ library.)
1619 In addition, these optimization, warning, and code generation options
1620 have meanings only for C++ programs:
1623 @item -fno-default-inline
1624 @opindex fno-default-inline
1625 Do not assume @samp{inline} for functions defined inside a class scope.
1626 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1627 functions will have linkage like inline functions; they just won't be
1630 @item -Wabi @r{(C++ only)}
1632 Warn when G++ generates code that is probably not compatible with the
1633 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1634 all such cases, there are probably some cases that are not warned about,
1635 even though G++ is generating incompatible code. There may also be
1636 cases where warnings are emitted even though the code that is generated
1639 You should rewrite your code to avoid these warnings if you are
1640 concerned about the fact that code generated by G++ may not be binary
1641 compatible with code generated by other compilers.
1643 The known incompatibilities at this point include:
1648 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1649 pack data into the same byte as a base class. For example:
1652 struct A @{ virtual void f(); int f1 : 1; @};
1653 struct B : public A @{ int f2 : 1; @};
1657 In this case, G++ will place @code{B::f2} into the same byte
1658 as@code{A::f1}; other compilers will not. You can avoid this problem
1659 by explicitly padding @code{A} so that its size is a multiple of the
1660 byte size on your platform; that will cause G++ and other compilers to
1661 layout @code{B} identically.
1664 Incorrect handling of tail-padding for virtual bases. G++ does not use
1665 tail padding when laying out virtual bases. For example:
1668 struct A @{ virtual void f(); char c1; @};
1669 struct B @{ B(); char c2; @};
1670 struct C : public A, public virtual B @{@};
1674 In this case, G++ will not place @code{B} into the tail-padding for
1675 @code{A}; other compilers will. You can avoid this problem by
1676 explicitly padding @code{A} so that its size is a multiple of its
1677 alignment (ignoring virtual base classes); that will cause G++ and other
1678 compilers to layout @code{C} identically.
1681 Incorrect handling of bit-fields with declared widths greater than that
1682 of their underlying types, when the bit-fields appear in a union. For
1686 union U @{ int i : 4096; @};
1690 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1691 union too small by the number of bits in an @code{int}.
1694 Empty classes can be placed at incorrect offsets. For example:
1704 struct C : public B, public A @{@};
1708 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1709 it should be placed at offset zero. G++ mistakenly believes that the
1710 @code{A} data member of @code{B} is already at offset zero.
1713 Names of template functions whose types involve @code{typename} or
1714 template template parameters can be mangled incorrectly.
1717 template <typename Q>
1718 void f(typename Q::X) @{@}
1720 template <template <typename> class Q>
1721 void f(typename Q<int>::X) @{@}
1725 Instantiations of these templates may be mangled incorrectly.
1729 @item -Wctor-dtor-privacy @r{(C++ only)}
1730 @opindex Wctor-dtor-privacy
1731 Warn when a class seems unusable because all the constructors or
1732 destructors in that class are private, and it has neither friends nor
1733 public static member functions.
1735 @item -Wnon-virtual-dtor @r{(C++ only)}
1736 @opindex Wnon-virtual-dtor
1737 Warn when a class appears to be polymorphic, thereby requiring a virtual
1738 destructor, yet it declares a non-virtual one.
1739 This warning is enabled by @option{-Wall}.
1741 @item -Wreorder @r{(C++ only)}
1743 @cindex reordering, warning
1744 @cindex warning for reordering of member initializers
1745 Warn when the order of member initializers given in the code does not
1746 match the order in which they must be executed. For instance:
1752 A(): j (0), i (1) @{ @}
1756 The compiler will rearrange the member initializers for @samp{i}
1757 and @samp{j} to match the declaration order of the members, emitting
1758 a warning to that effect. This warning is enabled by @option{-Wall}.
1761 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1764 @item -Weffc++ @r{(C++ only)}
1766 Warn about violations of the following style guidelines from Scott Meyers'
1767 @cite{Effective C++} book:
1771 Item 11: Define a copy constructor and an assignment operator for classes
1772 with dynamically allocated memory.
1775 Item 12: Prefer initialization to assignment in constructors.
1778 Item 14: Make destructors virtual in base classes.
1781 Item 15: Have @code{operator=} return a reference to @code{*this}.
1784 Item 23: Don't try to return a reference when you must return an object.
1788 Also warn about violations of the following style guidelines from
1789 Scott Meyers' @cite{More Effective C++} book:
1793 Item 6: Distinguish between prefix and postfix forms of increment and
1794 decrement operators.
1797 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1801 When selecting this option, be aware that the standard library
1802 headers do not obey all of these guidelines; use @samp{grep -v}
1803 to filter out those warnings.
1805 @item -Wno-deprecated @r{(C++ only)}
1806 @opindex Wno-deprecated
1807 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1809 @item -Wstrict-null-sentinel @r{(C++ only)}
1810 @opindex Wstrict-null-sentinel
1811 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1812 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1813 to @code{__null}. Although it is a null pointer constant not a null pointer,
1814 it is guaranteed to of the same size as a pointer. But this use is
1815 not portable across different compilers.
1817 @item -Wno-non-template-friend @r{(C++ only)}
1818 @opindex Wno-non-template-friend
1819 Disable warnings when non-templatized friend functions are declared
1820 within a template. Since the advent of explicit template specification
1821 support in G++, if the name of the friend is an unqualified-id (i.e.,
1822 @samp{friend foo(int)}), the C++ language specification demands that the
1823 friend declare or define an ordinary, nontemplate function. (Section
1824 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1825 could be interpreted as a particular specialization of a templatized
1826 function. Because this non-conforming behavior is no longer the default
1827 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1828 check existing code for potential trouble spots and is on by default.
1829 This new compiler behavior can be turned off with
1830 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1831 but disables the helpful warning.
1833 @item -Wold-style-cast @r{(C++ only)}
1834 @opindex Wold-style-cast
1835 Warn if an old-style (C-style) cast to a non-void type is used within
1836 a C++ program. The new-style casts (@samp{dynamic_cast},
1837 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1838 less vulnerable to unintended effects and much easier to search for.
1840 @item -Woverloaded-virtual @r{(C++ only)}
1841 @opindex Woverloaded-virtual
1842 @cindex overloaded virtual fn, warning
1843 @cindex warning for overloaded virtual fn
1844 Warn when a function declaration hides virtual functions from a
1845 base class. For example, in:
1852 struct B: public A @{
1857 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1865 will fail to compile.
1867 @item -Wno-pmf-conversions @r{(C++ only)}
1868 @opindex Wno-pmf-conversions
1869 Disable the diagnostic for converting a bound pointer to member function
1872 @item -Wsign-promo @r{(C++ only)}
1873 @opindex Wsign-promo
1874 Warn when overload resolution chooses a promotion from unsigned or
1875 enumerated type to a signed type, over a conversion to an unsigned type of
1876 the same size. Previous versions of G++ would try to preserve
1877 unsignedness, but the standard mandates the current behavior.
1882 A& operator = (int);
1892 In this example, G++ will synthesize a default @samp{A& operator =
1893 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1896 @node Objective-C and Objective-C++ Dialect Options
1897 @section Options Controlling Objective-C and Objective-C++ Dialects
1899 @cindex compiler options, Objective-C and Objective-C++
1900 @cindex Objective-C and Objective-C++ options, command line
1901 @cindex options, Objective-C and Objective-C++
1902 (NOTE: This manual does not describe the Objective-C and Objective-C++
1903 languages themselves. See @xref{Standards,,Language Standards
1904 Supported by GCC}, for references.)
1906 This section describes the command-line options that are only meaningful
1907 for Objective-C and Objective-C++ programs, but you can also use most of
1908 the language-independent GNU compiler options.
1909 For example, you might compile a file @code{some_class.m} like this:
1912 gcc -g -fgnu-runtime -O -c some_class.m
1916 In this example, @option{-fgnu-runtime} is an option meant only for
1917 Objective-C and Objective-C++ programs; you can use the other options with
1918 any language supported by GCC@.
1920 Note that since Objective-C is an extension of the C language, Objective-C
1921 compilations may also use options specific to the C front-end (e.g.,
1922 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1923 C++-specific options (e.g., @option{-Wabi}).
1925 Here is a list of options that are @emph{only} for compiling Objective-C
1926 and Objective-C++ programs:
1929 @item -fconstant-string-class=@var{class-name}
1930 @opindex fconstant-string-class
1931 Use @var{class-name} as the name of the class to instantiate for each
1932 literal string specified with the syntax @code{@@"@dots{}"}. The default
1933 class name is @code{NXConstantString} if the GNU runtime is being used, and
1934 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1935 @option{-fconstant-cfstrings} option, if also present, will override the
1936 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1937 to be laid out as constant CoreFoundation strings.
1940 @opindex fgnu-runtime
1941 Generate object code compatible with the standard GNU Objective-C
1942 runtime. This is the default for most types of systems.
1944 @item -fnext-runtime
1945 @opindex fnext-runtime
1946 Generate output compatible with the NeXT runtime. This is the default
1947 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1948 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1951 @item -fno-nil-receivers
1952 @opindex fno-nil-receivers
1953 Assume that all Objective-C message dispatches (e.g.,
1954 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1955 is not @code{nil}. This allows for more efficient entry points in the runtime
1956 to be used. Currently, this option is only available in conjunction with
1957 the NeXT runtime on Mac OS X 10.3 and later.
1959 @item -fobjc-call-cxx-cdtors
1960 @opindex fobjc-call-cxx-cdtors
1961 For each Objective-C class, check if any of its instance variables is a
1962 C++ object with a non-trivial default constructor. If so, synthesize a
1963 special @code{- (id) .cxx_construct} instance method that will run
1964 non-trivial default constructors on any such instance variables, in order,
1965 and then return @code{self}. Similarly, check if any instance variable
1966 is a C++ object with a non-trivial destructor, and if so, synthesize a
1967 special @code{- (void) .cxx_destruct} method that will run
1968 all such default destructors, in reverse order.
1970 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1971 thusly generated will only operate on instance variables declared in the
1972 current Objective-C class, and not those inherited from superclasses. It
1973 is the responsibility of the Objective-C runtime to invoke all such methods
1974 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1975 will be invoked by the runtime immediately after a new object
1976 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1977 be invoked immediately before the runtime deallocates an object instance.
1979 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1980 support for invoking the @code{- (id) .cxx_construct} and
1981 @code{- (void) .cxx_destruct} methods.
1983 @item -fobjc-direct-dispatch
1984 @opindex fobjc-direct-dispatch
1985 Allow fast jumps to the message dispatcher. On Darwin this is
1986 accomplished via the comm page.
1988 @item -fobjc-exceptions
1989 @opindex fobjc-exceptions
1990 Enable syntactic support for structured exception handling in Objective-C,
1991 similar to what is offered by C++ and Java. Currently, this option is only
1992 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2000 @@catch (AnObjCClass *exc) @{
2007 @@catch (AnotherClass *exc) @{
2010 @@catch (id allOthers) @{
2020 The @code{@@throw} statement may appear anywhere in an Objective-C or
2021 Objective-C++ program; when used inside of a @code{@@catch} block, the
2022 @code{@@throw} may appear without an argument (as shown above), in which case
2023 the object caught by the @code{@@catch} will be rethrown.
2025 Note that only (pointers to) Objective-C objects may be thrown and
2026 caught using this scheme. When an object is thrown, it will be caught
2027 by the nearest @code{@@catch} clause capable of handling objects of that type,
2028 analogously to how @code{catch} blocks work in C++ and Java. A
2029 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2030 any and all Objective-C exceptions not caught by previous @code{@@catch}
2033 The @code{@@finally} clause, if present, will be executed upon exit from the
2034 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2035 regardless of whether any exceptions are thrown, caught or rethrown
2036 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2037 of the @code{finally} clause in Java.
2039 There are several caveats to using the new exception mechanism:
2043 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2044 idioms provided by the @code{NSException} class, the new
2045 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2046 systems, due to additional functionality needed in the (NeXT) Objective-C
2050 As mentioned above, the new exceptions do not support handling
2051 types other than Objective-C objects. Furthermore, when used from
2052 Objective-C++, the Objective-C exception model does not interoperate with C++
2053 exceptions at this time. This means you cannot @code{@@throw} an exception
2054 from Objective-C and @code{catch} it in C++, or vice versa
2055 (i.e., @code{throw @dots{} @@catch}).
2058 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2059 blocks for thread-safe execution:
2062 @@synchronized (ObjCClass *guard) @{
2067 Upon entering the @code{@@synchronized} block, a thread of execution shall
2068 first check whether a lock has been placed on the corresponding @code{guard}
2069 object by another thread. If it has, the current thread shall wait until
2070 the other thread relinquishes its lock. Once @code{guard} becomes available,
2071 the current thread will place its own lock on it, execute the code contained in
2072 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2073 making @code{guard} available to other threads).
2075 Unlike Java, Objective-C does not allow for entire methods to be marked
2076 @code{@@synchronized}. Note that throwing exceptions out of
2077 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2078 to be unlocked properly.
2082 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2084 @item -freplace-objc-classes
2085 @opindex freplace-objc-classes
2086 Emit a special marker instructing @command{ld(1)} not to statically link in
2087 the resulting object file, and allow @command{dyld(1)} to load it in at
2088 run time instead. This is used in conjunction with the Fix-and-Continue
2089 debugging mode, where the object file in question may be recompiled and
2090 dynamically reloaded in the course of program execution, without the need
2091 to restart the program itself. Currently, Fix-and-Continue functionality
2092 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2097 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2098 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2099 compile time) with static class references that get initialized at load time,
2100 which improves run-time performance. Specifying the @option{-fzero-link} flag
2101 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2102 to be retained. This is useful in Zero-Link debugging mode, since it allows
2103 for individual class implementations to be modified during program execution.
2107 Dump interface declarations for all classes seen in the source file to a
2108 file named @file{@var{sourcename}.decl}.
2110 @item -Wassign-intercept
2111 @opindex Wassign-intercept
2112 Warn whenever an Objective-C assignment is being intercepted by the
2116 @opindex Wno-protocol
2117 If a class is declared to implement a protocol, a warning is issued for
2118 every method in the protocol that is not implemented by the class. The
2119 default behavior is to issue a warning for every method not explicitly
2120 implemented in the class, even if a method implementation is inherited
2121 from the superclass. If you use the @option{-Wno-protocol} option, then
2122 methods inherited from the superclass are considered to be implemented,
2123 and no warning is issued for them.
2127 Warn if multiple methods of different types for the same selector are
2128 found during compilation. The check is performed on the list of methods
2129 in the final stage of compilation. Additionally, a check is performed
2130 for each selector appearing in a @code{@@selector(@dots{})}
2131 expression, and a corresponding method for that selector has been found
2132 during compilation. Because these checks scan the method table only at
2133 the end of compilation, these warnings are not produced if the final
2134 stage of compilation is not reached, for example because an error is
2135 found during compilation, or because the @option{-fsyntax-only} option is
2138 @item -Wstrict-selector-match
2139 @opindex Wstrict-selector-match
2140 Warn if multiple methods with differing argument and/or return types are
2141 found for a given selector when attempting to send a message using this
2142 selector to a receiver of type @code{id} or @code{Class}. When this flag
2143 is off (which is the default behavior), the compiler will omit such warnings
2144 if any differences found are confined to types which share the same size
2147 @item -Wundeclared-selector
2148 @opindex Wundeclared-selector
2149 Warn if a @code{@@selector(@dots{})} expression referring to an
2150 undeclared selector is found. A selector is considered undeclared if no
2151 method with that name has been declared before the
2152 @code{@@selector(@dots{})} expression, either explicitly in an
2153 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2154 an @code{@@implementation} section. This option always performs its
2155 checks as soon as a @code{@@selector(@dots{})} expression is found,
2156 while @option{-Wselector} only performs its checks in the final stage of
2157 compilation. This also enforces the coding style convention
2158 that methods and selectors must be declared before being used.
2160 @item -print-objc-runtime-info
2161 @opindex print-objc-runtime-info
2162 Generate C header describing the largest structure that is passed by
2167 @node Language Independent Options
2168 @section Options to Control Diagnostic Messages Formatting
2169 @cindex options to control diagnostics formatting
2170 @cindex diagnostic messages
2171 @cindex message formatting
2173 Traditionally, diagnostic messages have been formatted irrespective of
2174 the output device's aspect (e.g.@: its width, @dots{}). The options described
2175 below can be used to control the diagnostic messages formatting
2176 algorithm, e.g.@: how many characters per line, how often source location
2177 information should be reported. Right now, only the C++ front end can
2178 honor these options. However it is expected, in the near future, that
2179 the remaining front ends would be able to digest them correctly.
2182 @item -fmessage-length=@var{n}
2183 @opindex fmessage-length
2184 Try to format error messages so that they fit on lines of about @var{n}
2185 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2186 the front ends supported by GCC@. If @var{n} is zero, then no
2187 line-wrapping will be done; each error message will appear on a single
2190 @opindex fdiagnostics-show-location
2191 @item -fdiagnostics-show-location=once
2192 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2193 reporter to emit @emph{once} source location information; that is, in
2194 case the message is too long to fit on a single physical line and has to
2195 be wrapped, the source location won't be emitted (as prefix) again,
2196 over and over, in subsequent continuation lines. This is the default
2199 @item -fdiagnostics-show-location=every-line
2200 Only meaningful in line-wrapping mode. Instructs the diagnostic
2201 messages reporter to emit the same source location information (as
2202 prefix) for physical lines that result from the process of breaking
2203 a message which is too long to fit on a single line.
2205 @item -fdiagnostics-show-options
2206 @opindex fdiagnostics-show-options
2207 This option instructs the diagnostic machinery to add text to each
2208 diagnostic emitted, which indicates which command line option directly
2209 controls that diagnostic, when such an option is known to the
2210 diagnostic machinery.
2214 @node Warning Options
2215 @section Options to Request or Suppress Warnings
2216 @cindex options to control warnings
2217 @cindex warning messages
2218 @cindex messages, warning
2219 @cindex suppressing warnings
2221 Warnings are diagnostic messages that report constructions which
2222 are not inherently erroneous but which are risky or suggest there
2223 may have been an error.
2225 You can request many specific warnings with options beginning @samp{-W},
2226 for example @option{-Wimplicit} to request warnings on implicit
2227 declarations. Each of these specific warning options also has a
2228 negative form beginning @samp{-Wno-} to turn off warnings;
2229 for example, @option{-Wno-implicit}. This manual lists only one of the
2230 two forms, whichever is not the default.
2232 The following options control the amount and kinds of warnings produced
2233 by GCC; for further, language-specific options also refer to
2234 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2238 @cindex syntax checking
2240 @opindex fsyntax-only
2241 Check the code for syntax errors, but don't do anything beyond that.
2245 Issue all the warnings demanded by strict ISO C and ISO C++;
2246 reject all programs that use forbidden extensions, and some other
2247 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2248 version of the ISO C standard specified by any @option{-std} option used.
2250 Valid ISO C and ISO C++ programs should compile properly with or without
2251 this option (though a rare few will require @option{-ansi} or a
2252 @option{-std} option specifying the required version of ISO C)@. However,
2253 without this option, certain GNU extensions and traditional C and C++
2254 features are supported as well. With this option, they are rejected.
2256 @option{-pedantic} does not cause warning messages for use of the
2257 alternate keywords whose names begin and end with @samp{__}. Pedantic
2258 warnings are also disabled in the expression that follows
2259 @code{__extension__}. However, only system header files should use
2260 these escape routes; application programs should avoid them.
2261 @xref{Alternate Keywords}.
2263 Some users try to use @option{-pedantic} to check programs for strict ISO
2264 C conformance. They soon find that it does not do quite what they want:
2265 it finds some non-ISO practices, but not all---only those for which
2266 ISO C @emph{requires} a diagnostic, and some others for which
2267 diagnostics have been added.
2269 A feature to report any failure to conform to ISO C might be useful in
2270 some instances, but would require considerable additional work and would
2271 be quite different from @option{-pedantic}. We don't have plans to
2272 support such a feature in the near future.
2274 Where the standard specified with @option{-std} represents a GNU
2275 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2276 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2277 extended dialect is based. Warnings from @option{-pedantic} are given
2278 where they are required by the base standard. (It would not make sense
2279 for such warnings to be given only for features not in the specified GNU
2280 C dialect, since by definition the GNU dialects of C include all
2281 features the compiler supports with the given option, and there would be
2282 nothing to warn about.)
2284 @item -pedantic-errors
2285 @opindex pedantic-errors
2286 Like @option{-pedantic}, except that errors are produced rather than
2291 Inhibit all warning messages.
2295 Inhibit warning messages about the use of @samp{#import}.
2297 @item -Wchar-subscripts
2298 @opindex Wchar-subscripts
2299 Warn if an array subscript has type @code{char}. This is a common cause
2300 of error, as programmers often forget that this type is signed on some
2302 This warning is enabled by @option{-Wall}.
2306 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2307 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2308 This warning is enabled by @option{-Wall}.
2310 @item -Wfatal-errors
2311 @opindex Wfatal-errors
2312 This option causes the compiler to abort compilation on the first error
2313 occurred rather than trying to keep going and printing further error
2318 @opindex ffreestanding
2319 @opindex fno-builtin
2320 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2321 the arguments supplied have types appropriate to the format string
2322 specified, and that the conversions specified in the format string make
2323 sense. This includes standard functions, and others specified by format
2324 attributes (@pxref{Function Attributes}), in the @code{printf},
2325 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2326 not in the C standard) families (or other target-specific families).
2327 Which functions are checked without format attributes having been
2328 specified depends on the standard version selected, and such checks of
2329 functions without the attribute specified are disabled by
2330 @option{-ffreestanding} or @option{-fno-builtin}.
2332 The formats are checked against the format features supported by GNU
2333 libc version 2.2. These include all ISO C90 and C99 features, as well
2334 as features from the Single Unix Specification and some BSD and GNU
2335 extensions. Other library implementations may not support all these
2336 features; GCC does not support warning about features that go beyond a
2337 particular library's limitations. However, if @option{-pedantic} is used
2338 with @option{-Wformat}, warnings will be given about format features not
2339 in the selected standard version (but not for @code{strfmon} formats,
2340 since those are not in any version of the C standard). @xref{C Dialect
2341 Options,,Options Controlling C Dialect}.
2343 Since @option{-Wformat} also checks for null format arguments for
2344 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2346 @option{-Wformat} is included in @option{-Wall}. For more control over some
2347 aspects of format checking, the options @option{-Wformat-y2k},
2348 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2349 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2350 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2353 @opindex Wformat-y2k
2354 If @option{-Wformat} is specified, also warn about @code{strftime}
2355 formats which may yield only a two-digit year.
2357 @item -Wno-format-extra-args
2358 @opindex Wno-format-extra-args
2359 If @option{-Wformat} is specified, do not warn about excess arguments to a
2360 @code{printf} or @code{scanf} format function. The C standard specifies
2361 that such arguments are ignored.
2363 Where the unused arguments lie between used arguments that are
2364 specified with @samp{$} operand number specifications, normally
2365 warnings are still given, since the implementation could not know what
2366 type to pass to @code{va_arg} to skip the unused arguments. However,
2367 in the case of @code{scanf} formats, this option will suppress the
2368 warning if the unused arguments are all pointers, since the Single
2369 Unix Specification says that such unused arguments are allowed.
2371 @item -Wno-format-zero-length
2372 @opindex Wno-format-zero-length
2373 If @option{-Wformat} is specified, do not warn about zero-length formats.
2374 The C standard specifies that zero-length formats are allowed.
2376 @item -Wformat-nonliteral
2377 @opindex Wformat-nonliteral
2378 If @option{-Wformat} is specified, also warn if the format string is not a
2379 string literal and so cannot be checked, unless the format function
2380 takes its format arguments as a @code{va_list}.
2382 @item -Wformat-security
2383 @opindex Wformat-security
2384 If @option{-Wformat} is specified, also warn about uses of format
2385 functions that represent possible security problems. At present, this
2386 warns about calls to @code{printf} and @code{scanf} functions where the
2387 format string is not a string literal and there are no format arguments,
2388 as in @code{printf (foo);}. This may be a security hole if the format
2389 string came from untrusted input and contains @samp{%n}. (This is
2390 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2391 in future warnings may be added to @option{-Wformat-security} that are not
2392 included in @option{-Wformat-nonliteral}.)
2396 Enable @option{-Wformat} plus format checks not included in
2397 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2398 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2402 Warn about passing a null pointer for arguments marked as
2403 requiring a non-null value by the @code{nonnull} function attribute.
2405 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2406 can be disabled with the @option{-Wno-nonnull} option.
2408 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2410 Warn about uninitialized variables which are initialized with themselves.
2411 Note this option can only be used with the @option{-Wuninitialized} option,
2412 which in turn only works with @option{-O1} and above.
2414 For example, GCC will warn about @code{i} being uninitialized in the
2415 following snippet only when @option{-Winit-self} has been specified:
2426 @item -Wimplicit-int
2427 @opindex Wimplicit-int
2428 Warn when a declaration does not specify a type.
2429 This warning is enabled by @option{-Wall}.
2431 @item -Wimplicit-function-declaration
2432 @itemx -Werror-implicit-function-declaration
2433 @opindex Wimplicit-function-declaration
2434 @opindex Werror-implicit-function-declaration
2435 Give a warning (or error) whenever a function is used before being
2436 declared. The form @option{-Wno-error-implicit-function-declaration}
2438 This warning is enabled by @option{-Wall} (as a warning, not an error).
2442 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2443 This warning is enabled by @option{-Wall}.
2447 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2448 function with external linkage, returning int, taking either zero
2449 arguments, two, or three arguments of appropriate types.
2450 This warning is enabled by @option{-Wall}.
2452 @item -Wmissing-braces
2453 @opindex Wmissing-braces
2454 Warn if an aggregate or union initializer is not fully bracketed. In
2455 the following example, the initializer for @samp{a} is not fully
2456 bracketed, but that for @samp{b} is fully bracketed.
2459 int a[2][2] = @{ 0, 1, 2, 3 @};
2460 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2463 This warning is enabled by @option{-Wall}.
2465 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2466 @opindex Wmissing-include-dirs
2467 Warn if a user-supplied include directory does not exist.
2470 @opindex Wparentheses
2471 Warn if parentheses are omitted in certain contexts, such
2472 as when there is an assignment in a context where a truth value
2473 is expected, or when operators are nested whose precedence people
2474 often get confused about. Only the warning for an assignment used as
2475 a truth value is supported when compiling C++; the other warnings are
2476 only supported when compiling C@.
2478 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2479 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2480 interpretation from that of ordinary mathematical notation.
2482 Also warn about constructions where there may be confusion to which
2483 @code{if} statement an @code{else} branch belongs. Here is an example of
2498 In C, every @code{else} branch belongs to the innermost possible @code{if}
2499 statement, which in this example is @code{if (b)}. This is often not
2500 what the programmer expected, as illustrated in the above example by
2501 indentation the programmer chose. When there is the potential for this
2502 confusion, GCC will issue a warning when this flag is specified.
2503 To eliminate the warning, add explicit braces around the innermost
2504 @code{if} statement so there is no way the @code{else} could belong to
2505 the enclosing @code{if}. The resulting code would look like this:
2521 This warning is enabled by @option{-Wall}.
2523 @item -Wsequence-point
2524 @opindex Wsequence-point
2525 Warn about code that may have undefined semantics because of violations
2526 of sequence point rules in the C standard.
2528 The C standard defines the order in which expressions in a C program are
2529 evaluated in terms of @dfn{sequence points}, which represent a partial
2530 ordering between the execution of parts of the program: those executed
2531 before the sequence point, and those executed after it. These occur
2532 after the evaluation of a full expression (one which is not part of a
2533 larger expression), after the evaluation of the first operand of a
2534 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2535 function is called (but after the evaluation of its arguments and the
2536 expression denoting the called function), and in certain other places.
2537 Other than as expressed by the sequence point rules, the order of
2538 evaluation of subexpressions of an expression is not specified. All
2539 these rules describe only a partial order rather than a total order,
2540 since, for example, if two functions are called within one expression
2541 with no sequence point between them, the order in which the functions
2542 are called is not specified. However, the standards committee have
2543 ruled that function calls do not overlap.
2545 It is not specified when between sequence points modifications to the
2546 values of objects take effect. Programs whose behavior depends on this
2547 have undefined behavior; the C standard specifies that ``Between the
2548 previous and next sequence point an object shall have its stored value
2549 modified at most once by the evaluation of an expression. Furthermore,
2550 the prior value shall be read only to determine the value to be
2551 stored.''. If a program breaks these rules, the results on any
2552 particular implementation are entirely unpredictable.
2554 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2555 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2556 diagnosed by this option, and it may give an occasional false positive
2557 result, but in general it has been found fairly effective at detecting
2558 this sort of problem in programs.
2560 The present implementation of this option only works for C programs. A
2561 future implementation may also work for C++ programs.
2563 The C standard is worded confusingly, therefore there is some debate
2564 over the precise meaning of the sequence point rules in subtle cases.
2565 Links to discussions of the problem, including proposed formal
2566 definitions, may be found on the GCC readings page, at
2567 @w{@uref{http://gcc.gnu.org/readings.html}}.
2569 This warning is enabled by @option{-Wall}.
2572 @opindex Wreturn-type
2573 Warn whenever a function is defined with a return-type that defaults to
2574 @code{int}. Also warn about any @code{return} statement with no
2575 return-value in a function whose return-type is not @code{void}.
2577 For C, also warn if the return type of a function has a type qualifier
2578 such as @code{const}. Such a type qualifier has no effect, since the
2579 value returned by a function is not an lvalue. ISO C prohibits
2580 qualified @code{void} return types on function definitions, so such
2581 return types always receive a warning even without this option.
2583 For C++, a function without return type always produces a diagnostic
2584 message, even when @option{-Wno-return-type} is specified. The only
2585 exceptions are @samp{main} and functions defined in system headers.
2587 This warning is enabled by @option{-Wall}.
2591 Warn whenever a @code{switch} statement has an index of enumerated type
2592 and lacks a @code{case} for one or more of the named codes of that
2593 enumeration. (The presence of a @code{default} label prevents this
2594 warning.) @code{case} labels outside the enumeration range also
2595 provoke warnings when this option is used.
2596 This warning is enabled by @option{-Wall}.
2598 @item -Wswitch-default
2599 @opindex Wswitch-switch
2600 Warn whenever a @code{switch} statement does not have a @code{default}
2604 @opindex Wswitch-enum
2605 Warn whenever a @code{switch} statement has an index of enumerated type
2606 and lacks a @code{case} for one or more of the named codes of that
2607 enumeration. @code{case} labels outside the enumeration range also
2608 provoke warnings when this option is used.
2612 Warn if any trigraphs are encountered that might change the meaning of
2613 the program (trigraphs within comments are not warned about).
2614 This warning is enabled by @option{-Wall}.
2616 @item -Wunused-function
2617 @opindex Wunused-function
2618 Warn whenever a static function is declared but not defined or a
2619 non-inline static function is unused.
2620 This warning is enabled by @option{-Wall}.
2622 @item -Wunused-label
2623 @opindex Wunused-label
2624 Warn whenever a label is declared but not used.
2625 This warning is enabled by @option{-Wall}.
2627 To suppress this warning use the @samp{unused} attribute
2628 (@pxref{Variable Attributes}).
2630 @item -Wunused-parameter
2631 @opindex Wunused-parameter
2632 Warn whenever a function parameter is unused aside from its declaration.
2634 To suppress this warning use the @samp{unused} attribute
2635 (@pxref{Variable Attributes}).
2637 @item -Wunused-variable
2638 @opindex Wunused-variable
2639 Warn whenever a local variable or non-constant static variable is unused
2640 aside from its declaration
2641 This warning is enabled by @option{-Wall}.
2643 To suppress this warning use the @samp{unused} attribute
2644 (@pxref{Variable Attributes}).
2646 @item -Wunused-value
2647 @opindex Wunused-value
2648 Warn whenever a statement computes a result that is explicitly not used.
2649 This warning is enabled by @option{-Wall}.
2651 To suppress this warning cast the expression to @samp{void}.
2655 All the above @option{-Wunused} options combined.
2657 In order to get a warning about an unused function parameter, you must
2658 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2659 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2661 @item -Wuninitialized
2662 @opindex Wuninitialized
2663 Warn if an automatic variable is used without first being initialized or
2664 if a variable may be clobbered by a @code{setjmp} call.
2666 These warnings are possible only in optimizing compilation,
2667 because they require data flow information that is computed only
2668 when optimizing. If you don't specify @option{-O}, you simply won't
2671 If you want to warn about code which uses the uninitialized value of the
2672 variable in its own initializer, use the @option{-Winit-self} option.
2674 These warnings occur for individual uninitialized or clobbered
2675 elements of structure, union or array variables as well as for
2676 variables which are uninitialized or clobbered as a whole. They do
2677 not occur for variables or elements declared @code{volatile}. Because
2678 these warnings depend on optimization, the exact variables or elements
2679 for which there are warnings will depend on the precise optimization
2680 options and version of GCC used.
2682 Note that there may be no warning about a variable that is used only
2683 to compute a value that itself is never used, because such
2684 computations may be deleted by data flow analysis before the warnings
2687 These warnings are made optional because GCC is not smart
2688 enough to see all the reasons why the code might be correct
2689 despite appearing to have an error. Here is one example of how
2710 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2711 always initialized, but GCC doesn't know this. Here is
2712 another common case:
2717 if (change_y) save_y = y, y = new_y;
2719 if (change_y) y = save_y;
2724 This has no bug because @code{save_y} is used only if it is set.
2726 @cindex @code{longjmp} warnings
2727 This option also warns when a non-volatile automatic variable might be
2728 changed by a call to @code{longjmp}. These warnings as well are possible
2729 only in optimizing compilation.
2731 The compiler sees only the calls to @code{setjmp}. It cannot know
2732 where @code{longjmp} will be called; in fact, a signal handler could
2733 call it at any point in the code. As a result, you may get a warning
2734 even when there is in fact no problem because @code{longjmp} cannot
2735 in fact be called at the place which would cause a problem.
2737 Some spurious warnings can be avoided if you declare all the functions
2738 you use that never return as @code{noreturn}. @xref{Function
2741 This warning is enabled by @option{-Wall}.
2743 @item -Wunknown-pragmas
2744 @opindex Wunknown-pragmas
2745 @cindex warning for unknown pragmas
2746 @cindex unknown pragmas, warning
2747 @cindex pragmas, warning of unknown
2748 Warn when a #pragma directive is encountered which is not understood by
2749 GCC@. If this command line option is used, warnings will even be issued
2750 for unknown pragmas in system header files. This is not the case if
2751 the warnings were only enabled by the @option{-Wall} command line option.
2754 @opindex Wno-pragmas
2756 Do not warn about misuses of pragmas, such as incorrect parameters,
2757 invalid syntax, or conflicts between pragmas. See also
2758 @samp{-Wunknown-pragmas}.
2760 @item -Wstrict-aliasing
2761 @opindex Wstrict-aliasing
2762 This option is only active when @option{-fstrict-aliasing} is active.
2763 It warns about code which might break the strict aliasing rules that the
2764 compiler is using for optimization. The warning does not catch all
2765 cases, but does attempt to catch the more common pitfalls. It is
2766 included in @option{-Wall}.
2768 @item -Wstrict-aliasing=2
2769 @opindex Wstrict-aliasing=2
2770 This option is only active when @option{-fstrict-aliasing} is active.
2771 It warns about code which might break the strict aliasing rules that the
2772 compiler is using for optimization. This warning catches more cases than
2773 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2774 cases that are safe.
2778 All of the above @samp{-W} options combined. This enables all the
2779 warnings about constructions that some users consider questionable, and
2780 that are easy to avoid (or modify to prevent the warning), even in
2781 conjunction with macros. This also enables some language-specific
2782 warnings described in @ref{C++ Dialect Options} and
2783 @ref{Objective-C and Objective-C++ Dialect Options}.
2786 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2787 Some of them warn about constructions that users generally do not
2788 consider questionable, but which occasionally you might wish to check
2789 for; others warn about constructions that are necessary or hard to avoid
2790 in some cases, and there is no simple way to modify the code to suppress
2797 (This option used to be called @option{-W}. The older name is still
2798 supported, but the newer name is more descriptive.) Print extra warning
2799 messages for these events:
2803 A function can return either with or without a value. (Falling
2804 off the end of the function body is considered returning without
2805 a value.) For example, this function would evoke such a
2819 An expression-statement or the left-hand side of a comma expression
2820 contains no side effects.
2821 To suppress the warning, cast the unused expression to void.
2822 For example, an expression such as @samp{x[i,j]} will cause a warning,
2823 but @samp{x[(void)i,j]} will not.
2826 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2829 Storage-class specifiers like @code{static} are not the first things in
2830 a declaration. According to the C Standard, this usage is obsolescent.
2833 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2837 A comparison between signed and unsigned values could produce an
2838 incorrect result when the signed value is converted to unsigned.
2839 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2842 An aggregate has an initializer which does not initialize all members.
2843 This warning can be independently controlled by
2844 @option{-Wmissing-field-initializers}.
2847 A function parameter is declared without a type specifier in K&R-style
2855 An empty body occurs in an @samp{if} or @samp{else} statement.
2858 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2859 @samp{>}, or @samp{>=}.
2862 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2865 Any of several floating-point events that often indicate errors, such as
2866 overflow, underflow, loss of precision, etc.
2868 @item @r{(C++ only)}
2869 An enumerator and a non-enumerator both appear in a conditional expression.
2871 @item @r{(C++ only)}
2872 A non-static reference or non-static @samp{const} member appears in a
2873 class without constructors.
2875 @item @r{(C++ only)}
2876 Ambiguous virtual bases.
2878 @item @r{(C++ only)}
2879 Subscripting an array which has been declared @samp{register}.
2881 @item @r{(C++ only)}
2882 Taking the address of a variable which has been declared @samp{register}.
2884 @item @r{(C++ only)}
2885 A base class is not initialized in a derived class' copy constructor.
2888 @item -Wno-div-by-zero
2889 @opindex Wno-div-by-zero
2890 @opindex Wdiv-by-zero
2891 Do not warn about compile-time integer division by zero. Floating point
2892 division by zero is not warned about, as it can be a legitimate way of
2893 obtaining infinities and NaNs.
2895 @item -Wsystem-headers
2896 @opindex Wsystem-headers
2897 @cindex warnings from system headers
2898 @cindex system headers, warnings from
2899 Print warning messages for constructs found in system header files.
2900 Warnings from system headers are normally suppressed, on the assumption
2901 that they usually do not indicate real problems and would only make the
2902 compiler output harder to read. Using this command line option tells
2903 GCC to emit warnings from system headers as if they occurred in user
2904 code. However, note that using @option{-Wall} in conjunction with this
2905 option will @emph{not} warn about unknown pragmas in system
2906 headers---for that, @option{-Wunknown-pragmas} must also be used.
2909 @opindex Wfloat-equal
2910 Warn if floating point values are used in equality comparisons.
2912 The idea behind this is that sometimes it is convenient (for the
2913 programmer) to consider floating-point values as approximations to
2914 infinitely precise real numbers. If you are doing this, then you need
2915 to compute (by analyzing the code, or in some other way) the maximum or
2916 likely maximum error that the computation introduces, and allow for it
2917 when performing comparisons (and when producing output, but that's a
2918 different problem). In particular, instead of testing for equality, you
2919 would check to see whether the two values have ranges that overlap; and
2920 this is done with the relational operators, so equality comparisons are
2923 @item -Wtraditional @r{(C only)}
2924 @opindex Wtraditional
2925 Warn about certain constructs that behave differently in traditional and
2926 ISO C@. Also warn about ISO C constructs that have no traditional C
2927 equivalent, and/or problematic constructs which should be avoided.
2931 Macro parameters that appear within string literals in the macro body.
2932 In traditional C macro replacement takes place within string literals,
2933 but does not in ISO C@.
2936 In traditional C, some preprocessor directives did not exist.
2937 Traditional preprocessors would only consider a line to be a directive
2938 if the @samp{#} appeared in column 1 on the line. Therefore
2939 @option{-Wtraditional} warns about directives that traditional C
2940 understands but would ignore because the @samp{#} does not appear as the
2941 first character on the line. It also suggests you hide directives like
2942 @samp{#pragma} not understood by traditional C by indenting them. Some
2943 traditional implementations would not recognize @samp{#elif}, so it
2944 suggests avoiding it altogether.
2947 A function-like macro that appears without arguments.
2950 The unary plus operator.
2953 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2954 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2955 constants.) Note, these suffixes appear in macros defined in the system
2956 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2957 Use of these macros in user code might normally lead to spurious
2958 warnings, however GCC's integrated preprocessor has enough context to
2959 avoid warning in these cases.
2962 A function declared external in one block and then used after the end of
2966 A @code{switch} statement has an operand of type @code{long}.
2969 A non-@code{static} function declaration follows a @code{static} one.
2970 This construct is not accepted by some traditional C compilers.
2973 The ISO type of an integer constant has a different width or
2974 signedness from its traditional type. This warning is only issued if
2975 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2976 typically represent bit patterns, are not warned about.
2979 Usage of ISO string concatenation is detected.
2982 Initialization of automatic aggregates.
2985 Identifier conflicts with labels. Traditional C lacks a separate
2986 namespace for labels.
2989 Initialization of unions. If the initializer is zero, the warning is
2990 omitted. This is done under the assumption that the zero initializer in
2991 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2992 initializer warnings and relies on default initialization to zero in the
2996 Conversions by prototypes between fixed/floating point values and vice
2997 versa. The absence of these prototypes when compiling with traditional
2998 C would cause serious problems. This is a subset of the possible
2999 conversion warnings, for the full set use @option{-Wconversion}.
3002 Use of ISO C style function definitions. This warning intentionally is
3003 @emph{not} issued for prototype declarations or variadic functions
3004 because these ISO C features will appear in your code when using
3005 libiberty's traditional C compatibility macros, @code{PARAMS} and
3006 @code{VPARAMS}. This warning is also bypassed for nested functions
3007 because that feature is already a GCC extension and thus not relevant to
3008 traditional C compatibility.
3011 @item -Wdeclaration-after-statement @r{(C only)}
3012 @opindex Wdeclaration-after-statement
3013 Warn when a declaration is found after a statement in a block. This
3014 construct, known from C++, was introduced with ISO C99 and is by default
3015 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3016 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3020 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3022 @item -Wno-endif-labels
3023 @opindex Wno-endif-labels
3024 @opindex Wendif-labels
3025 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3029 Warn whenever a local variable shadows another local variable, parameter or
3030 global variable or whenever a built-in function is shadowed.
3032 @item -Wlarger-than-@var{len}
3033 @opindex Wlarger-than
3034 Warn whenever an object of larger than @var{len} bytes is defined.
3036 @item -Wunsafe-loop-optimizations
3037 @opindex Wunsafe-loop-optimizations
3038 Warn if the loop cannot be optimized because the compiler could not
3039 assume anything on the bounds of the loop indices. With
3040 @option{-funsafe-loop-optimizations} warn if the compiler made
3043 @item -Wpointer-arith
3044 @opindex Wpointer-arith
3045 Warn about anything that depends on the ``size of'' a function type or
3046 of @code{void}. GNU C assigns these types a size of 1, for
3047 convenience in calculations with @code{void *} pointers and pointers
3050 @item -Wbad-function-cast @r{(C only)}
3051 @opindex Wbad-function-cast
3052 Warn whenever a function call is cast to a non-matching type.
3053 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3056 Warn about ISO C constructs that are outside of the common subset of
3057 ISO C and ISO C++, e.g.@: request for implicit conversion from
3058 @code{void *} to a pointer to non-@code{void} type.
3062 Warn whenever a pointer is cast so as to remove a type qualifier from
3063 the target type. For example, warn if a @code{const char *} is cast
3064 to an ordinary @code{char *}.
3067 @opindex Wcast-align
3068 Warn whenever a pointer is cast such that the required alignment of the
3069 target is increased. For example, warn if a @code{char *} is cast to
3070 an @code{int *} on machines where integers can only be accessed at
3071 two- or four-byte boundaries.
3073 @item -Wwrite-strings
3074 @opindex Wwrite-strings
3075 When compiling C, give string constants the type @code{const
3076 char[@var{length}]} so that
3077 copying the address of one into a non-@code{const} @code{char *}
3078 pointer will get a warning; when compiling C++, warn about the
3079 deprecated conversion from string constants to @code{char *}.
3080 These warnings will help you find at
3081 compile time code that can try to write into a string constant, but
3082 only if you have been very careful about using @code{const} in
3083 declarations and prototypes. Otherwise, it will just be a nuisance;
3084 this is why we did not make @option{-Wall} request these warnings.
3087 @opindex Wconversion
3088 Warn if a prototype causes a type conversion that is different from what
3089 would happen to the same argument in the absence of a prototype. This
3090 includes conversions of fixed point to floating and vice versa, and
3091 conversions changing the width or signedness of a fixed point argument
3092 except when the same as the default promotion.
3094 Also, warn if a negative integer constant expression is implicitly
3095 converted to an unsigned type. For example, warn about the assignment
3096 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3097 casts like @code{(unsigned) -1}.
3099 @item -Wsign-compare
3100 @opindex Wsign-compare
3101 @cindex warning for comparison of signed and unsigned values
3102 @cindex comparison of signed and unsigned values, warning
3103 @cindex signed and unsigned values, comparison warning
3104 Warn when a comparison between signed and unsigned values could produce
3105 an incorrect result when the signed value is converted to unsigned.
3106 This warning is also enabled by @option{-Wextra}; to get the other warnings
3107 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3109 @item -Waggregate-return
3110 @opindex Waggregate-return
3111 Warn if any functions that return structures or unions are defined or
3112 called. (In languages where you can return an array, this also elicits
3116 @opindex Walways-true
3117 Warn about comparisons which are always true such as testing if unsigned
3118 values are greater than zero.
3120 @item -Wno-attributes
3121 @opindex Wno-attributes
3122 @opindex Wattributes
3123 Do not warn if an unexpected @code{__attribute__} is used, such as
3124 unrecognized attributes, function attributes applied to variables,
3125 etc. This will not stop errors for incorrect use of supported
3128 @item -Wstrict-prototypes @r{(C only)}
3129 @opindex Wstrict-prototypes
3130 Warn if a function is declared or defined without specifying the
3131 argument types. (An old-style function definition is permitted without
3132 a warning if preceded by a declaration which specifies the argument
3135 @item -Wold-style-definition @r{(C only)}
3136 @opindex Wold-style-definition
3137 Warn if an old-style function definition is used. A warning is given
3138 even if there is a previous prototype.
3140 @item -Wmissing-prototypes @r{(C only)}
3141 @opindex Wmissing-prototypes
3142 Warn if a global function is defined without a previous prototype
3143 declaration. This warning is issued even if the definition itself
3144 provides a prototype. The aim is to detect global functions that fail
3145 to be declared in header files.
3147 @item -Wmissing-declarations @r{(C only)}
3148 @opindex Wmissing-declarations
3149 Warn if a global function is defined without a previous declaration.
3150 Do so even if the definition itself provides a prototype.
3151 Use this option to detect global functions that are not declared in
3154 @item -Wmissing-field-initializers
3155 @opindex Wmissing-field-initializers
3158 Warn if a structure's initializer has some fields missing. For
3159 example, the following code would cause such a warning, because
3160 @code{x.h} is implicitly zero:
3163 struct s @{ int f, g, h; @};
3164 struct s x = @{ 3, 4 @};
3167 This option does not warn about designated initializers, so the following
3168 modification would not trigger a warning:
3171 struct s @{ int f, g, h; @};
3172 struct s x = @{ .f = 3, .g = 4 @};
3175 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3176 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3178 @item -Wmissing-noreturn
3179 @opindex Wmissing-noreturn
3180 Warn about functions which might be candidates for attribute @code{noreturn}.
3181 Note these are only possible candidates, not absolute ones. Care should
3182 be taken to manually verify functions actually do not ever return before
3183 adding the @code{noreturn} attribute, otherwise subtle code generation
3184 bugs could be introduced. You will not get a warning for @code{main} in
3185 hosted C environments.
3187 @item -Wmissing-format-attribute
3188 @opindex Wmissing-format-attribute
3190 Warn about function pointers which might be candidates for @code{format}
3191 attributes. Note these are only possible candidates, not absolute ones.
3192 GCC will guess that function pointers with @code{format} attributes that
3193 are used in assignment, initialization, parameter passing or return
3194 statements should have a corresponding @code{format} attribute in the
3195 resulting type. I.e.@: the left-hand side of the assignment or
3196 initialization, the type of the parameter variable, or the return type
3197 of the containing function respectively should also have a @code{format}
3198 attribute to avoid the warning.
3200 GCC will also warn about function definitions which might be
3201 candidates for @code{format} attributes. Again, these are only
3202 possible candidates. GCC will guess that @code{format} attributes
3203 might be appropriate for any function that calls a function like
3204 @code{vprintf} or @code{vscanf}, but this might not always be the
3205 case, and some functions for which @code{format} attributes are
3206 appropriate may not be detected.
3208 @item -Wno-multichar
3209 @opindex Wno-multichar
3211 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3212 Usually they indicate a typo in the user's code, as they have
3213 implementation-defined values, and should not be used in portable code.
3215 @item -Wnormalized=<none|id|nfc|nfkc>
3216 @opindex Wnormalized
3219 @cindex character set, input normalization
3220 In ISO C and ISO C++, two identifiers are different if they are
3221 different sequences of characters. However, sometimes when characters
3222 outside the basic ASCII character set are used, you can have two
3223 different character sequences that look the same. To avoid confusion,
3224 the ISO 10646 standard sets out some @dfn{normalization rules} which
3225 when applied ensure that two sequences that look the same are turned into
3226 the same sequence. GCC can warn you if you are using identifiers which
3227 have not been normalized; this option controls that warning.
3229 There are four levels of warning that GCC supports. The default is
3230 @option{-Wnormalized=nfc}, which warns about any identifier which is
3231 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3232 recommended form for most uses.
3234 Unfortunately, there are some characters which ISO C and ISO C++ allow
3235 in identifiers that when turned into NFC aren't allowable as
3236 identifiers. That is, there's no way to use these symbols in portable
3237 ISO C or C++ and have all your identifiers in NFC.
3238 @option{-Wnormalized=id} suppresses the warning for these characters.
3239 It is hoped that future versions of the standards involved will correct
3240 this, which is why this option is not the default.
3242 You can switch the warning off for all characters by writing
3243 @option{-Wnormalized=none}. You would only want to do this if you
3244 were using some other normalization scheme (like ``D''), because
3245 otherwise you can easily create bugs that are literally impossible to see.
3247 Some characters in ISO 10646 have distinct meanings but look identical
3248 in some fonts or display methodologies, especially once formatting has
3249 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3250 LETTER N'', will display just like a regular @code{n} which has been
3251 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3252 normalisation scheme to convert all these into a standard form as
3253 well, and GCC will warn if your code is not in NFKC if you use
3254 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3255 about every identifier that contains the letter O because it might be
3256 confused with the digit 0, and so is not the default, but may be
3257 useful as a local coding convention if the programming environment is
3258 unable to be fixed to display these characters distinctly.
3260 @item -Wno-deprecated-declarations
3261 @opindex Wno-deprecated-declarations
3262 Do not warn about uses of functions, variables, and types marked as
3263 deprecated by using the @code{deprecated} attribute.
3264 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3265 @pxref{Type Attributes}.)
3269 Warn if a structure is given the packed attribute, but the packed
3270 attribute has no effect on the layout or size of the structure.
3271 Such structures may be mis-aligned for little benefit. For
3272 instance, in this code, the variable @code{f.x} in @code{struct bar}
3273 will be misaligned even though @code{struct bar} does not itself
3274 have the packed attribute:
3281 @} __attribute__((packed));
3291 Warn if padding is included in a structure, either to align an element
3292 of the structure or to align the whole structure. Sometimes when this
3293 happens it is possible to rearrange the fields of the structure to
3294 reduce the padding and so make the structure smaller.
3296 @item -Wredundant-decls
3297 @opindex Wredundant-decls
3298 Warn if anything is declared more than once in the same scope, even in
3299 cases where multiple declaration is valid and changes nothing.
3301 @item -Wnested-externs @r{(C only)}
3302 @opindex Wnested-externs
3303 Warn if an @code{extern} declaration is encountered within a function.
3305 @item -Wunreachable-code
3306 @opindex Wunreachable-code
3307 Warn if the compiler detects that code will never be executed.
3309 This option is intended to warn when the compiler detects that at
3310 least a whole line of source code will never be executed, because
3311 some condition is never satisfied or because it is after a
3312 procedure that never returns.
3314 It is possible for this option to produce a warning even though there
3315 are circumstances under which part of the affected line can be executed,
3316 so care should be taken when removing apparently-unreachable code.
3318 For instance, when a function is inlined, a warning may mean that the
3319 line is unreachable in only one inlined copy of the function.
3321 This option is not made part of @option{-Wall} because in a debugging
3322 version of a program there is often substantial code which checks
3323 correct functioning of the program and is, hopefully, unreachable
3324 because the program does work. Another common use of unreachable
3325 code is to provide behavior which is selectable at compile-time.
3329 Warn if a function can not be inlined and it was declared as inline.
3330 Even with this option, the compiler will not warn about failures to
3331 inline functions declared in system headers.
3333 The compiler uses a variety of heuristics to determine whether or not
3334 to inline a function. For example, the compiler takes into account
3335 the size of the function being inlined and the amount of inlining
3336 that has already been done in the current function. Therefore,
3337 seemingly insignificant changes in the source program can cause the
3338 warnings produced by @option{-Winline} to appear or disappear.
3340 @item -Wno-invalid-offsetof @r{(C++ only)}
3341 @opindex Wno-invalid-offsetof
3342 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3343 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3344 to a non-POD type is undefined. In existing C++ implementations,
3345 however, @samp{offsetof} typically gives meaningful results even when
3346 applied to certain kinds of non-POD types. (Such as a simple
3347 @samp{struct} that fails to be a POD type only by virtue of having a
3348 constructor.) This flag is for users who are aware that they are
3349 writing nonportable code and who have deliberately chosen to ignore the
3352 The restrictions on @samp{offsetof} may be relaxed in a future version
3353 of the C++ standard.
3355 @item -Wno-int-to-pointer-cast @r{(C only)}
3356 @opindex Wno-int-to-pointer-cast
3357 Suppress warnings from casts to pointer type of an integer of a
3360 @item -Wno-pointer-to-int-cast @r{(C only)}
3361 @opindex Wno-pointer-to-int-cast
3362 Suppress warnings from casts from a pointer to an integer type of a
3366 @opindex Winvalid-pch
3367 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3368 the search path but can't be used.
3372 @opindex Wno-long-long
3373 Warn if @samp{long long} type is used. This is default. To inhibit
3374 the warning messages, use @option{-Wno-long-long}. Flags
3375 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3376 only when @option{-pedantic} flag is used.
3378 @item -Wvariadic-macros
3379 @opindex Wvariadic-macros
3380 @opindex Wno-variadic-macros
3381 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3382 alternate syntax when in pedantic ISO C99 mode. This is default.
3383 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3385 @item -Wvolatile-register-var
3386 @opindex Wvolatile-register-var
3387 @opindex Wno-volatile-register-var
3388 Warn if a register variable is declared volatile. The volatile
3389 modifier does not inhibit all optimizations that may eliminate reads
3390 and/or writes to register variables.
3392 @item -Wdisabled-optimization
3393 @opindex Wdisabled-optimization
3394 Warn if a requested optimization pass is disabled. This warning does
3395 not generally indicate that there is anything wrong with your code; it
3396 merely indicates that GCC's optimizers were unable to handle the code
3397 effectively. Often, the problem is that your code is too big or too
3398 complex; GCC will refuse to optimize programs when the optimization
3399 itself is likely to take inordinate amounts of time.
3401 @item -Wno-pointer-sign
3402 @opindex Wno-pointer-sign
3403 Don't warn for pointer argument passing or assignment with different signedness.
3404 Only useful in the negative form since this warning is enabled by default.
3405 This option is only supported for C and Objective-C@.
3409 Make all warnings into errors.
3411 @item -Wstack-protector
3412 @opindex Wstack-protector
3413 This option is only active when @option{-fstack-protector} is active. It
3414 warns about functions that will not be protected against stack smashing.
3416 @item -Wstring-literal-comparison
3417 @opindex Wstring-literal-comparison
3418 Warn about suspicious comparisons to string literal constants. In C,
3419 direct comparisons against the memory address of a string literal, such
3420 as @code{if (x == "abc")}, typically indicate a programmer error, and
3421 even when intentional, result in unspecified behavior and are not portable.
3422 Usually these warnings alert that the programmer intended to use
3423 @code{strcmp}. This warning is enabled by @option{-Wall}.
3427 @node Debugging Options
3428 @section Options for Debugging Your Program or GCC
3429 @cindex options, debugging
3430 @cindex debugging information options
3432 GCC has various special options that are used for debugging
3433 either your program or GCC:
3438 Produce debugging information in the operating system's native format
3439 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3442 On most systems that use stabs format, @option{-g} enables use of extra
3443 debugging information that only GDB can use; this extra information
3444 makes debugging work better in GDB but will probably make other debuggers
3446 refuse to read the program. If you want to control for certain whether
3447 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3448 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3450 GCC allows you to use @option{-g} with
3451 @option{-O}. The shortcuts taken by optimized code may occasionally
3452 produce surprising results: some variables you declared may not exist
3453 at all; flow of control may briefly move where you did not expect it;
3454 some statements may not be executed because they compute constant
3455 results or their values were already at hand; some statements may
3456 execute in different places because they were moved out of loops.
3458 Nevertheless it proves possible to debug optimized output. This makes
3459 it reasonable to use the optimizer for programs that might have bugs.
3461 The following options are useful when GCC is generated with the
3462 capability for more than one debugging format.
3466 Produce debugging information for use by GDB@. This means to use the
3467 most expressive format available (DWARF 2, stabs, or the native format
3468 if neither of those are supported), including GDB extensions if at all
3473 Produce debugging information in stabs format (if that is supported),
3474 without GDB extensions. This is the format used by DBX on most BSD
3475 systems. On MIPS, Alpha and System V Release 4 systems this option
3476 produces stabs debugging output which is not understood by DBX or SDB@.
3477 On System V Release 4 systems this option requires the GNU assembler.
3479 @item -feliminate-unused-debug-symbols
3480 @opindex feliminate-unused-debug-symbols
3481 Produce debugging information in stabs format (if that is supported),
3482 for only symbols that are actually used.
3486 Produce debugging information in stabs format (if that is supported),
3487 using GNU extensions understood only by the GNU debugger (GDB)@. The
3488 use of these extensions is likely to make other debuggers crash or
3489 refuse to read the program.
3493 Produce debugging information in COFF format (if that is supported).
3494 This is the format used by SDB on most System V systems prior to
3499 Produce debugging information in XCOFF format (if that is supported).
3500 This is the format used by the DBX debugger on IBM RS/6000 systems.
3504 Produce debugging information in XCOFF format (if that is supported),
3505 using GNU extensions understood only by the GNU debugger (GDB)@. The
3506 use of these extensions is likely to make other debuggers crash or
3507 refuse to read the program, and may cause assemblers other than the GNU
3508 assembler (GAS) to fail with an error.
3512 Produce debugging information in DWARF version 2 format (if that is
3513 supported). This is the format used by DBX on IRIX 6. With this
3514 option, GCC uses features of DWARF version 3 when they are useful;
3515 version 3 is upward compatible with version 2, but may still cause
3516 problems for older debuggers.
3520 Produce debugging information in VMS debug format (if that is
3521 supported). This is the format used by DEBUG on VMS systems.
3524 @itemx -ggdb@var{level}
3525 @itemx -gstabs@var{level}
3526 @itemx -gcoff@var{level}
3527 @itemx -gxcoff@var{level}
3528 @itemx -gvms@var{level}
3529 Request debugging information and also use @var{level} to specify how
3530 much information. The default level is 2.
3532 Level 1 produces minimal information, enough for making backtraces in
3533 parts of the program that you don't plan to debug. This includes
3534 descriptions of functions and external variables, but no information
3535 about local variables and no line numbers.
3537 Level 3 includes extra information, such as all the macro definitions
3538 present in the program. Some debuggers support macro expansion when
3539 you use @option{-g3}.
3541 @option{-gdwarf-2} does not accept a concatenated debug level, because
3542 GCC used to support an option @option{-gdwarf} that meant to generate
3543 debug information in version 1 of the DWARF format (which is very
3544 different from version 2), and it would have been too confusing. That
3545 debug format is long obsolete, but the option cannot be changed now.
3546 Instead use an additional @option{-g@var{level}} option to change the
3547 debug level for DWARF2.
3549 @item -feliminate-dwarf2-dups
3550 @opindex feliminate-dwarf2-dups
3551 Compress DWARF2 debugging information by eliminating duplicated
3552 information about each symbol. This option only makes sense when
3553 generating DWARF2 debugging information with @option{-gdwarf-2}.
3555 @cindex @command{prof}
3558 Generate extra code to write profile information suitable for the
3559 analysis program @command{prof}. You must use this option when compiling
3560 the source files you want data about, and you must also use it when
3563 @cindex @command{gprof}
3566 Generate extra code to write profile information suitable for the
3567 analysis program @command{gprof}. You must use this option when compiling
3568 the source files you want data about, and you must also use it when
3573 Makes the compiler print out each function name as it is compiled, and
3574 print some statistics about each pass when it finishes.
3577 @opindex ftime-report
3578 Makes the compiler print some statistics about the time consumed by each
3579 pass when it finishes.
3582 @opindex fmem-report
3583 Makes the compiler print some statistics about permanent memory
3584 allocation when it finishes.
3586 @item -fprofile-arcs
3587 @opindex fprofile-arcs
3588 Add code so that program flow @dfn{arcs} are instrumented. During
3589 execution the program records how many times each branch and call is
3590 executed and how many times it is taken or returns. When the compiled
3591 program exits it saves this data to a file called
3592 @file{@var{auxname}.gcda} for each source file. The data may be used for
3593 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3594 test coverage analysis (@option{-ftest-coverage}). Each object file's
3595 @var{auxname} is generated from the name of the output file, if
3596 explicitly specified and it is not the final executable, otherwise it is
3597 the basename of the source file. In both cases any suffix is removed
3598 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3599 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3600 @xref{Cross-profiling}.
3602 @cindex @command{gcov}
3606 This option is used to compile and link code instrumented for coverage
3607 analysis. The option is a synonym for @option{-fprofile-arcs}
3608 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3609 linking). See the documentation for those options for more details.
3614 Compile the source files with @option{-fprofile-arcs} plus optimization
3615 and code generation options. For test coverage analysis, use the
3616 additional @option{-ftest-coverage} option. You do not need to profile
3617 every source file in a program.
3620 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3621 (the latter implies the former).
3624 Run the program on a representative workload to generate the arc profile
3625 information. This may be repeated any number of times. You can run
3626 concurrent instances of your program, and provided that the file system
3627 supports locking, the data files will be correctly updated. Also
3628 @code{fork} calls are detected and correctly handled (double counting
3632 For profile-directed optimizations, compile the source files again with
3633 the same optimization and code generation options plus
3634 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3635 Control Optimization}).
3638 For test coverage analysis, use @command{gcov} to produce human readable
3639 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3640 @command{gcov} documentation for further information.
3644 With @option{-fprofile-arcs}, for each function of your program GCC
3645 creates a program flow graph, then finds a spanning tree for the graph.
3646 Only arcs that are not on the spanning tree have to be instrumented: the
3647 compiler adds code to count the number of times that these arcs are
3648 executed. When an arc is the only exit or only entrance to a block, the
3649 instrumentation code can be added to the block; otherwise, a new basic
3650 block must be created to hold the instrumentation code.
3653 @item -ftest-coverage
3654 @opindex ftest-coverage
3655 Produce a notes file that the @command{gcov} code-coverage utility
3656 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3657 show program coverage. Each source file's note file is called
3658 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3659 above for a description of @var{auxname} and instructions on how to
3660 generate test coverage data. Coverage data will match the source files
3661 more closely, if you do not optimize.
3663 @item -d@var{letters}
3664 @item -fdump-rtl-@var{pass}
3666 Says to make debugging dumps during compilation at times specified by
3667 @var{letters}. This is used for debugging the RTL-based passes of the
3668 compiler. The file names for most of the dumps are made by appending a
3669 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3670 from the name of the output file, if explicitly specified and it is not
3671 an executable, otherwise it is the basename of the source file.
3673 Most debug dumps can be enabled either passing a letter to the @option{-d}
3674 option, or with a long @option{-fdump-rtl} switch; here are the possible
3675 letters for use in @var{letters} and @var{pass}, and their meanings:
3680 Annotate the assembler output with miscellaneous debugging information.
3683 @itemx -fdump-rtl-bp
3685 @opindex fdump-rtl-bp
3686 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3689 @itemx -fdump-rtl-bbro
3691 @opindex fdump-rtl-bbro
3692 Dump after block reordering, to @file{@var{file}.30.bbro}.
3695 @itemx -fdump-rtl-combine
3697 @opindex fdump-rtl-combine
3698 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3701 @itemx -fdump-rtl-ce1
3702 @itemx -fdump-rtl-ce2
3704 @opindex fdump-rtl-ce1
3705 @opindex fdump-rtl-ce2
3706 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3707 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3708 and @option{-fdump-rtl-ce2} enable dumping after the second if
3709 conversion, to the file @file{@var{file}.18.ce2}.
3712 @itemx -fdump-rtl-btl
3713 @itemx -fdump-rtl-dbr
3715 @opindex fdump-rtl-btl
3716 @opindex fdump-rtl-dbr
3717 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3718 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3719 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3720 scheduling, to @file{@var{file}.36.dbr}.
3724 Dump all macro definitions, at the end of preprocessing, in addition to
3728 @itemx -fdump-rtl-ce3
3730 @opindex fdump-rtl-ce3
3731 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3734 @itemx -fdump-rtl-cfg
3735 @itemx -fdump-rtl-life
3737 @opindex fdump-rtl-cfg
3738 @opindex fdump-rtl-life
3739 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3740 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3741 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3742 to @file{@var{file}.16.life}.
3745 @itemx -fdump-rtl-greg
3747 @opindex fdump-rtl-greg
3748 Dump after global register allocation, to @file{@var{file}.23.greg}.
3751 @itemx -fdump-rtl-gcse
3752 @itemx -fdump-rtl-bypass
3754 @opindex fdump-rtl-gcse
3755 @opindex fdump-rtl-bypass
3756 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3757 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3758 enable dumping after jump bypassing and control flow optimizations, to
3759 @file{@var{file}.07.bypass}.
3762 @itemx -fdump-rtl-eh
3764 @opindex fdump-rtl-eh
3765 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3768 @itemx -fdump-rtl-sibling
3770 @opindex fdump-rtl-sibling
3771 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3774 @itemx -fdump-rtl-jump
3776 @opindex fdump-rtl-jump
3777 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3780 @itemx -fdump-rtl-stack
3782 @opindex fdump-rtl-stack
3783 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3786 @itemx -fdump-rtl-lreg
3788 @opindex fdump-rtl-lreg
3789 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3792 @itemx -fdump-rtl-loop
3793 @itemx -fdump-rtl-loop2
3795 @opindex fdump-rtl-loop
3796 @opindex fdump-rtl-loop2
3797 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3798 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3799 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3800 @file{@var{file}.13.loop2}.
3803 @itemx -fdump-rtl-sms
3805 @opindex fdump-rtl-sms
3806 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3809 @itemx -fdump-rtl-mach
3811 @opindex fdump-rtl-mach
3812 Dump after performing the machine dependent reorganization pass, to
3813 @file{@var{file}.35.mach}.
3816 @itemx -fdump-rtl-rnreg
3818 @opindex fdump-rtl-rnreg
3819 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3822 @itemx -fdump-rtl-regmove
3824 @opindex fdump-rtl-regmove
3825 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3828 @itemx -fdump-rtl-postreload
3830 @opindex fdump-rtl-postreload
3831 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3834 @itemx -fdump-rtl-expand
3836 @opindex fdump-rtl-expand
3837 Dump after RTL generation, to @file{@var{file}.00.expand}.
3840 @itemx -fdump-rtl-sched2
3842 @opindex fdump-rtl-sched2
3843 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3846 @itemx -fdump-rtl-cse
3848 @opindex fdump-rtl-cse
3849 Dump after CSE (including the jump optimization that sometimes follows
3850 CSE), to @file{@var{file}.04.cse}.
3853 @itemx -fdump-rtl-sched
3855 @opindex fdump-rtl-sched
3856 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3859 @itemx -fdump-rtl-cse2
3861 @opindex fdump-rtl-cse2
3862 Dump after the second CSE pass (including the jump optimization that
3863 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3866 @itemx -fdump-rtl-tracer
3868 @opindex fdump-rtl-tracer
3869 Dump after running tracer, to @file{@var{file}.12.tracer}.
3872 @itemx -fdump-rtl-vpt
3873 @itemx -fdump-rtl-vartrack
3875 @opindex fdump-rtl-vpt
3876 @opindex fdump-rtl-vartrack
3877 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3878 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3879 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3880 to @file{@var{file}.34.vartrack}.
3883 @itemx -fdump-rtl-flow2
3885 @opindex fdump-rtl-flow2
3886 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3889 @itemx -fdump-rtl-peephole2
3891 @opindex fdump-rtl-peephole2
3892 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3895 @itemx -fdump-rtl-web
3897 @opindex fdump-rtl-web
3898 Dump after live range splitting, to @file{@var{file}.14.web}.
3901 @itemx -fdump-rtl-all
3903 @opindex fdump-rtl-all
3904 Produce all the dumps listed above.
3908 Produce a core dump whenever an error occurs.
3912 Print statistics on memory usage, at the end of the run, to
3917 Annotate the assembler output with a comment indicating which
3918 pattern and alternative was used. The length of each instruction is
3923 Dump the RTL in the assembler output as a comment before each instruction.
3924 Also turns on @option{-dp} annotation.
3928 For each of the other indicated dump files (either with @option{-d} or
3929 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3930 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3934 Just generate RTL for a function instead of compiling it. Usually used
3935 with @samp{r} (@option{-fdump-rtl-expand}).
3939 Dump debugging information during parsing, to standard error.
3942 @item -fdump-unnumbered
3943 @opindex fdump-unnumbered
3944 When doing debugging dumps (see @option{-d} option above), suppress instruction
3945 numbers and line number note output. This makes it more feasible to
3946 use diff on debugging dumps for compiler invocations with different
3947 options, in particular with and without @option{-g}.
3949 @item -fdump-translation-unit @r{(C++ only)}
3950 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3951 @opindex fdump-translation-unit
3952 Dump a representation of the tree structure for the entire translation
3953 unit to a file. The file name is made by appending @file{.tu} to the
3954 source file name. If the @samp{-@var{options}} form is used, @var{options}
3955 controls the details of the dump as described for the
3956 @option{-fdump-tree} options.
3958 @item -fdump-class-hierarchy @r{(C++ only)}
3959 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3960 @opindex fdump-class-hierarchy
3961 Dump a representation of each class's hierarchy and virtual function
3962 table layout to a file. The file name is made by appending @file{.class}
3963 to the source file name. If the @samp{-@var{options}} form is used,
3964 @var{options} controls the details of the dump as described for the
3965 @option{-fdump-tree} options.
3967 @item -fdump-ipa-@var{switch}
3969 Control the dumping at various stages of inter-procedural analysis
3970 language tree to a file. The file name is generated by appending a switch
3971 specific suffix to the source file name. The following dumps are possible:
3975 Enables all inter-procedural analysis dumps; currently the only produced
3976 dump is the @samp{cgraph} dump.
3979 Dumps information about call-graph optimization, unused function removal,
3980 and inlining decisions.
3983 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3984 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3986 Control the dumping at various stages of processing the intermediate
3987 language tree to a file. The file name is generated by appending a switch
3988 specific suffix to the source file name. If the @samp{-@var{options}}
3989 form is used, @var{options} is a list of @samp{-} separated options that
3990 control the details of the dump. Not all options are applicable to all
3991 dumps, those which are not meaningful will be ignored. The following
3992 options are available
3996 Print the address of each node. Usually this is not meaningful as it
3997 changes according to the environment and source file. Its primary use
3998 is for tying up a dump file with a debug environment.
4000 Inhibit dumping of members of a scope or body of a function merely
4001 because that scope has been reached. Only dump such items when they
4002 are directly reachable by some other path. When dumping pretty-printed
4003 trees, this option inhibits dumping the bodies of control structures.
4005 Print a raw representation of the tree. By default, trees are
4006 pretty-printed into a C-like representation.
4008 Enable more detailed dumps (not honored by every dump option).
4010 Enable dumping various statistics about the pass (not honored by every dump
4013 Enable showing basic block boundaries (disabled in raw dumps).
4015 Enable showing virtual operands for every statement.
4017 Enable showing line numbers for statements.
4019 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4021 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4024 The following tree dumps are possible:
4028 Dump before any tree based optimization, to @file{@var{file}.original}.
4031 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4034 Dump after function inlining, to @file{@var{file}.inlined}.
4037 @opindex fdump-tree-gimple
4038 Dump each function before and after the gimplification pass to a file. The
4039 file name is made by appending @file{.gimple} to the source file name.
4042 @opindex fdump-tree-cfg
4043 Dump the control flow graph of each function to a file. The file name is
4044 made by appending @file{.cfg} to the source file name.
4047 @opindex fdump-tree-vcg
4048 Dump the control flow graph of each function to a file in VCG format. The
4049 file name is made by appending @file{.vcg} to the source file name. Note
4050 that if the file contains more than one function, the generated file cannot
4051 be used directly by VCG@. You will need to cut and paste each function's
4052 graph into its own separate file first.
4055 @opindex fdump-tree-ch
4056 Dump each function after copying loop headers. The file name is made by
4057 appending @file{.ch} to the source file name.
4060 @opindex fdump-tree-ssa
4061 Dump SSA related information to a file. The file name is made by appending
4062 @file{.ssa} to the source file name.
4065 @opindex fdump-tree-salias
4066 Dump structure aliasing variable information to a file. This file name
4067 is made by appending @file{.salias} to the source file name.
4070 @opindex fdump-tree-alias
4071 Dump aliasing information for each function. The file name is made by
4072 appending @file{.alias} to the source file name.
4075 @opindex fdump-tree-ccp
4076 Dump each function after CCP@. The file name is made by appending
4077 @file{.ccp} to the source file name.
4080 @opindex fdump-tree-storeccp
4081 Dump each function after STORE-CCP. The file name is made by appending
4082 @file{.storeccp} to the source file name.
4085 @opindex fdump-tree-pre
4086 Dump trees after partial redundancy elimination. The file name is made
4087 by appending @file{.pre} to the source file name.
4090 @opindex fdump-tree-fre
4091 Dump trees after full redundancy elimination. The file name is made
4092 by appending @file{.fre} to the source file name.
4095 @opindex fdump-tree-copyprop
4096 Dump trees after copy propagation. The file name is made
4097 by appending @file{.copyprop} to the source file name.
4099 @item store_copyprop
4100 @opindex fdump-tree-store_copyprop
4101 Dump trees after store copy-propagation. The file name is made
4102 by appending @file{.store_copyprop} to the source file name.
4105 @opindex fdump-tree-dce
4106 Dump each function after dead code elimination. The file name is made by
4107 appending @file{.dce} to the source file name.
4110 @opindex fdump-tree-mudflap
4111 Dump each function after adding mudflap instrumentation. The file name is
4112 made by appending @file{.mudflap} to the source file name.
4115 @opindex fdump-tree-sra
4116 Dump each function after performing scalar replacement of aggregates. The
4117 file name is made by appending @file{.sra} to the source file name.
4120 @opindex fdump-tree-sink
4121 Dump each function after performing code sinking. The file name is made
4122 by appending @file{.sink} to the source file name.
4125 @opindex fdump-tree-dom
4126 Dump each function after applying dominator tree optimizations. The file
4127 name is made by appending @file{.dom} to the source file name.
4130 @opindex fdump-tree-dse
4131 Dump each function after applying dead store elimination. The file
4132 name is made by appending @file{.dse} to the source file name.
4135 @opindex fdump-tree-phiopt
4136 Dump each function after optimizing PHI nodes into straightline code. The file
4137 name is made by appending @file{.phiopt} to the source file name.
4140 @opindex fdump-tree-forwprop
4141 Dump each function after forward propagating single use variables. The file
4142 name is made by appending @file{.forwprop} to the source file name.
4145 @opindex fdump-tree-copyrename
4146 Dump each function after applying the copy rename optimization. The file
4147 name is made by appending @file{.copyrename} to the source file name.
4150 @opindex fdump-tree-nrv
4151 Dump each function after applying the named return value optimization on
4152 generic trees. The file name is made by appending @file{.nrv} to the source
4156 @opindex fdump-tree-vect
4157 Dump each function after applying vectorization of loops. The file name is
4158 made by appending @file{.vect} to the source file name.
4161 @opindex fdump-tree-vrp
4162 Dump each function after Value Range Propagation (VRP). The file name
4163 is made by appending @file{.vrp} to the source file name.
4166 @opindex fdump-tree-all
4167 Enable all the available tree dumps with the flags provided in this option.
4170 @item -ftree-vectorizer-verbose=@var{n}
4171 @opindex ftree-vectorizer-verbose
4172 This option controls the amount of debugging output the vectorizer prints.
4173 This information is written to standard error, unless @option{-fdump-tree-all}
4174 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4175 usual dump listing file, @file{.vect}.
4177 @item -frandom-seed=@var{string}
4178 @opindex frandom-string
4179 This option provides a seed that GCC uses when it would otherwise use
4180 random numbers. It is used to generate certain symbol names
4181 that have to be different in every compiled file. It is also used to
4182 place unique stamps in coverage data files and the object files that
4183 produce them. You can use the @option{-frandom-seed} option to produce
4184 reproducibly identical object files.
4186 The @var{string} should be different for every file you compile.
4188 @item -fsched-verbose=@var{n}
4189 @opindex fsched-verbose
4190 On targets that use instruction scheduling, this option controls the
4191 amount of debugging output the scheduler prints. This information is
4192 written to standard error, unless @option{-dS} or @option{-dR} is
4193 specified, in which case it is output to the usual dump
4194 listing file, @file{.sched} or @file{.sched2} respectively. However
4195 for @var{n} greater than nine, the output is always printed to standard
4198 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4199 same information as @option{-dRS}. For @var{n} greater than one, it
4200 also output basic block probabilities, detailed ready list information
4201 and unit/insn info. For @var{n} greater than two, it includes RTL
4202 at abort point, control-flow and regions info. And for @var{n} over
4203 four, @option{-fsched-verbose} also includes dependence info.
4207 Store the usual ``temporary'' intermediate files permanently; place them
4208 in the current directory and name them based on the source file. Thus,
4209 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4210 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4211 preprocessed @file{foo.i} output file even though the compiler now
4212 normally uses an integrated preprocessor.
4214 When used in combination with the @option{-x} command line option,
4215 @option{-save-temps} is sensible enough to avoid over writing an
4216 input source file with the same extension as an intermediate file.
4217 The corresponding intermediate file may be obtained by renaming the
4218 source file before using @option{-save-temps}.
4222 Report the CPU time taken by each subprocess in the compilation
4223 sequence. For C source files, this is the compiler proper and assembler
4224 (plus the linker if linking is done). The output looks like this:
4231 The first number on each line is the ``user time'', that is time spent
4232 executing the program itself. The second number is ``system time'',
4233 time spent executing operating system routines on behalf of the program.
4234 Both numbers are in seconds.
4236 @item -fvar-tracking
4237 @opindex fvar-tracking
4238 Run variable tracking pass. It computes where variables are stored at each
4239 position in code. Better debugging information is then generated
4240 (if the debugging information format supports this information).
4242 It is enabled by default when compiling with optimization (@option{-Os},
4243 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4244 the debug info format supports it.
4246 @item -print-file-name=@var{library}
4247 @opindex print-file-name
4248 Print the full absolute name of the library file @var{library} that
4249 would be used when linking---and don't do anything else. With this
4250 option, GCC does not compile or link anything; it just prints the
4253 @item -print-multi-directory
4254 @opindex print-multi-directory
4255 Print the directory name corresponding to the multilib selected by any
4256 other switches present in the command line. This directory is supposed
4257 to exist in @env{GCC_EXEC_PREFIX}.
4259 @item -print-multi-lib
4260 @opindex print-multi-lib
4261 Print the mapping from multilib directory names to compiler switches
4262 that enable them. The directory name is separated from the switches by
4263 @samp{;}, and each switch starts with an @samp{@@} instead of the
4264 @samp{-}, without spaces between multiple switches. This is supposed to
4265 ease shell-processing.
4267 @item -print-prog-name=@var{program}
4268 @opindex print-prog-name
4269 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4271 @item -print-libgcc-file-name
4272 @opindex print-libgcc-file-name
4273 Same as @option{-print-file-name=libgcc.a}.
4275 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4276 but you do want to link with @file{libgcc.a}. You can do
4279 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4282 @item -print-search-dirs
4283 @opindex print-search-dirs
4284 Print the name of the configured installation directory and a list of
4285 program and library directories @command{gcc} will search---and don't do anything else.
4287 This is useful when @command{gcc} prints the error message
4288 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4289 To resolve this you either need to put @file{cpp0} and the other compiler
4290 components where @command{gcc} expects to find them, or you can set the environment
4291 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4292 Don't forget the trailing @samp{/}.
4293 @xref{Environment Variables}.
4296 @opindex dumpmachine
4297 Print the compiler's target machine (for example,
4298 @samp{i686-pc-linux-gnu})---and don't do anything else.
4301 @opindex dumpversion
4302 Print the compiler version (for example, @samp{3.0})---and don't do
4307 Print the compiler's built-in specs---and don't do anything else. (This
4308 is used when GCC itself is being built.) @xref{Spec Files}.
4310 @item -feliminate-unused-debug-types
4311 @opindex feliminate-unused-debug-types
4312 Normally, when producing DWARF2 output, GCC will emit debugging
4313 information for all types declared in a compilation
4314 unit, regardless of whether or not they are actually used
4315 in that compilation unit. Sometimes this is useful, such as
4316 if, in the debugger, you want to cast a value to a type that is
4317 not actually used in your program (but is declared). More often,
4318 however, this results in a significant amount of wasted space.
4319 With this option, GCC will avoid producing debug symbol output
4320 for types that are nowhere used in the source file being compiled.
4323 @node Optimize Options
4324 @section Options That Control Optimization
4325 @cindex optimize options
4326 @cindex options, optimization
4328 These options control various sorts of optimizations.
4330 Without any optimization option, the compiler's goal is to reduce the
4331 cost of compilation and to make debugging produce the expected
4332 results. Statements are independent: if you stop the program with a
4333 breakpoint between statements, you can then assign a new value to any
4334 variable or change the program counter to any other statement in the
4335 function and get exactly the results you would expect from the source
4338 Turning on optimization flags makes the compiler attempt to improve
4339 the performance and/or code size at the expense of compilation time
4340 and possibly the ability to debug the program.
4342 The compiler performs optimization based on the knowledge it has of
4343 the program. Optimization levels @option{-O2} and above, in
4344 particular, enable @emph{unit-at-a-time} mode, which allows the
4345 compiler to consider information gained from later functions in
4346 the file when compiling a function. Compiling multiple files at
4347 once to a single output file in @emph{unit-at-a-time} mode allows
4348 the compiler to use information gained from all of the files when
4349 compiling each of them.
4351 Not all optimizations are controlled directly by a flag. Only
4352 optimizations that have a flag are listed.
4359 Optimize. Optimizing compilation takes somewhat more time, and a lot
4360 more memory for a large function.
4362 With @option{-O}, the compiler tries to reduce code size and execution
4363 time, without performing any optimizations that take a great deal of
4366 @option{-O} turns on the following optimization flags:
4367 @gccoptlist{-fdefer-pop @gol
4368 -fdelayed-branch @gol
4369 -fguess-branch-probability @gol
4370 -fcprop-registers @gol
4371 -floop-optimize @gol
4372 -fif-conversion @gol
4373 -fif-conversion2 @gol
4376 -ftree-dominator-opts @gol
4381 -ftree-copyrename @gol
4386 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4387 where doing so does not interfere with debugging.
4391 Optimize even more. GCC performs nearly all supported optimizations
4392 that do not involve a space-speed tradeoff. The compiler does not
4393 perform loop unrolling or function inlining when you specify @option{-O2}.
4394 As compared to @option{-O}, this option increases both compilation time
4395 and the performance of the generated code.
4397 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4398 also turns on the following optimization flags:
4399 @gccoptlist{-fthread-jumps @gol
4401 -foptimize-sibling-calls @gol
4402 -fcse-follow-jumps -fcse-skip-blocks @gol
4403 -fgcse -fgcse-lm @gol
4404 -fexpensive-optimizations @gol
4405 -fstrength-reduce @gol
4406 -frerun-cse-after-loop -frerun-loop-opt @gol
4409 -fschedule-insns -fschedule-insns2 @gol
4410 -fsched-interblock -fsched-spec @gol
4412 -fstrict-aliasing @gol
4413 -fdelete-null-pointer-checks @gol
4414 -freorder-blocks -freorder-functions @gol
4415 -funit-at-a-time @gol
4416 -falign-functions -falign-jumps @gol
4417 -falign-loops -falign-labels @gol
4421 Please note the warning under @option{-fgcse} about
4422 invoking @option{-O2} on programs that use computed gotos.
4426 Optimize yet more. @option{-O3} turns on all optimizations specified by
4427 @option{-O2} and also turns on the @option{-finline-functions},
4428 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4432 Do not optimize. This is the default.
4436 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4437 do not typically increase code size. It also performs further
4438 optimizations designed to reduce code size.
4440 @option{-Os} disables the following optimization flags:
4441 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4442 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4443 -fprefetch-loop-arrays -ftree-vect-loop-version}
4445 If you use multiple @option{-O} options, with or without level numbers,
4446 the last such option is the one that is effective.
4449 Options of the form @option{-f@var{flag}} specify machine-independent
4450 flags. Most flags have both positive and negative forms; the negative
4451 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4452 below, only one of the forms is listed---the one you typically will
4453 use. You can figure out the other form by either removing @samp{no-}
4456 The following options control specific optimizations. They are either
4457 activated by @option{-O} options or are related to ones that are. You
4458 can use the following flags in the rare cases when ``fine-tuning'' of
4459 optimizations to be performed is desired.
4462 @item -fno-default-inline
4463 @opindex fno-default-inline
4464 Do not make member functions inline by default merely because they are
4465 defined inside the class scope (C++ only). Otherwise, when you specify
4466 @w{@option{-O}}, member functions defined inside class scope are compiled
4467 inline by default; i.e., you don't need to add @samp{inline} in front of
4468 the member function name.
4470 @item -fno-defer-pop
4471 @opindex fno-defer-pop
4472 Always pop the arguments to each function call as soon as that function
4473 returns. For machines which must pop arguments after a function call,
4474 the compiler normally lets arguments accumulate on the stack for several
4475 function calls and pops them all at once.
4477 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4481 Force memory operands to be copied into registers before doing
4482 arithmetic on them. This produces better code by making all memory
4483 references potential common subexpressions. When they are not common
4484 subexpressions, instruction combination should eliminate the separate
4485 register-load. This option is now a nop and will be removed in 4.2.
4488 @opindex fforce-addr
4489 Force memory address constants to be copied into registers before
4490 doing arithmetic on them.
4492 @item -fomit-frame-pointer
4493 @opindex fomit-frame-pointer
4494 Don't keep the frame pointer in a register for functions that
4495 don't need one. This avoids the instructions to save, set up and
4496 restore frame pointers; it also makes an extra register available
4497 in many functions. @strong{It also makes debugging impossible on
4500 On some machines, such as the VAX, this flag has no effect, because
4501 the standard calling sequence automatically handles the frame pointer
4502 and nothing is saved by pretending it doesn't exist. The
4503 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4504 whether a target machine supports this flag. @xref{Registers,,Register
4505 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4507 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4509 @item -foptimize-sibling-calls
4510 @opindex foptimize-sibling-calls
4511 Optimize sibling and tail recursive calls.
4513 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4517 Don't pay attention to the @code{inline} keyword. Normally this option
4518 is used to keep the compiler from expanding any functions inline.
4519 Note that if you are not optimizing, no functions can be expanded inline.
4521 @item -finline-functions
4522 @opindex finline-functions
4523 Integrate all simple functions into their callers. The compiler
4524 heuristically decides which functions are simple enough to be worth
4525 integrating in this way.
4527 If all calls to a given function are integrated, and the function is
4528 declared @code{static}, then the function is normally not output as
4529 assembler code in its own right.
4531 Enabled at level @option{-O3}.
4533 @item -finline-functions-called-once
4534 @opindex finline-functions-called-once
4535 Consider all @code{static} functions called once for inlining into their
4536 caller even if they are not marked @code{inline}. If a call to a given
4537 function is integrated, then the function is not output as assembler code
4540 Enabled if @option{-funit-at-a-time} is enabled.
4542 @item -fearly-inlining
4543 @opindex fearly-inlining
4544 Inline functions marked by @code{always_inline} and functions whose body seems
4545 smaller than the function call overhead early before doing
4546 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4547 makes profiling significantly cheaper and usually inlining faster on programs
4548 having large chains of nested wrapper functions.
4552 @item -finline-limit=@var{n}
4553 @opindex finline-limit
4554 By default, GCC limits the size of functions that can be inlined. This flag
4555 allows the control of this limit for functions that are explicitly marked as
4556 inline (i.e., marked with the inline keyword or defined within the class
4557 definition in c++). @var{n} is the size of functions that can be inlined in
4558 number of pseudo instructions (not counting parameter handling). The default
4559 value of @var{n} is 600.
4560 Increasing this value can result in more inlined code at
4561 the cost of compilation time and memory consumption. Decreasing usually makes
4562 the compilation faster and less code will be inlined (which presumably
4563 means slower programs). This option is particularly useful for programs that
4564 use inlining heavily such as those based on recursive templates with C++.
4566 Inlining is actually controlled by a number of parameters, which may be
4567 specified individually by using @option{--param @var{name}=@var{value}}.
4568 The @option{-finline-limit=@var{n}} option sets some of these parameters
4572 @item max-inline-insns-single
4573 is set to @var{n}/2.
4574 @item max-inline-insns-auto
4575 is set to @var{n}/2.
4576 @item min-inline-insns
4577 is set to 130 or @var{n}/4, whichever is smaller.
4578 @item max-inline-insns-rtl
4582 See below for a documentation of the individual
4583 parameters controlling inlining.
4585 @emph{Note:} pseudo instruction represents, in this particular context, an
4586 abstract measurement of function's size. In no way does it represent a count
4587 of assembly instructions and as such its exact meaning might change from one
4588 release to an another.
4590 @item -fkeep-inline-functions
4591 @opindex fkeep-inline-functions
4592 In C, emit @code{static} functions that are declared @code{inline}
4593 into the object file, even if the function has been inlined into all
4594 of its callers. This switch does not affect functions using the
4595 @code{extern inline} extension in GNU C@. In C++, emit any and all
4596 inline functions into the object file.
4598 @item -fkeep-static-consts
4599 @opindex fkeep-static-consts
4600 Emit variables declared @code{static const} when optimization isn't turned
4601 on, even if the variables aren't referenced.
4603 GCC enables this option by default. If you want to force the compiler to
4604 check if the variable was referenced, regardless of whether or not
4605 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4607 @item -fmerge-constants
4608 Attempt to merge identical constants (string constants and floating point
4609 constants) across compilation units.
4611 This option is the default for optimized compilation if the assembler and
4612 linker support it. Use @option{-fno-merge-constants} to inhibit this
4615 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4617 @item -fmerge-all-constants
4618 Attempt to merge identical constants and identical variables.
4620 This option implies @option{-fmerge-constants}. In addition to
4621 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4622 arrays or initialized constant variables with integral or floating point
4623 types. Languages like C or C++ require each non-automatic variable to
4624 have distinct location, so using this option will result in non-conforming
4627 @item -fmodulo-sched
4628 @opindex fmodulo-sched
4629 Perform swing modulo scheduling immediately before the first scheduling
4630 pass. This pass looks at innermost loops and reorders their
4631 instructions by overlapping different iterations.
4633 @item -fno-branch-count-reg
4634 @opindex fno-branch-count-reg
4635 Do not use ``decrement and branch'' instructions on a count register,
4636 but instead generate a sequence of instructions that decrement a
4637 register, compare it against zero, then branch based upon the result.
4638 This option is only meaningful on architectures that support such
4639 instructions, which include x86, PowerPC, IA-64 and S/390.
4641 The default is @option{-fbranch-count-reg}, enabled when
4642 @option{-fstrength-reduce} is enabled.
4644 @item -fno-function-cse
4645 @opindex fno-function-cse
4646 Do not put function addresses in registers; make each instruction that
4647 calls a constant function contain the function's address explicitly.
4649 This option results in less efficient code, but some strange hacks
4650 that alter the assembler output may be confused by the optimizations
4651 performed when this option is not used.
4653 The default is @option{-ffunction-cse}
4655 @item -fno-zero-initialized-in-bss
4656 @opindex fno-zero-initialized-in-bss
4657 If the target supports a BSS section, GCC by default puts variables that
4658 are initialized to zero into BSS@. This can save space in the resulting
4661 This option turns off this behavior because some programs explicitly
4662 rely on variables going to the data section. E.g., so that the
4663 resulting executable can find the beginning of that section and/or make
4664 assumptions based on that.
4666 The default is @option{-fzero-initialized-in-bss}.
4668 @item -fbounds-check
4669 @opindex fbounds-check
4670 For front-ends that support it, generate additional code to check that
4671 indices used to access arrays are within the declared range. This is
4672 currently only supported by the Java and Fortran front-ends, where
4673 this option defaults to true and false respectively.
4675 @item -fmudflap -fmudflapth -fmudflapir
4679 @cindex bounds checking
4681 For front-ends that support it (C and C++), instrument all risky
4682 pointer/array dereferencing operations, some standard library
4683 string/heap functions, and some other associated constructs with
4684 range/validity tests. Modules so instrumented should be immune to
4685 buffer overflows, invalid heap use, and some other classes of C/C++
4686 programming errors. The instrumentation relies on a separate runtime
4687 library (@file{libmudflap}), which will be linked into a program if
4688 @option{-fmudflap} is given at link time. Run-time behavior of the
4689 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4690 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4693 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4694 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4695 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4696 instrumentation should ignore pointer reads. This produces less
4697 instrumentation (and therefore faster execution) and still provides
4698 some protection against outright memory corrupting writes, but allows
4699 erroneously read data to propagate within a program.
4701 @item -fstrength-reduce
4702 @opindex fstrength-reduce
4703 Perform the optimizations of loop strength reduction and
4704 elimination of iteration variables.
4706 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4708 @item -fthread-jumps
4709 @opindex fthread-jumps
4710 Perform optimizations where we check to see if a jump branches to a
4711 location where another comparison subsumed by the first is found. If
4712 so, the first branch is redirected to either the destination of the
4713 second branch or a point immediately following it, depending on whether
4714 the condition is known to be true or false.
4716 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4718 @item -fcse-follow-jumps
4719 @opindex fcse-follow-jumps
4720 In common subexpression elimination, scan through jump instructions
4721 when the target of the jump is not reached by any other path. For
4722 example, when CSE encounters an @code{if} statement with an
4723 @code{else} clause, CSE will follow the jump when the condition
4726 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4728 @item -fcse-skip-blocks
4729 @opindex fcse-skip-blocks
4730 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4731 follow jumps which conditionally skip over blocks. When CSE
4732 encounters a simple @code{if} statement with no else clause,
4733 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4734 body of the @code{if}.
4736 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4738 @item -frerun-cse-after-loop
4739 @opindex frerun-cse-after-loop
4740 Re-run common subexpression elimination after loop optimizations has been
4743 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4745 @item -frerun-loop-opt
4746 @opindex frerun-loop-opt
4747 Run the loop optimizer twice.
4749 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4753 Perform a global common subexpression elimination pass.
4754 This pass also performs global constant and copy propagation.
4756 @emph{Note:} When compiling a program using computed gotos, a GCC
4757 extension, you may get better runtime performance if you disable
4758 the global common subexpression elimination pass by adding
4759 @option{-fno-gcse} to the command line.
4761 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4765 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4766 attempt to move loads which are only killed by stores into themselves. This
4767 allows a loop containing a load/store sequence to be changed to a load outside
4768 the loop, and a copy/store within the loop.
4770 Enabled by default when gcse is enabled.
4774 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4775 global common subexpression elimination. This pass will attempt to move
4776 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4777 loops containing a load/store sequence can be changed to a load before
4778 the loop and a store after the loop.
4780 Not enabled at any optimization level.
4784 When @option{-fgcse-las} is enabled, the global common subexpression
4785 elimination pass eliminates redundant loads that come after stores to the
4786 same memory location (both partial and full redundancies).
4788 Not enabled at any optimization level.
4790 @item -fgcse-after-reload
4791 @opindex fgcse-after-reload
4792 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4793 pass is performed after reload. The purpose of this pass is to cleanup
4796 @item -floop-optimize
4797 @opindex floop-optimize
4798 Perform loop optimizations: move constant expressions out of loops, simplify
4799 exit test conditions and optionally do strength-reduction as well.
4801 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4803 @item -floop-optimize2
4804 @opindex floop-optimize2
4805 Perform loop optimizations using the new loop optimizer. The optimizations
4806 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4809 @item -funsafe-loop-optimizations
4810 @opindex funsafe-loop-optimizations
4811 If given, the loop optimizer will assume that loop indices do not
4812 overflow, and that the loops with nontrivial exit condition are not
4813 infinite. This enables a wider range of loop optimizations even if
4814 the loop optimizer itself cannot prove that these assumptions are valid.
4815 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4816 if it finds this kind of loop.
4818 @item -fcrossjumping
4819 @opindex crossjumping
4820 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4821 resulting code may or may not perform better than without cross-jumping.
4823 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4825 @item -fif-conversion
4826 @opindex if-conversion
4827 Attempt to transform conditional jumps into branch-less equivalents. This
4828 include use of conditional moves, min, max, set flags and abs instructions, and
4829 some tricks doable by standard arithmetics. The use of conditional execution
4830 on chips where it is available is controlled by @code{if-conversion2}.
4832 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4834 @item -fif-conversion2
4835 @opindex if-conversion2
4836 Use conditional execution (where available) to transform conditional jumps into
4837 branch-less equivalents.
4839 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4841 @item -fdelete-null-pointer-checks
4842 @opindex fdelete-null-pointer-checks
4843 Use global dataflow analysis to identify and eliminate useless checks
4844 for null pointers. The compiler assumes that dereferencing a null
4845 pointer would have halted the program. If a pointer is checked after
4846 it has already been dereferenced, it cannot be null.
4848 In some environments, this assumption is not true, and programs can
4849 safely dereference null pointers. Use
4850 @option{-fno-delete-null-pointer-checks} to disable this optimization
4851 for programs which depend on that behavior.
4853 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4855 @item -fexpensive-optimizations
4856 @opindex fexpensive-optimizations
4857 Perform a number of minor optimizations that are relatively expensive.
4859 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4861 @item -foptimize-register-move
4863 @opindex foptimize-register-move
4865 Attempt to reassign register numbers in move instructions and as
4866 operands of other simple instructions in order to maximize the amount of
4867 register tying. This is especially helpful on machines with two-operand
4870 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4873 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4875 @item -fdelayed-branch
4876 @opindex fdelayed-branch
4877 If supported for the target machine, attempt to reorder instructions
4878 to exploit instruction slots available after delayed branch
4881 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4883 @item -fschedule-insns
4884 @opindex fschedule-insns
4885 If supported for the target machine, attempt to reorder instructions to
4886 eliminate execution stalls due to required data being unavailable. This
4887 helps machines that have slow floating point or memory load instructions
4888 by allowing other instructions to be issued until the result of the load
4889 or floating point instruction is required.
4891 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4893 @item -fschedule-insns2
4894 @opindex fschedule-insns2
4895 Similar to @option{-fschedule-insns}, but requests an additional pass of
4896 instruction scheduling after register allocation has been done. This is
4897 especially useful on machines with a relatively small number of
4898 registers and where memory load instructions take more than one cycle.
4900 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4902 @item -fno-sched-interblock
4903 @opindex fno-sched-interblock
4904 Don't schedule instructions across basic blocks. This is normally
4905 enabled by default when scheduling before register allocation, i.e.@:
4906 with @option{-fschedule-insns} or at @option{-O2} or higher.
4908 @item -fno-sched-spec
4909 @opindex fno-sched-spec
4910 Don't allow speculative motion of non-load instructions. This is normally
4911 enabled by default when scheduling before register allocation, i.e.@:
4912 with @option{-fschedule-insns} or at @option{-O2} or higher.
4914 @item -fsched-spec-load
4915 @opindex fsched-spec-load
4916 Allow speculative motion of some load instructions. This only makes
4917 sense when scheduling before register allocation, i.e.@: with
4918 @option{-fschedule-insns} or at @option{-O2} or higher.
4920 @item -fsched-spec-load-dangerous
4921 @opindex fsched-spec-load-dangerous
4922 Allow speculative motion of more load instructions. This only makes
4923 sense when scheduling before register allocation, i.e.@: with
4924 @option{-fschedule-insns} or at @option{-O2} or higher.
4926 @item -fsched-stalled-insns=@var{n}
4927 @opindex fsched-stalled-insns
4928 Define how many insns (if any) can be moved prematurely from the queue
4929 of stalled insns into the ready list, during the second scheduling pass.
4931 @item -fsched-stalled-insns-dep=@var{n}
4932 @opindex fsched-stalled-insns-dep
4933 Define how many insn groups (cycles) will be examined for a dependency
4934 on a stalled insn that is candidate for premature removal from the queue
4935 of stalled insns. Has an effect only during the second scheduling pass,
4936 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4938 @item -fsched2-use-superblocks
4939 @opindex fsched2-use-superblocks
4940 When scheduling after register allocation, do use superblock scheduling
4941 algorithm. Superblock scheduling allows motion across basic block boundaries
4942 resulting on faster schedules. This option is experimental, as not all machine
4943 descriptions used by GCC model the CPU closely enough to avoid unreliable
4944 results from the algorithm.
4946 This only makes sense when scheduling after register allocation, i.e.@: with
4947 @option{-fschedule-insns2} or at @option{-O2} or higher.
4949 @item -fsched2-use-traces
4950 @opindex fsched2-use-traces
4951 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4952 allocation and additionally perform code duplication in order to increase the
4953 size of superblocks using tracer pass. See @option{-ftracer} for details on
4956 This mode should produce faster but significantly longer programs. Also
4957 without @option{-fbranch-probabilities} the traces constructed may not
4958 match the reality and hurt the performance. This only makes
4959 sense when scheduling after register allocation, i.e.@: with
4960 @option{-fschedule-insns2} or at @option{-O2} or higher.
4962 @item -freschedule-modulo-scheduled-loops
4963 @opindex fscheduling-in-modulo-scheduled-loops
4964 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4965 we may want to prevent the later scheduling passes from changing its schedule, we use this
4966 option to control that.
4968 @item -fcaller-saves
4969 @opindex fcaller-saves
4970 Enable values to be allocated in registers that will be clobbered by
4971 function calls, by emitting extra instructions to save and restore the
4972 registers around such calls. Such allocation is done only when it
4973 seems to result in better code than would otherwise be produced.
4975 This option is always enabled by default on certain machines, usually
4976 those which have no call-preserved registers to use instead.
4978 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4981 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4982 enabled by default at @option{-O2} and @option{-O3}.
4985 Perform Full Redundancy Elimination (FRE) on trees. The difference
4986 between FRE and PRE is that FRE only considers expressions
4987 that are computed on all paths leading to the redundant computation.
4988 This analysis faster than PRE, though it exposes fewer redundancies.
4989 This flag is enabled by default at @option{-O} and higher.
4991 @item -ftree-copy-prop
4992 Perform copy propagation on trees. This pass eliminates unnecessary
4993 copy operations. This flag is enabled by default at @option{-O} and
4996 @item -ftree-store-copy-prop
4997 Perform copy propagation of memory loads and stores. This pass
4998 eliminates unnecessary copy operations in memory references
4999 (structures, global variables, arrays, etc). This flag is enabled by
5000 default at @option{-O2} and higher.
5003 Perform structural alias analysis on trees. This flag
5004 is enabled by default at @option{-O} and higher.
5007 Perform forward store motion on trees. This flag is
5008 enabled by default at @option{-O} and higher.
5011 Perform sparse conditional constant propagation (CCP) on trees. This
5012 pass only operates on local scalar variables and is enabled by default
5013 at @option{-O} and higher.
5015 @item -ftree-store-ccp
5016 Perform sparse conditional constant propagation (CCP) on trees. This
5017 pass operates on both local scalar variables and memory stores and
5018 loads (global variables, structures, arrays, etc). This flag is
5019 enabled by default at @option{-O2} and higher.
5022 Perform dead code elimination (DCE) on trees. This flag is enabled by
5023 default at @option{-O} and higher.
5025 @item -ftree-dominator-opts
5026 Perform a variety of simple scalar cleanups (constant/copy
5027 propagation, redundancy elimination, range propagation and expression
5028 simplification) based on a dominator tree traversal. This also
5029 performs jump threading (to reduce jumps to jumps). This flag is
5030 enabled by default at @option{-O} and higher.
5033 Perform loop header copying on trees. This is beneficial since it increases
5034 effectiveness of code motion optimizations. It also saves one jump. This flag
5035 is enabled by default at @option{-O} and higher. It is not enabled
5036 for @option{-Os}, since it usually increases code size.
5038 @item -ftree-loop-optimize
5039 Perform loop optimizations on trees. This flag is enabled by default
5040 at @option{-O} and higher.
5042 @item -ftree-loop-linear
5043 Perform linear loop transformations on tree. This flag can improve cache
5044 performance and allow further loop optimizations to take place.
5046 @item -ftree-loop-im
5047 Perform loop invariant motion on trees. This pass moves only invariants that
5048 would be hard to handle at RTL level (function calls, operations that expand to
5049 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5050 operands of conditions that are invariant out of the loop, so that we can use
5051 just trivial invariantness analysis in loop unswitching. The pass also includes
5054 @item -ftree-loop-ivcanon
5055 Create a canonical counter for number of iterations in the loop for that
5056 determining number of iterations requires complicated analysis. Later
5057 optimizations then may determine the number easily. Useful especially
5058 in connection with unrolling.
5061 Perform induction variable optimizations (strength reduction, induction
5062 variable merging and induction variable elimination) on trees.
5065 Perform scalar replacement of aggregates. This pass replaces structure
5066 references with scalars to prevent committing structures to memory too
5067 early. This flag is enabled by default at @option{-O} and higher.
5069 @item -ftree-copyrename
5070 Perform copy renaming on trees. This pass attempts to rename compiler
5071 temporaries to other variables at copy locations, usually resulting in
5072 variable names which more closely resemble the original variables. This flag
5073 is enabled by default at @option{-O} and higher.
5076 Perform temporary expression replacement during the SSA->normal phase. Single
5077 use/single def temporaries are replaced at their use location with their
5078 defining expression. This results in non-GIMPLE code, but gives the expanders
5079 much more complex trees to work on resulting in better RTL generation. This is
5080 enabled by default at @option{-O} and higher.
5083 Perform live range splitting during the SSA->normal phase. Distinct live
5084 ranges of a variable are split into unique variables, allowing for better
5085 optimization later. This is enabled by default at @option{-O} and higher.
5087 @item -ftree-vectorize
5088 Perform loop vectorization on trees.
5090 @item -ftree-vect-loop-version
5091 @opindex ftree-vect-loop-version
5092 Perform loop versioning when doing loop vectorization on trees. When a loop
5093 appears to be vectorizable except that data alignment or data dependence cannot
5094 be determined at compile time then vectorized and non-vectorized versions of
5095 the loop are generated along with runtime checks for alignment or dependence
5096 to control which version is executed. This option is enabled by default
5097 except at level @option{-Os} where it is disabled.
5100 Perform Value Range Propagation on trees. This is similar to the
5101 constant propagation pass, but instead of values, ranges of values are
5102 propagated. This allows the optimizers to remove unnecessary range
5103 checks like array bound checks and null pointer checks. This is
5104 enabled by default at @option{-O2} and higher. Null pointer check
5105 elimination is only done if @option{-fdelete-null-pointer-checks} is
5110 Perform tail duplication to enlarge superblock size. This transformation
5111 simplifies the control flow of the function allowing other optimizations to do
5114 @item -funroll-loops
5115 @opindex funroll-loops
5116 Unroll loops whose number of iterations can be determined at compile
5117 time or upon entry to the loop. @option{-funroll-loops} implies both
5118 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
5119 option makes code larger, and may or may not make it run faster.
5121 @item -funroll-all-loops
5122 @opindex funroll-all-loops
5123 Unroll all loops, even if their number of iterations is uncertain when
5124 the loop is entered. This usually makes programs run more slowly.
5125 @option{-funroll-all-loops} implies the same options as
5126 @option{-funroll-loops},
5128 @item -fsplit-ivs-in-unroller
5129 @opindex -fsplit-ivs-in-unroller
5130 Enables expressing of values of induction variables in later iterations
5131 of the unrolled loop using the value in the first iteration. This breaks
5132 long dependency chains, thus improving efficiency of the scheduling passes.
5134 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5135 same effect. However in cases the loop body is more complicated than
5136 a single basic block, this is not reliable. It also does not work at all
5137 on some of the architectures due to restrictions in the CSE pass.
5139 This optimization is enabled by default.
5141 @item -fvariable-expansion-in-unroller
5142 @opindex -fvariable-expansion-in-unroller
5143 With this option, the compiler will create multiple copies of some
5144 local variables when unrolling a loop which can result in superior code.
5146 @item -fprefetch-loop-arrays
5147 @opindex fprefetch-loop-arrays
5148 If supported by the target machine, generate instructions to prefetch
5149 memory to improve the performance of loops that access large arrays.
5151 These options may generate better or worse code; results are highly
5152 dependent on the structure of loops within the source code.
5155 @itemx -fno-peephole2
5156 @opindex fno-peephole
5157 @opindex fno-peephole2
5158 Disable any machine-specific peephole optimizations. The difference
5159 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5160 are implemented in the compiler; some targets use one, some use the
5161 other, a few use both.
5163 @option{-fpeephole} is enabled by default.
5164 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5166 @item -fno-guess-branch-probability
5167 @opindex fno-guess-branch-probability
5168 Do not guess branch probabilities using heuristics.
5170 GCC will use heuristics to guess branch probabilities if they are
5171 not provided by profiling feedback (@option{-fprofile-arcs}). These
5172 heuristics are based on the control flow graph. If some branch probabilities
5173 are specified by @samp{__builtin_expect}, then the heuristics will be
5174 used to guess branch probabilities for the rest of the control flow graph,
5175 taking the @samp{__builtin_expect} info into account. The interactions
5176 between the heuristics and @samp{__builtin_expect} can be complex, and in
5177 some cases, it may be useful to disable the heuristics so that the effects
5178 of @samp{__builtin_expect} are easier to understand.
5180 The default is @option{-fguess-branch-probability} at levels
5181 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5183 @item -freorder-blocks
5184 @opindex freorder-blocks
5185 Reorder basic blocks in the compiled function in order to reduce number of
5186 taken branches and improve code locality.
5188 Enabled at levels @option{-O2}, @option{-O3}.
5190 @item -freorder-blocks-and-partition
5191 @opindex freorder-blocks-and-partition
5192 In addition to reordering basic blocks in the compiled function, in order
5193 to reduce number of taken branches, partitions hot and cold basic blocks
5194 into separate sections of the assembly and .o files, to improve
5195 paging and cache locality performance.
5197 This optimization is automatically turned off in the presence of
5198 exception handling, for linkonce sections, for functions with a user-defined
5199 section attribute and on any architecture that does not support named
5202 @item -freorder-functions
5203 @opindex freorder-functions
5204 Reorder functions in the object file in order to
5205 improve code locality. This is implemented by using special
5206 subsections @code{.text.hot} for most frequently executed functions and
5207 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5208 the linker so object file format must support named sections and linker must
5209 place them in a reasonable way.
5211 Also profile feedback must be available in to make this option effective. See
5212 @option{-fprofile-arcs} for details.
5214 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5216 @item -fstrict-aliasing
5217 @opindex fstrict-aliasing
5218 Allows the compiler to assume the strictest aliasing rules applicable to
5219 the language being compiled. For C (and C++), this activates
5220 optimizations based on the type of expressions. In particular, an
5221 object of one type is assumed never to reside at the same address as an
5222 object of a different type, unless the types are almost the same. For
5223 example, an @code{unsigned int} can alias an @code{int}, but not a
5224 @code{void*} or a @code{double}. A character type may alias any other
5227 Pay special attention to code like this:
5240 The practice of reading from a different union member than the one most
5241 recently written to (called ``type-punning'') is common. Even with
5242 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5243 is accessed through the union type. So, the code above will work as
5244 expected. However, this code might not:
5255 Every language that wishes to perform language-specific alias analysis
5256 should define a function that computes, given an @code{tree}
5257 node, an alias set for the node. Nodes in different alias sets are not
5258 allowed to alias. For an example, see the C front-end function
5259 @code{c_get_alias_set}.
5261 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5263 @item -falign-functions
5264 @itemx -falign-functions=@var{n}
5265 @opindex falign-functions
5266 Align the start of functions to the next power-of-two greater than
5267 @var{n}, skipping up to @var{n} bytes. For instance,
5268 @option{-falign-functions=32} aligns functions to the next 32-byte
5269 boundary, but @option{-falign-functions=24} would align to the next
5270 32-byte boundary only if this can be done by skipping 23 bytes or less.
5272 @option{-fno-align-functions} and @option{-falign-functions=1} are
5273 equivalent and mean that functions will not be aligned.
5275 Some assemblers only support this flag when @var{n} is a power of two;
5276 in that case, it is rounded up.
5278 If @var{n} is not specified or is zero, use a machine-dependent default.
5280 Enabled at levels @option{-O2}, @option{-O3}.
5282 @item -falign-labels
5283 @itemx -falign-labels=@var{n}
5284 @opindex falign-labels
5285 Align all branch targets to a power-of-two boundary, skipping up to
5286 @var{n} bytes like @option{-falign-functions}. This option can easily
5287 make code slower, because it must insert dummy operations for when the
5288 branch target is reached in the usual flow of the code.
5290 @option{-fno-align-labels} and @option{-falign-labels=1} are
5291 equivalent and mean that labels will not be aligned.
5293 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5294 are greater than this value, then their values are used instead.
5296 If @var{n} is not specified or is zero, use a machine-dependent default
5297 which is very likely to be @samp{1}, meaning no alignment.
5299 Enabled at levels @option{-O2}, @option{-O3}.
5302 @itemx -falign-loops=@var{n}
5303 @opindex falign-loops
5304 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5305 like @option{-falign-functions}. The hope is that the loop will be
5306 executed many times, which will make up for any execution of the dummy
5309 @option{-fno-align-loops} and @option{-falign-loops=1} are
5310 equivalent and mean that loops will not be aligned.
5312 If @var{n} is not specified or is zero, use a machine-dependent default.
5314 Enabled at levels @option{-O2}, @option{-O3}.
5317 @itemx -falign-jumps=@var{n}
5318 @opindex falign-jumps
5319 Align branch targets to a power-of-two boundary, for branch targets
5320 where the targets can only be reached by jumping, skipping up to @var{n}
5321 bytes like @option{-falign-functions}. In this case, no dummy operations
5324 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5325 equivalent and mean that loops will not be aligned.
5327 If @var{n} is not specified or is zero, use a machine-dependent default.
5329 Enabled at levels @option{-O2}, @option{-O3}.
5331 @item -funit-at-a-time
5332 @opindex funit-at-a-time
5333 Parse the whole compilation unit before starting to produce code.
5334 This allows some extra optimizations to take place but consumes
5335 more memory (in general). There are some compatibility issues
5336 with @emph{unit-at-at-time} mode:
5339 enabling @emph{unit-at-a-time} mode may change the order
5340 in which functions, variables, and top-level @code{asm} statements
5341 are emitted, and will likely break code relying on some particular
5342 ordering. The majority of such top-level @code{asm} statements,
5343 though, can be replaced by @code{section} attributes.
5346 @emph{unit-at-a-time} mode removes unreferenced static variables
5347 and functions. This may result in undefined references
5348 when an @code{asm} statement refers directly to variables or functions
5349 that are otherwise unused. In that case either the variable/function
5350 shall be listed as an operand of the @code{asm} statement operand or,
5351 in the case of top-level @code{asm} statements the attribute @code{used}
5352 shall be used on the declaration.
5355 Static functions now can use non-standard passing conventions that
5356 may break @code{asm} statements calling functions directly. Again,
5357 attribute @code{used} will prevent this behavior.
5360 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5361 but this scheme may not be supported by future releases of GCC@.
5363 Enabled at levels @option{-O2}, @option{-O3}.
5367 Constructs webs as commonly used for register allocation purposes and assign
5368 each web individual pseudo register. This allows the register allocation pass
5369 to operate on pseudos directly, but also strengthens several other optimization
5370 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5371 however, make debugging impossible, since variables will no longer stay in a
5374 Enabled by default with @option{-funroll-loops}.
5376 @item -fwhole-program
5377 @opindex fwhole-program
5378 Assume that the current compilation unit represents whole program being
5379 compiled. All public functions and variables with the exception of @code{main}
5380 and those merged by attribute @code{externally_visible} become static functions
5381 and in a affect gets more aggressively optimized by interprocedural optimizers.
5382 While this option is equivalent to proper use of @code{static} keyword for
5383 programs consisting of single file, in combination with option
5384 @option{--combine} this flag can be used to compile most of smaller scale C
5385 programs since the functions and variables become local for the whole combined
5386 compilation unit, not for the single source file itself.
5389 @item -fno-cprop-registers
5390 @opindex fno-cprop-registers
5391 After register allocation and post-register allocation instruction splitting,
5392 we perform a copy-propagation pass to try to reduce scheduling dependencies
5393 and occasionally eliminate the copy.
5395 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5397 @item -fprofile-generate
5398 @opindex fprofile-generate
5400 Enable options usually used for instrumenting application to produce
5401 profile useful for later recompilation with profile feedback based
5402 optimization. You must use @option{-fprofile-generate} both when
5403 compiling and when linking your program.
5405 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5408 @opindex fprofile-use
5409 Enable profile feedback directed optimizations, and optimizations
5410 generally profitable only with profile feedback available.
5412 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5413 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer},
5414 @code{-fno-loop-optimize}.
5418 The following options control compiler behavior regarding floating
5419 point arithmetic. These options trade off between speed and
5420 correctness. All must be specifically enabled.
5424 @opindex ffloat-store
5425 Do not store floating point variables in registers, and inhibit other
5426 options that might change whether a floating point value is taken from a
5429 @cindex floating point precision
5430 This option prevents undesirable excess precision on machines such as
5431 the 68000 where the floating registers (of the 68881) keep more
5432 precision than a @code{double} is supposed to have. Similarly for the
5433 x86 architecture. For most programs, the excess precision does only
5434 good, but a few programs rely on the precise definition of IEEE floating
5435 point. Use @option{-ffloat-store} for such programs, after modifying
5436 them to store all pertinent intermediate computations into variables.
5440 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5441 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5442 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5443 and @option{fcx-limited-range}.
5445 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5447 This option should never be turned on by any @option{-O} option since
5448 it can result in incorrect output for programs which depend on
5449 an exact implementation of IEEE or ISO rules/specifications for
5452 @item -fno-math-errno
5453 @opindex fno-math-errno
5454 Do not set ERRNO after calling math functions that are executed
5455 with a single instruction, e.g., sqrt. A program that relies on
5456 IEEE exceptions for math error handling may want to use this flag
5457 for speed while maintaining IEEE arithmetic compatibility.
5459 This option should never be turned on by any @option{-O} option since
5460 it can result in incorrect output for programs which depend on
5461 an exact implementation of IEEE or ISO rules/specifications for
5464 The default is @option{-fmath-errno}.
5466 On Darwin systems, the math library never sets @code{errno}. There is therefore
5467 no reason for the compiler to consider the possibility that it might,
5468 and @option{-fno-math-errno} is the default.
5470 @item -funsafe-math-optimizations
5471 @opindex funsafe-math-optimizations
5472 Allow optimizations for floating-point arithmetic that (a) assume
5473 that arguments and results are valid and (b) may violate IEEE or
5474 ANSI standards. When used at link-time, it may include libraries
5475 or startup files that change the default FPU control word or other
5476 similar optimizations.
5478 This option should never be turned on by any @option{-O} option since
5479 it can result in incorrect output for programs which depend on
5480 an exact implementation of IEEE or ISO rules/specifications for
5483 The default is @option{-fno-unsafe-math-optimizations}.
5485 @item -ffinite-math-only
5486 @opindex ffinite-math-only
5487 Allow optimizations for floating-point arithmetic that assume
5488 that arguments and results are not NaNs or +-Infs.
5490 This option should never be turned on by any @option{-O} option since
5491 it can result in incorrect output for programs which depend on
5492 an exact implementation of IEEE or ISO rules/specifications.
5494 The default is @option{-fno-finite-math-only}.
5496 @item -fno-trapping-math
5497 @opindex fno-trapping-math
5498 Compile code assuming that floating-point operations cannot generate
5499 user-visible traps. These traps include division by zero, overflow,
5500 underflow, inexact result and invalid operation. This option implies
5501 @option{-fno-signaling-nans}. Setting this option may allow faster
5502 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5504 This option should never be turned on by any @option{-O} option since
5505 it can result in incorrect output for programs which depend on
5506 an exact implementation of IEEE or ISO rules/specifications for
5509 The default is @option{-ftrapping-math}.
5511 @item -frounding-math
5512 @opindex frounding-math
5513 Disable transformations and optimizations that assume default floating
5514 point rounding behavior. This is round-to-zero for all floating point
5515 to integer conversions, and round-to-nearest for all other arithmetic
5516 truncations. This option should be specified for programs that change
5517 the FP rounding mode dynamically, or that may be executed with a
5518 non-default rounding mode. This option disables constant folding of
5519 floating point expressions at compile-time (which may be affected by
5520 rounding mode) and arithmetic transformations that are unsafe in the
5521 presence of sign-dependent rounding modes.
5523 The default is @option{-fno-rounding-math}.
5525 This option is experimental and does not currently guarantee to
5526 disable all GCC optimizations that are affected by rounding mode.
5527 Future versions of GCC may provide finer control of this setting
5528 using C99's @code{FENV_ACCESS} pragma. This command line option
5529 will be used to specify the default state for @code{FENV_ACCESS}.
5531 @item -fsignaling-nans
5532 @opindex fsignaling-nans
5533 Compile code assuming that IEEE signaling NaNs may generate user-visible
5534 traps during floating-point operations. Setting this option disables
5535 optimizations that may change the number of exceptions visible with
5536 signaling NaNs. This option implies @option{-ftrapping-math}.
5538 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5541 The default is @option{-fno-signaling-nans}.
5543 This option is experimental and does not currently guarantee to
5544 disable all GCC optimizations that affect signaling NaN behavior.
5546 @item -fsingle-precision-constant
5547 @opindex fsingle-precision-constant
5548 Treat floating point constant as single precision constant instead of
5549 implicitly converting it to double precision constant.
5551 @item -fcx-limited-range
5552 @itemx -fno-cx-limited-range
5553 @opindex fcx-limited-range
5554 @opindex fno-cx-limited-range
5555 When enabled, this option states that a range reduction step is not
5556 needed when performing complex division. The default is
5557 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5559 This option controls the default setting of the ISO C99
5560 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5565 The following options control optimizations that may improve
5566 performance, but are not enabled by any @option{-O} options. This
5567 section includes experimental options that may produce broken code.
5570 @item -fbranch-probabilities
5571 @opindex fbranch-probabilities
5572 After running a program compiled with @option{-fprofile-arcs}
5573 (@pxref{Debugging Options,, Options for Debugging Your Program or
5574 @command{gcc}}), you can compile it a second time using
5575 @option{-fbranch-probabilities}, to improve optimizations based on
5576 the number of times each branch was taken. When the program
5577 compiled with @option{-fprofile-arcs} exits it saves arc execution
5578 counts to a file called @file{@var{sourcename}.gcda} for each source
5579 file The information in this data file is very dependent on the
5580 structure of the generated code, so you must use the same source code
5581 and the same optimization options for both compilations.
5583 With @option{-fbranch-probabilities}, GCC puts a
5584 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5585 These can be used to improve optimization. Currently, they are only
5586 used in one place: in @file{reorg.c}, instead of guessing which path a
5587 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5588 exactly determine which path is taken more often.
5590 @item -fprofile-values
5591 @opindex fprofile-values
5592 If combined with @option{-fprofile-arcs}, it adds code so that some
5593 data about values of expressions in the program is gathered.
5595 With @option{-fbranch-probabilities}, it reads back the data gathered
5596 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5597 notes to instructions for their later usage in optimizations.
5599 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5603 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5604 a code to gather information about values of expressions.
5606 With @option{-fbranch-probabilities}, it reads back the data gathered
5607 and actually performs the optimizations based on them.
5608 Currently the optimizations include specialization of division operation
5609 using the knowledge about the value of the denominator.
5611 @item -frename-registers
5612 @opindex frename-registers
5613 Attempt to avoid false dependencies in scheduled code by making use
5614 of registers left over after register allocation. This optimization
5615 will most benefit processors with lots of registers. Depending on the
5616 debug information format adopted by the target, however, it can
5617 make debugging impossible, since variables will no longer stay in
5618 a ``home register''.
5620 Enabled by default with @option{-funroll-loops}.
5624 Perform tail duplication to enlarge superblock size. This transformation
5625 simplifies the control flow of the function allowing other optimizations to do
5628 Enabled with @option{-fprofile-use}.
5630 @item -funroll-loops
5631 @opindex funroll-loops
5632 Unroll loops whose number of iterations can be determined at compile time or
5633 upon entry to the loop. @option{-funroll-loops} implies
5634 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5635 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5636 small constant number of iterations). This option makes code larger, and may
5637 or may not make it run faster.
5639 Enabled with @option{-fprofile-use}.
5641 @item -funroll-all-loops
5642 @opindex funroll-all-loops
5643 Unroll all loops, even if their number of iterations is uncertain when
5644 the loop is entered. This usually makes programs run more slowly.
5645 @option{-funroll-all-loops} implies the same options as
5646 @option{-funroll-loops}.
5649 @opindex fpeel-loops
5650 Peels the loops for that there is enough information that they do not
5651 roll much (from profile feedback). It also turns on complete loop peeling
5652 (i.e.@: complete removal of loops with small constant number of iterations).
5654 Enabled with @option{-fprofile-use}.
5656 @item -fmove-loop-invariants
5657 @opindex fmove-loop-invariants
5658 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5659 at level @option{-O1}
5661 @item -funswitch-loops
5662 @opindex funswitch-loops
5663 Move branches with loop invariant conditions out of the loop, with duplicates
5664 of the loop on both branches (modified according to result of the condition).
5666 @item -fprefetch-loop-arrays
5667 @opindex fprefetch-loop-arrays
5668 If supported by the target machine, generate instructions to prefetch
5669 memory to improve the performance of loops that access large arrays.
5671 Disabled at level @option{-Os}.
5673 @item -ffunction-sections
5674 @itemx -fdata-sections
5675 @opindex ffunction-sections
5676 @opindex fdata-sections
5677 Place each function or data item into its own section in the output
5678 file if the target supports arbitrary sections. The name of the
5679 function or the name of the data item determines the section's name
5682 Use these options on systems where the linker can perform optimizations
5683 to improve locality of reference in the instruction space. Most systems
5684 using the ELF object format and SPARC processors running Solaris 2 have
5685 linkers with such optimizations. AIX may have these optimizations in
5688 Only use these options when there are significant benefits from doing
5689 so. When you specify these options, the assembler and linker will
5690 create larger object and executable files and will also be slower.
5691 You will not be able to use @code{gprof} on all systems if you
5692 specify this option and you may have problems with debugging if
5693 you specify both this option and @option{-g}.
5695 @item -fbranch-target-load-optimize
5696 @opindex fbranch-target-load-optimize
5697 Perform branch target register load optimization before prologue / epilogue
5699 The use of target registers can typically be exposed only during reload,
5700 thus hoisting loads out of loops and doing inter-block scheduling needs
5701 a separate optimization pass.
5703 @item -fbranch-target-load-optimize2
5704 @opindex fbranch-target-load-optimize2
5705 Perform branch target register load optimization after prologue / epilogue
5708 @item -fbtr-bb-exclusive
5709 @opindex fbtr-bb-exclusive
5710 When performing branch target register load optimization, don't reuse
5711 branch target registers in within any basic block.
5713 @item -fstack-protector
5714 Emit extra code to check for buffer overflows, such as stack smashing
5715 attacks. This is done by adding a guard variable to functions with
5716 vulnerable objects. This includes functions that call alloca, and
5717 functions with buffers larger than 8 bytes. The guards are initialized
5718 when a function is entered and then checked when the function exits.
5719 If a guard check fails, an error message is printed and the program exits.
5721 @item -fstack-protector-all
5722 Like @option{-fstack-protector} except that all functions are protected.
5724 @item --param @var{name}=@var{value}
5726 In some places, GCC uses various constants to control the amount of
5727 optimization that is done. For example, GCC will not inline functions
5728 that contain more that a certain number of instructions. You can
5729 control some of these constants on the command-line using the
5730 @option{--param} option.
5732 The names of specific parameters, and the meaning of the values, are
5733 tied to the internals of the compiler, and are subject to change
5734 without notice in future releases.
5736 In each case, the @var{value} is an integer. The allowable choices for
5737 @var{name} are given in the following table:
5740 @item salias-max-implicit-fields
5741 The maximum number of fields in a variable without direct
5742 structure accesses for which structure aliasing will consider trying
5743 to track each field. The default is 5
5745 @item sra-max-structure-size
5746 The maximum structure size, in bytes, at which the scalar replacement
5747 of aggregates (SRA) optimization will perform block copies. The
5748 default value, 0, implies that GCC will select the most appropriate
5751 @item sra-field-structure-ratio
5752 The threshold ratio (as a percentage) between instantiated fields and
5753 the complete structure size. We say that if the ratio of the number
5754 of bytes in instantiated fields to the number of bytes in the complete
5755 structure exceeds this parameter, then block copies are not used. The
5758 @item max-crossjump-edges
5759 The maximum number of incoming edges to consider for crossjumping.
5760 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5761 the number of edges incoming to each block. Increasing values mean
5762 more aggressive optimization, making the compile time increase with
5763 probably small improvement in executable size.
5765 @item min-crossjump-insns
5766 The minimum number of instructions which must be matched at the end
5767 of two blocks before crossjumping will be performed on them. This
5768 value is ignored in the case where all instructions in the block being
5769 crossjumped from are matched. The default value is 5.
5771 @item max-grow-copy-bb-insns
5772 The maximum code size expansion factor when copying basic blocks
5773 instead of jumping. The expansion is relative to a jump instruction.
5774 The default value is 8.
5776 @item max-goto-duplication-insns
5777 The maximum number of instructions to duplicate to a block that jumps
5778 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5779 passes, GCC factors computed gotos early in the compilation process,
5780 and unfactors them as late as possible. Only computed jumps at the
5781 end of a basic blocks with no more than max-goto-duplication-insns are
5782 unfactored. The default value is 8.
5784 @item max-delay-slot-insn-search
5785 The maximum number of instructions to consider when looking for an
5786 instruction to fill a delay slot. If more than this arbitrary number of
5787 instructions is searched, the time savings from filling the delay slot
5788 will be minimal so stop searching. Increasing values mean more
5789 aggressive optimization, making the compile time increase with probably
5790 small improvement in executable run time.
5792 @item max-delay-slot-live-search
5793 When trying to fill delay slots, the maximum number of instructions to
5794 consider when searching for a block with valid live register
5795 information. Increasing this arbitrarily chosen value means more
5796 aggressive optimization, increasing the compile time. This parameter
5797 should be removed when the delay slot code is rewritten to maintain the
5800 @item max-gcse-memory
5801 The approximate maximum amount of memory that will be allocated in
5802 order to perform the global common subexpression elimination
5803 optimization. If more memory than specified is required, the
5804 optimization will not be done.
5806 @item max-gcse-passes
5807 The maximum number of passes of GCSE to run. The default is 1.
5809 @item max-pending-list-length
5810 The maximum number of pending dependencies scheduling will allow
5811 before flushing the current state and starting over. Large functions
5812 with few branches or calls can create excessively large lists which
5813 needlessly consume memory and resources.
5815 @item max-inline-insns-single
5816 Several parameters control the tree inliner used in gcc.
5817 This number sets the maximum number of instructions (counted in GCC's
5818 internal representation) in a single function that the tree inliner
5819 will consider for inlining. This only affects functions declared
5820 inline and methods implemented in a class declaration (C++).
5821 The default value is 450.
5823 @item max-inline-insns-auto
5824 When you use @option{-finline-functions} (included in @option{-O3}),
5825 a lot of functions that would otherwise not be considered for inlining
5826 by the compiler will be investigated. To those functions, a different
5827 (more restrictive) limit compared to functions declared inline can
5829 The default value is 90.
5831 @item large-function-insns
5832 The limit specifying really large functions. For functions larger than this
5833 limit after inlining inlining is constrained by
5834 @option{--param large-function-growth}. This parameter is useful primarily
5835 to avoid extreme compilation time caused by non-linear algorithms used by the
5837 This parameter is ignored when @option{-funit-at-a-time} is not used.
5838 The default value is 2700.
5840 @item large-function-growth
5841 Specifies maximal growth of large function caused by inlining in percents.
5842 This parameter is ignored when @option{-funit-at-a-time} is not used.
5843 The default value is 100 which limits large function growth to 2.0 times
5846 @item large-unit-insns
5847 The limit specifying large translation unit. Growth caused by inlining of
5848 units larger than this limit is limited by @option{--param inline-unit-growth}.
5849 For small units this might be too tight (consider unit consisting of function A
5850 that is inline and B that just calls A three time. If B is small relative to
5851 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5852 large units consisting of small inlininable functions however the overall unit
5853 growth limit is needed to avoid exponential explosion of code size. Thus for
5854 smaller units, the size is increased to @option{--param large-unit-insns}
5855 before aplying @option{--param inline-unit-growth}. The default is 10000
5857 @item inline-unit-growth
5858 Specifies maximal overall growth of the compilation unit caused by inlining.
5859 This parameter is ignored when @option{-funit-at-a-time} is not used.
5860 The default value is 50 which limits unit growth to 1.5 times the original
5863 @item max-inline-insns-recursive
5864 @itemx max-inline-insns-recursive-auto
5865 Specifies maximum number of instructions out-of-line copy of self recursive inline
5866 function can grow into by performing recursive inlining.
5868 For functions declared inline @option{--param max-inline-insns-recursive} is
5869 taken into acount. For function not declared inline, recursive inlining
5870 happens only when @option{-finline-functions} (included in @option{-O3}) is
5871 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5872 default value is 450.
5874 @item max-inline-recursive-depth
5875 @itemx max-inline-recursive-depth-auto
5876 Specifies maximum recursion depth used by the recursive inlining.
5878 For functions declared inline @option{--param max-inline-recursive-depth} is
5879 taken into acount. For function not declared inline, recursive inlining
5880 happens only when @option{-finline-functions} (included in @option{-O3}) is
5881 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5882 default value is 450.
5884 @item min-inline-recursive-probability
5885 Recursive inlining is profitable only for function having deep recursion
5886 in average and can hurt for function having little recursion depth by
5887 increasing the prologue size or complexity of function body to other
5890 When profile feedback is available (see @option{-fprofile-generate}) the actual
5891 recursion depth can be guessed from probability that function will recurse via
5892 given call expression. This parameter limits inlining only to call expression
5893 whose probability exceeds given threshold (in percents). The default value is
5896 @item inline-call-cost
5897 Specify cost of call instruction relative to simple arithmetics operations
5898 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5899 functions and at the same time increases size of leaf function that is believed to
5900 reduce function size by being inlined. In effect it increases amount of
5901 inlining for code having large abstraction penalty (many functions that just
5902 pass the arguments to other functions) and decrease inlining for code with low
5903 abstraction penalty. The default value is 16.
5905 @item max-unrolled-insns
5906 The maximum number of instructions that a loop should have if that loop
5907 is unrolled, and if the loop is unrolled, it determines how many times
5908 the loop code is unrolled.
5910 @item max-average-unrolled-insns
5911 The maximum number of instructions biased by probabilities of their execution
5912 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5913 it determines how many times the loop code is unrolled.
5915 @item max-unroll-times
5916 The maximum number of unrollings of a single loop.
5918 @item max-peeled-insns
5919 The maximum number of instructions that a loop should have if that loop
5920 is peeled, and if the loop is peeled, it determines how many times
5921 the loop code is peeled.
5923 @item max-peel-times
5924 The maximum number of peelings of a single loop.
5926 @item max-completely-peeled-insns
5927 The maximum number of insns of a completely peeled loop.
5929 @item max-completely-peel-times
5930 The maximum number of iterations of a loop to be suitable for complete peeling.
5932 @item max-unswitch-insns
5933 The maximum number of insns of an unswitched loop.
5935 @item max-unswitch-level
5936 The maximum number of branches unswitched in a single loop.
5939 The minimum cost of an expensive expression in the loop invariant motion.
5941 @item iv-consider-all-candidates-bound
5942 Bound on number of candidates for induction variables below that
5943 all candidates are considered for each use in induction variable
5944 optimizations. Only the most relevant candidates are considered
5945 if there are more candidates, to avoid quadratic time complexity.
5947 @item iv-max-considered-uses
5948 The induction variable optimizations give up on loops that contain more
5949 induction variable uses.
5951 @item iv-always-prune-cand-set-bound
5952 If number of candidates in the set is smaller than this value,
5953 we always try to remove unnecessary ivs from the set during its
5954 optimization when a new iv is added to the set.
5956 @item scev-max-expr-size
5957 Bound on size of expressions used in the scalar evolutions analyzer.
5958 Large expressions slow the analyzer.
5960 @item vect-max-version-checks
5961 The maximum number of runtime checks that can be performed when doing
5962 loop versioning in the vectorizer. See option ftree-vect-loop-version
5963 for more information.
5965 @item max-iterations-to-track
5967 The maximum number of iterations of a loop the brute force algorithm
5968 for analysis of # of iterations of the loop tries to evaluate.
5970 @item hot-bb-count-fraction
5971 Select fraction of the maximal count of repetitions of basic block in program
5972 given basic block needs to have to be considered hot.
5974 @item hot-bb-frequency-fraction
5975 Select fraction of the maximal frequency of executions of basic block in
5976 function given basic block needs to have to be considered hot
5978 @item max-predicted-iterations
5979 The maximum number of loop iterations we predict statically. This is useful
5980 in cases where function contain single loop with known bound and other loop
5981 with unknown. We predict the known number of iterations correctly, while
5982 the unknown number of iterations average to roughly 10. This means that the
5983 loop without bounds would appear artificially cold relative to the other one.
5985 @item tracer-dynamic-coverage
5986 @itemx tracer-dynamic-coverage-feedback
5988 This value is used to limit superblock formation once the given percentage of
5989 executed instructions is covered. This limits unnecessary code size
5992 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5993 feedback is available. The real profiles (as opposed to statically estimated
5994 ones) are much less balanced allowing the threshold to be larger value.
5996 @item tracer-max-code-growth
5997 Stop tail duplication once code growth has reached given percentage. This is
5998 rather hokey argument, as most of the duplicates will be eliminated later in
5999 cross jumping, so it may be set to much higher values than is the desired code
6002 @item tracer-min-branch-ratio
6004 Stop reverse growth when the reverse probability of best edge is less than this
6005 threshold (in percent).
6007 @item tracer-min-branch-ratio
6008 @itemx tracer-min-branch-ratio-feedback
6010 Stop forward growth if the best edge do have probability lower than this
6013 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6014 compilation for profile feedback and one for compilation without. The value
6015 for compilation with profile feedback needs to be more conservative (higher) in
6016 order to make tracer effective.
6018 @item max-cse-path-length
6020 Maximum number of basic blocks on path that cse considers. The default is 10.
6023 The maximum instructions CSE process before flushing. The default is 1000.
6025 @item global-var-threshold
6027 Counts the number of function calls (@var{n}) and the number of
6028 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6029 single artificial variable will be created to represent all the
6030 call-clobbered variables at function call sites. This artificial
6031 variable will then be made to alias every call-clobbered variable.
6032 (done as @code{int * size_t} on the host machine; beware overflow).
6034 @item max-aliased-vops
6036 Maximum number of virtual operands allowed to represent aliases
6037 before triggering the alias grouping heuristic. Alias grouping
6038 reduces compile times and memory consumption needed for aliasing at
6039 the expense of precision loss in alias information.
6041 @item ggc-min-expand
6043 GCC uses a garbage collector to manage its own memory allocation. This
6044 parameter specifies the minimum percentage by which the garbage
6045 collector's heap should be allowed to expand between collections.
6046 Tuning this may improve compilation speed; it has no effect on code
6049 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6050 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6051 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6052 GCC is not able to calculate RAM on a particular platform, the lower
6053 bound of 30% is used. Setting this parameter and
6054 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6055 every opportunity. This is extremely slow, but can be useful for
6058 @item ggc-min-heapsize
6060 Minimum size of the garbage collector's heap before it begins bothering
6061 to collect garbage. The first collection occurs after the heap expands
6062 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6063 tuning this may improve compilation speed, and has no effect on code
6066 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6067 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6068 with a lower bound of 4096 (four megabytes) and an upper bound of
6069 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6070 particular platform, the lower bound is used. Setting this parameter
6071 very large effectively disables garbage collection. Setting this
6072 parameter and @option{ggc-min-expand} to zero causes a full collection
6073 to occur at every opportunity.
6075 @item max-reload-search-insns
6076 The maximum number of instruction reload should look backward for equivalent
6077 register. Increasing values mean more aggressive optimization, making the
6078 compile time increase with probably slightly better performance. The default
6081 @item max-cselib-memory-location
6082 The maximum number of memory locations cselib should take into acount.
6083 Increasing values mean more aggressive optimization, making the compile time
6084 increase with probably slightly better performance. The default value is 500.
6086 @item max-flow-memory-location
6087 Similar as @option{max-cselib-memory-location} but for dataflow liveness.
6088 The default value is 100.
6090 @item reorder-blocks-duplicate
6091 @itemx reorder-blocks-duplicate-feedback
6093 Used by basic block reordering pass to decide whether to use unconditional
6094 branch or duplicate the code on its destination. Code is duplicated when its
6095 estimated size is smaller than this value multiplied by the estimated size of
6096 unconditional jump in the hot spots of the program.
6098 The @option{reorder-block-duplicate-feedback} is used only when profile
6099 feedback is available and may be set to higher values than
6100 @option{reorder-block-duplicate} since information about the hot spots is more
6103 @item max-sched-region-blocks
6104 The maximum number of blocks in a region to be considered for
6105 interblock scheduling. The default value is 10.
6107 @item max-sched-region-insns
6108 The maximum number of insns in a region to be considered for
6109 interblock scheduling. The default value is 100.
6111 @item min-sched-prob
6112 The minimum probability of reaching a source block for interblock
6113 speculative scheduling. The default value is 40.
6115 @item max-last-value-rtl
6117 The maximum size measured as number of RTLs that can be recorded in an expression
6118 in combiner for a pseudo register as last known value of that register. The default
6121 @item integer-share-limit
6122 Small integer constants can use a shared data structure, reducing the
6123 compiler's memory usage and increasing its speed. This sets the maximum
6124 value of a shared integer constant's. The default value is 256.
6126 @item min-virtual-mappings
6127 Specifies the minimum number of virtual mappings in the incremental
6128 SSA updater that should be registered to trigger the virtual mappings
6129 heuristic defined by virtual-mappings-ratio. The default value is
6132 @item virtual-mappings-ratio
6133 If the number of virtual mappings is virtual-mappings-ratio bigger
6134 than the number of virtual symbols to be updated, then the incremental
6135 SSA updater switches to a full update for those symbols. The default
6138 @item ssp-buffer-size
6139 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6140 protection when @option{-fstack-protection} is used.
6142 @item max-jump-thread-duplication-stmts
6143 Maximum number of statements allowed in a block that needs to be
6144 duplicated when threading jumps.
6148 @node Preprocessor Options
6149 @section Options Controlling the Preprocessor
6150 @cindex preprocessor options
6151 @cindex options, preprocessor
6153 These options control the C preprocessor, which is run on each C source
6154 file before actual compilation.
6156 If you use the @option{-E} option, nothing is done except preprocessing.
6157 Some of these options make sense only together with @option{-E} because
6158 they cause the preprocessor output to be unsuitable for actual
6163 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6164 and pass @var{option} directly through to the preprocessor. If
6165 @var{option} contains commas, it is split into multiple options at the
6166 commas. However, many options are modified, translated or interpreted
6167 by the compiler driver before being passed to the preprocessor, and
6168 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6169 interface is undocumented and subject to change, so whenever possible
6170 you should avoid using @option{-Wp} and let the driver handle the
6173 @item -Xpreprocessor @var{option}
6174 @opindex preprocessor
6175 Pass @var{option} as an option to the preprocessor. You can use this to
6176 supply system-specific preprocessor options which GCC does not know how to
6179 If you want to pass an option that takes an argument, you must use
6180 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6183 @include cppopts.texi
6185 @node Assembler Options
6186 @section Passing Options to the Assembler
6188 @c prevent bad page break with this line
6189 You can pass options to the assembler.
6192 @item -Wa,@var{option}
6194 Pass @var{option} as an option to the assembler. If @var{option}
6195 contains commas, it is split into multiple options at the commas.
6197 @item -Xassembler @var{option}
6199 Pass @var{option} as an option to the assembler. You can use this to
6200 supply system-specific assembler options which GCC does not know how to
6203 If you want to pass an option that takes an argument, you must use
6204 @option{-Xassembler} twice, once for the option and once for the argument.
6209 @section Options for Linking
6210 @cindex link options
6211 @cindex options, linking
6213 These options come into play when the compiler links object files into
6214 an executable output file. They are meaningless if the compiler is
6215 not doing a link step.
6219 @item @var{object-file-name}
6220 A file name that does not end in a special recognized suffix is
6221 considered to name an object file or library. (Object files are
6222 distinguished from libraries by the linker according to the file
6223 contents.) If linking is done, these object files are used as input
6232 If any of these options is used, then the linker is not run, and
6233 object file names should not be used as arguments. @xref{Overall
6237 @item -l@var{library}
6238 @itemx -l @var{library}
6240 Search the library named @var{library} when linking. (The second
6241 alternative with the library as a separate argument is only for
6242 POSIX compliance and is not recommended.)
6244 It makes a difference where in the command you write this option; the
6245 linker searches and processes libraries and object files in the order they
6246 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6247 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6248 to functions in @samp{z}, those functions may not be loaded.
6250 The linker searches a standard list of directories for the library,
6251 which is actually a file named @file{lib@var{library}.a}. The linker
6252 then uses this file as if it had been specified precisely by name.
6254 The directories searched include several standard system directories
6255 plus any that you specify with @option{-L}.
6257 Normally the files found this way are library files---archive files
6258 whose members are object files. The linker handles an archive file by
6259 scanning through it for members which define symbols that have so far
6260 been referenced but not defined. But if the file that is found is an
6261 ordinary object file, it is linked in the usual fashion. The only
6262 difference between using an @option{-l} option and specifying a file name
6263 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6264 and searches several directories.
6268 You need this special case of the @option{-l} option in order to
6269 link an Objective-C or Objective-C++ program.
6272 @opindex nostartfiles
6273 Do not use the standard system startup files when linking.
6274 The standard system libraries are used normally, unless @option{-nostdlib}
6275 or @option{-nodefaultlibs} is used.
6277 @item -nodefaultlibs
6278 @opindex nodefaultlibs
6279 Do not use the standard system libraries when linking.
6280 Only the libraries you specify will be passed to the linker.
6281 The standard startup files are used normally, unless @option{-nostartfiles}
6282 is used. The compiler may generate calls to @code{memcmp},
6283 @code{memset}, @code{memcpy} and @code{memmove}.
6284 These entries are usually resolved by entries in
6285 libc. These entry points should be supplied through some other
6286 mechanism when this option is specified.
6290 Do not use the standard system startup files or libraries when linking.
6291 No startup files and only the libraries you specify will be passed to
6292 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6293 @code{memcpy} and @code{memmove}.
6294 These entries are usually resolved by entries in
6295 libc. These entry points should be supplied through some other
6296 mechanism when this option is specified.
6298 @cindex @option{-lgcc}, use with @option{-nostdlib}
6299 @cindex @option{-nostdlib} and unresolved references
6300 @cindex unresolved references and @option{-nostdlib}
6301 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6302 @cindex @option{-nodefaultlibs} and unresolved references
6303 @cindex unresolved references and @option{-nodefaultlibs}
6304 One of the standard libraries bypassed by @option{-nostdlib} and
6305 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6306 that GCC uses to overcome shortcomings of particular machines, or special
6307 needs for some languages.
6308 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6309 Collection (GCC) Internals},
6310 for more discussion of @file{libgcc.a}.)
6311 In most cases, you need @file{libgcc.a} even when you want to avoid
6312 other standard libraries. In other words, when you specify @option{-nostdlib}
6313 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6314 This ensures that you have no unresolved references to internal GCC
6315 library subroutines. (For example, @samp{__main}, used to ensure C++
6316 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6317 GNU Compiler Collection (GCC) Internals}.)
6321 Produce a position independent executable on targets which support it.
6322 For predictable results, you must also specify the same set of options
6323 that were used to generate code (@option{-fpie}, @option{-fPIE},
6324 or model suboptions) when you specify this option.
6328 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6329 that support it. This instructs the linker to add all symbols, not
6330 only used ones, to the dynamic symbol table. This option is needed
6331 for some uses of @code{dlopen} or to allow obtaining backtraces
6332 from within a program.
6336 Remove all symbol table and relocation information from the executable.
6340 On systems that support dynamic linking, this prevents linking with the shared
6341 libraries. On other systems, this option has no effect.
6345 Produce a shared object which can then be linked with other objects to
6346 form an executable. Not all systems support this option. For predictable
6347 results, you must also specify the same set of options that were used to
6348 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6349 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6350 needs to build supplementary stub code for constructors to work. On
6351 multi-libbed systems, @samp{gcc -shared} must select the correct support
6352 libraries to link against. Failing to supply the correct flags may lead
6353 to subtle defects. Supplying them in cases where they are not necessary
6356 @item -shared-libgcc
6357 @itemx -static-libgcc
6358 @opindex shared-libgcc
6359 @opindex static-libgcc
6360 On systems that provide @file{libgcc} as a shared library, these options
6361 force the use of either the shared or static version respectively.
6362 If no shared version of @file{libgcc} was built when the compiler was
6363 configured, these options have no effect.
6365 There are several situations in which an application should use the
6366 shared @file{libgcc} instead of the static version. The most common
6367 of these is when the application wishes to throw and catch exceptions
6368 across different shared libraries. In that case, each of the libraries
6369 as well as the application itself should use the shared @file{libgcc}.
6371 Therefore, the G++ and GCJ drivers automatically add
6372 @option{-shared-libgcc} whenever you build a shared library or a main
6373 executable, because C++ and Java programs typically use exceptions, so
6374 this is the right thing to do.
6376 If, instead, you use the GCC driver to create shared libraries, you may
6377 find that they will not always be linked with the shared @file{libgcc}.
6378 If GCC finds, at its configuration time, that you have a non-GNU linker
6379 or a GNU linker that does not support option @option{--eh-frame-hdr},
6380 it will link the shared version of @file{libgcc} into shared libraries
6381 by default. Otherwise, it will take advantage of the linker and optimize
6382 away the linking with the shared version of @file{libgcc}, linking with
6383 the static version of libgcc by default. This allows exceptions to
6384 propagate through such shared libraries, without incurring relocation
6385 costs at library load time.
6387 However, if a library or main executable is supposed to throw or catch
6388 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6389 for the languages used in the program, or using the option
6390 @option{-shared-libgcc}, such that it is linked with the shared
6395 Bind references to global symbols when building a shared object. Warn
6396 about any unresolved references (unless overridden by the link editor
6397 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6400 @item -Xlinker @var{option}
6402 Pass @var{option} as an option to the linker. You can use this to
6403 supply system-specific linker options which GCC does not know how to
6406 If you want to pass an option that takes an argument, you must use
6407 @option{-Xlinker} twice, once for the option and once for the argument.
6408 For example, to pass @option{-assert definitions}, you must write
6409 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6410 @option{-Xlinker "-assert definitions"}, because this passes the entire
6411 string as a single argument, which is not what the linker expects.
6413 @item -Wl,@var{option}
6415 Pass @var{option} as an option to the linker. If @var{option} contains
6416 commas, it is split into multiple options at the commas.
6418 @item -u @var{symbol}
6420 Pretend the symbol @var{symbol} is undefined, to force linking of
6421 library modules to define it. You can use @option{-u} multiple times with
6422 different symbols to force loading of additional library modules.
6425 @node Directory Options
6426 @section Options for Directory Search
6427 @cindex directory options
6428 @cindex options, directory search
6431 These options specify directories to search for header files, for
6432 libraries and for parts of the compiler:
6437 Add the directory @var{dir} to the head of the list of directories to be
6438 searched for header files. This can be used to override a system header
6439 file, substituting your own version, since these directories are
6440 searched before the system header file directories. However, you should
6441 not use this option to add directories that contain vendor-supplied
6442 system header files (use @option{-isystem} for that). If you use more than
6443 one @option{-I} option, the directories are scanned in left-to-right
6444 order; the standard system directories come after.
6446 If a standard system include directory, or a directory specified with
6447 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6448 option will be ignored. The directory will still be searched but as a
6449 system directory at its normal position in the system include chain.
6450 This is to ensure that GCC's procedure to fix buggy system headers and
6451 the ordering for the include_next directive are not inadvertently changed.
6452 If you really need to change the search order for system directories,
6453 use the @option{-nostdinc} and/or @option{-isystem} options.
6455 @item -iquote@var{dir}
6457 Add the directory @var{dir} to the head of the list of directories to
6458 be searched for header files only for the case of @samp{#include
6459 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6460 otherwise just like @option{-I}.
6464 Add directory @var{dir} to the list of directories to be searched
6467 @item -B@var{prefix}
6469 This option specifies where to find the executables, libraries,
6470 include files, and data files of the compiler itself.
6472 The compiler driver program runs one or more of the subprograms
6473 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6474 @var{prefix} as a prefix for each program it tries to run, both with and
6475 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6477 For each subprogram to be run, the compiler driver first tries the
6478 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6479 was not specified, the driver tries two standard prefixes, which are
6480 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6481 those results in a file name that is found, the unmodified program
6482 name is searched for using the directories specified in your
6483 @env{PATH} environment variable.
6485 The compiler will check to see if the path provided by the @option{-B}
6486 refers to a directory, and if necessary it will add a directory
6487 separator character at the end of the path.
6489 @option{-B} prefixes that effectively specify directory names also apply
6490 to libraries in the linker, because the compiler translates these
6491 options into @option{-L} options for the linker. They also apply to
6492 includes files in the preprocessor, because the compiler translates these
6493 options into @option{-isystem} options for the preprocessor. In this case,
6494 the compiler appends @samp{include} to the prefix.
6496 The run-time support file @file{libgcc.a} can also be searched for using
6497 the @option{-B} prefix, if needed. If it is not found there, the two
6498 standard prefixes above are tried, and that is all. The file is left
6499 out of the link if it is not found by those means.
6501 Another way to specify a prefix much like the @option{-B} prefix is to use
6502 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6505 As a special kludge, if the path provided by @option{-B} is
6506 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6507 9, then it will be replaced by @file{[dir/]include}. This is to help
6508 with boot-strapping the compiler.
6510 @item -specs=@var{file}
6512 Process @var{file} after the compiler reads in the standard @file{specs}
6513 file, in order to override the defaults that the @file{gcc} driver
6514 program uses when determining what switches to pass to @file{cc1},
6515 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6516 @option{-specs=@var{file}} can be specified on the command line, and they
6517 are processed in order, from left to right.
6519 @item --sysroot=@var{dir}
6521 Use @var{dir} as the logical root directory for headers and libraries.
6522 For example, if the compiler would normally search for headers in
6523 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6524 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6526 If you use both this option and the @option{-isysroot} option, then
6527 the @option{--sysroot} option will apply to libraries, but the
6528 @option{-isysroot} option will apply to header files.
6530 The GNU linker (beginning with version 2.16) has the necessary support
6531 for this option. If your linker does not support this option, the
6532 header file aspect of @option{--sysroot} will still work, but the
6533 library aspect will not.
6537 This option has been deprecated. Please use @option{-iquote} instead for
6538 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6539 Any directories you specify with @option{-I} options before the @option{-I-}
6540 option are searched only for the case of @samp{#include "@var{file}"};
6541 they are not searched for @samp{#include <@var{file}>}.
6543 If additional directories are specified with @option{-I} options after
6544 the @option{-I-}, these directories are searched for all @samp{#include}
6545 directives. (Ordinarily @emph{all} @option{-I} directories are used
6548 In addition, the @option{-I-} option inhibits the use of the current
6549 directory (where the current input file came from) as the first search
6550 directory for @samp{#include "@var{file}"}. There is no way to
6551 override this effect of @option{-I-}. With @option{-I.} you can specify
6552 searching the directory which was current when the compiler was
6553 invoked. That is not exactly the same as what the preprocessor does
6554 by default, but it is often satisfactory.
6556 @option{-I-} does not inhibit the use of the standard system directories
6557 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6564 @section Specifying subprocesses and the switches to pass to them
6567 @command{gcc} is a driver program. It performs its job by invoking a
6568 sequence of other programs to do the work of compiling, assembling and
6569 linking. GCC interprets its command-line parameters and uses these to
6570 deduce which programs it should invoke, and which command-line options
6571 it ought to place on their command lines. This behavior is controlled
6572 by @dfn{spec strings}. In most cases there is one spec string for each
6573 program that GCC can invoke, but a few programs have multiple spec
6574 strings to control their behavior. The spec strings built into GCC can
6575 be overridden by using the @option{-specs=} command-line switch to specify
6578 @dfn{Spec files} are plaintext files that are used to construct spec
6579 strings. They consist of a sequence of directives separated by blank
6580 lines. The type of directive is determined by the first non-whitespace
6581 character on the line and it can be one of the following:
6584 @item %@var{command}
6585 Issues a @var{command} to the spec file processor. The commands that can
6589 @item %include <@var{file}>
6591 Search for @var{file} and insert its text at the current point in the
6594 @item %include_noerr <@var{file}>
6595 @cindex %include_noerr
6596 Just like @samp{%include}, but do not generate an error message if the include
6597 file cannot be found.
6599 @item %rename @var{old_name} @var{new_name}
6601 Rename the spec string @var{old_name} to @var{new_name}.
6605 @item *[@var{spec_name}]:
6606 This tells the compiler to create, override or delete the named spec
6607 string. All lines after this directive up to the next directive or
6608 blank line are considered to be the text for the spec string. If this
6609 results in an empty string then the spec will be deleted. (Or, if the
6610 spec did not exist, then nothing will happened.) Otherwise, if the spec
6611 does not currently exist a new spec will be created. If the spec does
6612 exist then its contents will be overridden by the text of this
6613 directive, unless the first character of that text is the @samp{+}
6614 character, in which case the text will be appended to the spec.
6616 @item [@var{suffix}]:
6617 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6618 and up to the next directive or blank line are considered to make up the
6619 spec string for the indicated suffix. When the compiler encounters an
6620 input file with the named suffix, it will processes the spec string in
6621 order to work out how to compile that file. For example:
6628 This says that any input file whose name ends in @samp{.ZZ} should be
6629 passed to the program @samp{z-compile}, which should be invoked with the
6630 command-line switch @option{-input} and with the result of performing the
6631 @samp{%i} substitution. (See below.)
6633 As an alternative to providing a spec string, the text that follows a
6634 suffix directive can be one of the following:
6637 @item @@@var{language}
6638 This says that the suffix is an alias for a known @var{language}. This is
6639 similar to using the @option{-x} command-line switch to GCC to specify a
6640 language explicitly. For example:
6647 Says that .ZZ files are, in fact, C++ source files.
6650 This causes an error messages saying:
6653 @var{name} compiler not installed on this system.
6657 GCC already has an extensive list of suffixes built into it.
6658 This directive will add an entry to the end of the list of suffixes, but
6659 since the list is searched from the end backwards, it is effectively
6660 possible to override earlier entries using this technique.
6664 GCC has the following spec strings built into it. Spec files can
6665 override these strings or create their own. Note that individual
6666 targets can also add their own spec strings to this list.
6669 asm Options to pass to the assembler
6670 asm_final Options to pass to the assembler post-processor
6671 cpp Options to pass to the C preprocessor
6672 cc1 Options to pass to the C compiler
6673 cc1plus Options to pass to the C++ compiler
6674 endfile Object files to include at the end of the link
6675 link Options to pass to the linker
6676 lib Libraries to include on the command line to the linker
6677 libgcc Decides which GCC support library to pass to the linker
6678 linker Sets the name of the linker
6679 predefines Defines to be passed to the C preprocessor
6680 signed_char Defines to pass to CPP to say whether @code{char} is signed
6682 startfile Object files to include at the start of the link
6685 Here is a small example of a spec file:
6691 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6694 This example renames the spec called @samp{lib} to @samp{old_lib} and
6695 then overrides the previous definition of @samp{lib} with a new one.
6696 The new definition adds in some extra command-line options before
6697 including the text of the old definition.
6699 @dfn{Spec strings} are a list of command-line options to be passed to their
6700 corresponding program. In addition, the spec strings can contain
6701 @samp{%}-prefixed sequences to substitute variable text or to
6702 conditionally insert text into the command line. Using these constructs
6703 it is possible to generate quite complex command lines.
6705 Here is a table of all defined @samp{%}-sequences for spec
6706 strings. Note that spaces are not generated automatically around the
6707 results of expanding these sequences. Therefore you can concatenate them
6708 together or combine them with constant text in a single argument.
6712 Substitute one @samp{%} into the program name or argument.
6715 Substitute the name of the input file being processed.
6718 Substitute the basename of the input file being processed.
6719 This is the substring up to (and not including) the last period
6720 and not including the directory.
6723 This is the same as @samp{%b}, but include the file suffix (text after
6727 Marks the argument containing or following the @samp{%d} as a
6728 temporary file name, so that that file will be deleted if GCC exits
6729 successfully. Unlike @samp{%g}, this contributes no text to the
6732 @item %g@var{suffix}
6733 Substitute a file name that has suffix @var{suffix} and is chosen
6734 once per compilation, and mark the argument in the same way as
6735 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6736 name is now chosen in a way that is hard to predict even when previously
6737 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6738 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6739 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6740 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6741 was simply substituted with a file name chosen once per compilation,
6742 without regard to any appended suffix (which was therefore treated
6743 just like ordinary text), making such attacks more likely to succeed.
6745 @item %u@var{suffix}
6746 Like @samp{%g}, but generates a new temporary file name even if
6747 @samp{%u@var{suffix}} was already seen.
6749 @item %U@var{suffix}
6750 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6751 new one if there is no such last file name. In the absence of any
6752 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6753 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6754 would involve the generation of two distinct file names, one
6755 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6756 simply substituted with a file name chosen for the previous @samp{%u},
6757 without regard to any appended suffix.
6759 @item %j@var{suffix}
6760 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6761 writable, and if save-temps is off; otherwise, substitute the name
6762 of a temporary file, just like @samp{%u}. This temporary file is not
6763 meant for communication between processes, but rather as a junk
6766 @item %|@var{suffix}
6767 @itemx %m@var{suffix}
6768 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6769 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6770 all. These are the two most common ways to instruct a program that it
6771 should read from standard input or write to standard output. If you
6772 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6773 construct: see for example @file{f/lang-specs.h}.
6775 @item %.@var{SUFFIX}
6776 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6777 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6778 terminated by the next space or %.
6781 Marks the argument containing or following the @samp{%w} as the
6782 designated output file of this compilation. This puts the argument
6783 into the sequence of arguments that @samp{%o} will substitute later.
6786 Substitutes the names of all the output files, with spaces
6787 automatically placed around them. You should write spaces
6788 around the @samp{%o} as well or the results are undefined.
6789 @samp{%o} is for use in the specs for running the linker.
6790 Input files whose names have no recognized suffix are not compiled
6791 at all, but they are included among the output files, so they will
6795 Substitutes the suffix for object files. Note that this is
6796 handled specially when it immediately follows @samp{%g, %u, or %U},
6797 because of the need for those to form complete file names. The
6798 handling is such that @samp{%O} is treated exactly as if it had already
6799 been substituted, except that @samp{%g, %u, and %U} do not currently
6800 support additional @var{suffix} characters following @samp{%O} as they would
6801 following, for example, @samp{.o}.
6804 Substitutes the standard macro predefinitions for the
6805 current target machine. Use this when running @code{cpp}.
6808 Like @samp{%p}, but puts @samp{__} before and after the name of each
6809 predefined macro, except for macros that start with @samp{__} or with
6810 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6814 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6815 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6816 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6820 Current argument is the name of a library or startup file of some sort.
6821 Search for that file in a standard list of directories and substitute
6822 the full name found.
6825 Print @var{str} as an error message. @var{str} is terminated by a newline.
6826 Use this when inconsistent options are detected.
6829 Substitute the contents of spec string @var{name} at this point.
6832 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6834 @item %x@{@var{option}@}
6835 Accumulate an option for @samp{%X}.
6838 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6842 Output the accumulated assembler options specified by @option{-Wa}.
6845 Output the accumulated preprocessor options specified by @option{-Wp}.
6848 Process the @code{asm} spec. This is used to compute the
6849 switches to be passed to the assembler.
6852 Process the @code{asm_final} spec. This is a spec string for
6853 passing switches to an assembler post-processor, if such a program is
6857 Process the @code{link} spec. This is the spec for computing the
6858 command line passed to the linker. Typically it will make use of the
6859 @samp{%L %G %S %D and %E} sequences.
6862 Dump out a @option{-L} option for each directory that GCC believes might
6863 contain startup files. If the target supports multilibs then the
6864 current multilib directory will be prepended to each of these paths.
6867 Process the @code{lib} spec. This is a spec string for deciding which
6868 libraries should be included on the command line to the linker.
6871 Process the @code{libgcc} spec. This is a spec string for deciding
6872 which GCC support library should be included on the command line to the linker.
6875 Process the @code{startfile} spec. This is a spec for deciding which
6876 object files should be the first ones passed to the linker. Typically
6877 this might be a file named @file{crt0.o}.
6880 Process the @code{endfile} spec. This is a spec string that specifies
6881 the last object files that will be passed to the linker.
6884 Process the @code{cpp} spec. This is used to construct the arguments
6885 to be passed to the C preprocessor.
6888 Process the @code{cc1} spec. This is used to construct the options to be
6889 passed to the actual C compiler (@samp{cc1}).
6892 Process the @code{cc1plus} spec. This is used to construct the options to be
6893 passed to the actual C++ compiler (@samp{cc1plus}).
6896 Substitute the variable part of a matched option. See below.
6897 Note that each comma in the substituted string is replaced by
6901 Remove all occurrences of @code{-S} from the command line. Note---this
6902 command is position dependent. @samp{%} commands in the spec string
6903 before this one will see @code{-S}, @samp{%} commands in the spec string
6904 after this one will not.
6906 @item %:@var{function}(@var{args})
6907 Call the named function @var{function}, passing it @var{args}.
6908 @var{args} is first processed as a nested spec string, then split
6909 into an argument vector in the usual fashion. The function returns
6910 a string which is processed as if it had appeared literally as part
6911 of the current spec.
6913 The following built-in spec functions are provided:
6916 @item @code{if-exists}
6917 The @code{if-exists} spec function takes one argument, an absolute
6918 pathname to a file. If the file exists, @code{if-exists} returns the
6919 pathname. Here is a small example of its usage:
6923 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6926 @item @code{if-exists-else}
6927 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6928 spec function, except that it takes two arguments. The first argument is
6929 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6930 returns the pathname. If it does not exist, it returns the second argument.
6931 This way, @code{if-exists-else} can be used to select one file or another,
6932 based on the existence of the first. Here is a small example of its usage:
6936 crt0%O%s %:if-exists(crti%O%s) \
6937 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6940 @item @code{replace-outfile}
6941 The @code{replace-outfile} spec function takes two arguments. It looks for the
6942 first argument in the outfiles array and replaces it with the second argument. Here
6943 is a small example of its usage:
6946 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6952 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6953 If that switch was not specified, this substitutes nothing. Note that
6954 the leading dash is omitted when specifying this option, and it is
6955 automatically inserted if the substitution is performed. Thus the spec
6956 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6957 and would output the command line option @option{-foo}.
6959 @item %W@{@code{S}@}
6960 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6963 @item %@{@code{S}*@}
6964 Substitutes all the switches specified to GCC whose names start
6965 with @code{-S}, but which also take an argument. This is used for
6966 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6967 GCC considers @option{-o foo} as being
6968 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6969 text, including the space. Thus two arguments would be generated.
6971 @item %@{@code{S}*&@code{T}*@}
6972 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6973 (the order of @code{S} and @code{T} in the spec is not significant).
6974 There can be any number of ampersand-separated variables; for each the
6975 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6977 @item %@{@code{S}:@code{X}@}
6978 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6980 @item %@{!@code{S}:@code{X}@}
6981 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6983 @item %@{@code{S}*:@code{X}@}
6984 Substitutes @code{X} if one or more switches whose names start with
6985 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6986 once, no matter how many such switches appeared. However, if @code{%*}
6987 appears somewhere in @code{X}, then @code{X} will be substituted once
6988 for each matching switch, with the @code{%*} replaced by the part of
6989 that switch that matched the @code{*}.
6991 @item %@{.@code{S}:@code{X}@}
6992 Substitutes @code{X}, if processing a file with suffix @code{S}.
6994 @item %@{!.@code{S}:@code{X}@}
6995 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6997 @item %@{@code{S}|@code{P}:@code{X}@}
6998 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6999 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7000 although they have a stronger binding than the @samp{|}. If @code{%*}
7001 appears in @code{X}, all of the alternatives must be starred, and only
7002 the first matching alternative is substituted.
7004 For example, a spec string like this:
7007 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7010 will output the following command-line options from the following input
7011 command-line options:
7016 -d fred.c -foo -baz -boggle
7017 -d jim.d -bar -baz -boggle
7020 @item %@{S:X; T:Y; :D@}
7022 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7023 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7024 be as many clauses as you need. This may be combined with @code{.},
7025 @code{!}, @code{|}, and @code{*} as needed.
7030 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7031 construct may contain other nested @samp{%} constructs or spaces, or
7032 even newlines. They are processed as usual, as described above.
7033 Trailing white space in @code{X} is ignored. White space may also
7034 appear anywhere on the left side of the colon in these constructs,
7035 except between @code{.} or @code{*} and the corresponding word.
7037 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7038 handled specifically in these constructs. If another value of
7039 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7040 @option{-W} switch is found later in the command line, the earlier
7041 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7042 just one letter, which passes all matching options.
7044 The character @samp{|} at the beginning of the predicate text is used to
7045 indicate that a command should be piped to the following command, but
7046 only if @option{-pipe} is specified.
7048 It is built into GCC which switches take arguments and which do not.
7049 (You might think it would be useful to generalize this to allow each
7050 compiler's spec to say which switches take arguments. But this cannot
7051 be done in a consistent fashion. GCC cannot even decide which input
7052 files have been specified without knowing which switches take arguments,
7053 and it must know which input files to compile in order to tell which
7056 GCC also knows implicitly that arguments starting in @option{-l} are to be
7057 treated as compiler output files, and passed to the linker in their
7058 proper position among the other output files.
7060 @c man begin OPTIONS
7062 @node Target Options
7063 @section Specifying Target Machine and Compiler Version
7064 @cindex target options
7065 @cindex cross compiling
7066 @cindex specifying machine version
7067 @cindex specifying compiler version and target machine
7068 @cindex compiler version, specifying
7069 @cindex target machine, specifying
7071 The usual way to run GCC is to run the executable called @file{gcc}, or
7072 @file{<machine>-gcc} when cross-compiling, or
7073 @file{<machine>-gcc-<version>} to run a version other than the one that
7074 was installed last. Sometimes this is inconvenient, so GCC provides
7075 options that will switch to another cross-compiler or version.
7078 @item -b @var{machine}
7080 The argument @var{machine} specifies the target machine for compilation.
7082 The value to use for @var{machine} is the same as was specified as the
7083 machine type when configuring GCC as a cross-compiler. For
7084 example, if a cross-compiler was configured with @samp{configure
7085 arm-elf}, meaning to compile for an arm processor with elf binaries,
7086 then you would specify @option{-b arm-elf} to run that cross compiler.
7087 Because there are other options beginning with @option{-b}, the
7088 configuration must contain a hyphen.
7090 @item -V @var{version}
7092 The argument @var{version} specifies which version of GCC to run.
7093 This is useful when multiple versions are installed. For example,
7094 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7097 The @option{-V} and @option{-b} options work by running the
7098 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7099 use them if you can just run that directly.
7101 @node Submodel Options
7102 @section Hardware Models and Configurations
7103 @cindex submodel options
7104 @cindex specifying hardware config
7105 @cindex hardware models and configurations, specifying
7106 @cindex machine dependent options
7108 Earlier we discussed the standard option @option{-b} which chooses among
7109 different installed compilers for completely different target
7110 machines, such as VAX vs.@: 68000 vs.@: 80386.
7112 In addition, each of these target machine types can have its own
7113 special options, starting with @samp{-m}, to choose among various
7114 hardware models or configurations---for example, 68010 vs 68020,
7115 floating coprocessor or none. A single installed version of the
7116 compiler can compile for any model or configuration, according to the
7119 Some configurations of the compiler also support additional special
7120 options, usually for compatibility with other compilers on the same
7123 @c This list is ordered alphanumerically by subsection name.
7124 @c It should be the same order and spelling as these options are listed
7125 @c in Machine Dependent Options
7131 * Blackfin Options::
7135 * DEC Alpha Options::
7136 * DEC Alpha/VMS Options::
7140 * i386 and x86-64 Options::
7153 * RS/6000 and PowerPC Options::
7154 * S/390 and zSeries Options::
7157 * System V Options::
7158 * TMS320C3x/C4x Options::
7162 * Xstormy16 Options::
7168 @subsection ARC Options
7171 These options are defined for ARC implementations:
7176 Compile code for little endian mode. This is the default.
7180 Compile code for big endian mode.
7183 @opindex mmangle-cpu
7184 Prepend the name of the cpu to all public symbol names.
7185 In multiple-processor systems, there are many ARC variants with different
7186 instruction and register set characteristics. This flag prevents code
7187 compiled for one cpu to be linked with code compiled for another.
7188 No facility exists for handling variants that are ``almost identical''.
7189 This is an all or nothing option.
7191 @item -mcpu=@var{cpu}
7193 Compile code for ARC variant @var{cpu}.
7194 Which variants are supported depend on the configuration.
7195 All variants support @option{-mcpu=base}, this is the default.
7197 @item -mtext=@var{text-section}
7198 @itemx -mdata=@var{data-section}
7199 @itemx -mrodata=@var{readonly-data-section}
7203 Put functions, data, and readonly data in @var{text-section},
7204 @var{data-section}, and @var{readonly-data-section} respectively
7205 by default. This can be overridden with the @code{section} attribute.
7206 @xref{Variable Attributes}.
7211 @subsection ARM Options
7214 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7218 @item -mabi=@var{name}
7220 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7221 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7224 @opindex mapcs-frame
7225 Generate a stack frame that is compliant with the ARM Procedure Call
7226 Standard for all functions, even if this is not strictly necessary for
7227 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7228 with this option will cause the stack frames not to be generated for
7229 leaf functions. The default is @option{-mno-apcs-frame}.
7233 This is a synonym for @option{-mapcs-frame}.
7236 @c not currently implemented
7237 @item -mapcs-stack-check
7238 @opindex mapcs-stack-check
7239 Generate code to check the amount of stack space available upon entry to
7240 every function (that actually uses some stack space). If there is
7241 insufficient space available then either the function
7242 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7243 called, depending upon the amount of stack space required. The run time
7244 system is required to provide these functions. The default is
7245 @option{-mno-apcs-stack-check}, since this produces smaller code.
7247 @c not currently implemented
7249 @opindex mapcs-float
7250 Pass floating point arguments using the float point registers. This is
7251 one of the variants of the APCS@. This option is recommended if the
7252 target hardware has a floating point unit or if a lot of floating point
7253 arithmetic is going to be performed by the code. The default is
7254 @option{-mno-apcs-float}, since integer only code is slightly increased in
7255 size if @option{-mapcs-float} is used.
7257 @c not currently implemented
7258 @item -mapcs-reentrant
7259 @opindex mapcs-reentrant
7260 Generate reentrant, position independent code. The default is
7261 @option{-mno-apcs-reentrant}.
7264 @item -mthumb-interwork
7265 @opindex mthumb-interwork
7266 Generate code which supports calling between the ARM and Thumb
7267 instruction sets. Without this option the two instruction sets cannot
7268 be reliably used inside one program. The default is
7269 @option{-mno-thumb-interwork}, since slightly larger code is generated
7270 when @option{-mthumb-interwork} is specified.
7272 @item -mno-sched-prolog
7273 @opindex mno-sched-prolog
7274 Prevent the reordering of instructions in the function prolog, or the
7275 merging of those instruction with the instructions in the function's
7276 body. This means that all functions will start with a recognizable set
7277 of instructions (or in fact one of a choice from a small set of
7278 different function prologues), and this information can be used to
7279 locate the start if functions inside an executable piece of code. The
7280 default is @option{-msched-prolog}.
7283 @opindex mhard-float
7284 Generate output containing floating point instructions. This is the
7288 @opindex msoft-float
7289 Generate output containing library calls for floating point.
7290 @strong{Warning:} the requisite libraries are not available for all ARM
7291 targets. Normally the facilities of the machine's usual C compiler are
7292 used, but this cannot be done directly in cross-compilation. You must make
7293 your own arrangements to provide suitable library functions for
7296 @option{-msoft-float} changes the calling convention in the output file;
7297 therefore, it is only useful if you compile @emph{all} of a program with
7298 this option. In particular, you need to compile @file{libgcc.a}, the
7299 library that comes with GCC, with @option{-msoft-float} in order for
7302 @item -mfloat-abi=@var{name}
7304 Specifies which ABI to use for floating point values. Permissible values
7305 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7307 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7308 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7309 of floating point instructions, but still uses the soft-float calling
7312 @item -mlittle-endian
7313 @opindex mlittle-endian
7314 Generate code for a processor running in little-endian mode. This is
7315 the default for all standard configurations.
7318 @opindex mbig-endian
7319 Generate code for a processor running in big-endian mode; the default is
7320 to compile code for a little-endian processor.
7322 @item -mwords-little-endian
7323 @opindex mwords-little-endian
7324 This option only applies when generating code for big-endian processors.
7325 Generate code for a little-endian word order but a big-endian byte
7326 order. That is, a byte order of the form @samp{32107654}. Note: this
7327 option should only be used if you require compatibility with code for
7328 big-endian ARM processors generated by versions of the compiler prior to
7331 @item -mcpu=@var{name}
7333 This specifies the name of the target ARM processor. GCC uses this name
7334 to determine what kind of instructions it can emit when generating
7335 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7336 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7337 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7338 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7339 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7340 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7341 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7342 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7343 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7344 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7345 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7346 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7347 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7348 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7351 @itemx -mtune=@var{name}
7353 This option is very similar to the @option{-mcpu=} option, except that
7354 instead of specifying the actual target processor type, and hence
7355 restricting which instructions can be used, it specifies that GCC should
7356 tune the performance of the code as if the target were of the type
7357 specified in this option, but still choosing the instructions that it
7358 will generate based on the cpu specified by a @option{-mcpu=} option.
7359 For some ARM implementations better performance can be obtained by using
7362 @item -march=@var{name}
7364 This specifies the name of the target ARM architecture. GCC uses this
7365 name to determine what kind of instructions it can emit when generating
7366 assembly code. This option can be used in conjunction with or instead
7367 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7368 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7369 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7370 @samp{iwmmxt}, @samp{ep9312}.
7372 @item -mfpu=@var{name}
7373 @itemx -mfpe=@var{number}
7374 @itemx -mfp=@var{number}
7378 This specifies what floating point hardware (or hardware emulation) is
7379 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7380 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7381 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7382 with older versions of GCC@.
7384 If @option{-msoft-float} is specified this specifies the format of
7385 floating point values.
7387 @item -mstructure-size-boundary=@var{n}
7388 @opindex mstructure-size-boundary
7389 The size of all structures and unions will be rounded up to a multiple
7390 of the number of bits set by this option. Permissible values are 8, 32
7391 and 64. The default value varies for different toolchains. For the COFF
7392 targeted toolchain the default value is 8. A value of 64 is only allowed
7393 if the underlying ABI supports it.
7395 Specifying the larger number can produce faster, more efficient code, but
7396 can also increase the size of the program. Different values are potentially
7397 incompatible. Code compiled with one value cannot necessarily expect to
7398 work with code or libraries compiled with another value, if they exchange
7399 information using structures or unions.
7401 @item -mabort-on-noreturn
7402 @opindex mabort-on-noreturn
7403 Generate a call to the function @code{abort} at the end of a
7404 @code{noreturn} function. It will be executed if the function tries to
7408 @itemx -mno-long-calls
7409 @opindex mlong-calls
7410 @opindex mno-long-calls
7411 Tells the compiler to perform function calls by first loading the
7412 address of the function into a register and then performing a subroutine
7413 call on this register. This switch is needed if the target function
7414 will lie outside of the 64 megabyte addressing range of the offset based
7415 version of subroutine call instruction.
7417 Even if this switch is enabled, not all function calls will be turned
7418 into long calls. The heuristic is that static functions, functions
7419 which have the @samp{short-call} attribute, functions that are inside
7420 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7421 definitions have already been compiled within the current compilation
7422 unit, will not be turned into long calls. The exception to this rule is
7423 that weak function definitions, functions with the @samp{long-call}
7424 attribute or the @samp{section} attribute, and functions that are within
7425 the scope of a @samp{#pragma long_calls} directive, will always be
7426 turned into long calls.
7428 This feature is not enabled by default. Specifying
7429 @option{-mno-long-calls} will restore the default behavior, as will
7430 placing the function calls within the scope of a @samp{#pragma
7431 long_calls_off} directive. Note these switches have no effect on how
7432 the compiler generates code to handle function calls via function
7435 @item -mnop-fun-dllimport
7436 @opindex mnop-fun-dllimport
7437 Disable support for the @code{dllimport} attribute.
7439 @item -msingle-pic-base
7440 @opindex msingle-pic-base
7441 Treat the register used for PIC addressing as read-only, rather than
7442 loading it in the prologue for each function. The run-time system is
7443 responsible for initializing this register with an appropriate value
7444 before execution begins.
7446 @item -mpic-register=@var{reg}
7447 @opindex mpic-register
7448 Specify the register to be used for PIC addressing. The default is R10
7449 unless stack-checking is enabled, when R9 is used.
7451 @item -mcirrus-fix-invalid-insns
7452 @opindex mcirrus-fix-invalid-insns
7453 @opindex mno-cirrus-fix-invalid-insns
7454 Insert NOPs into the instruction stream to in order to work around
7455 problems with invalid Maverick instruction combinations. This option
7456 is only valid if the @option{-mcpu=ep9312} option has been used to
7457 enable generation of instructions for the Cirrus Maverick floating
7458 point co-processor. This option is not enabled by default, since the
7459 problem is only present in older Maverick implementations. The default
7460 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7463 @item -mpoke-function-name
7464 @opindex mpoke-function-name
7465 Write the name of each function into the text section, directly
7466 preceding the function prologue. The generated code is similar to this:
7470 .ascii "arm_poke_function_name", 0
7473 .word 0xff000000 + (t1 - t0)
7474 arm_poke_function_name
7476 stmfd sp!, @{fp, ip, lr, pc@}
7480 When performing a stack backtrace, code can inspect the value of
7481 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7482 location @code{pc - 12} and the top 8 bits are set, then we know that
7483 there is a function name embedded immediately preceding this location
7484 and has length @code{((pc[-3]) & 0xff000000)}.
7488 Generate code for the 16-bit Thumb instruction set. The default is to
7489 use the 32-bit ARM instruction set.
7492 @opindex mtpcs-frame
7493 Generate a stack frame that is compliant with the Thumb Procedure Call
7494 Standard for all non-leaf functions. (A leaf function is one that does
7495 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7497 @item -mtpcs-leaf-frame
7498 @opindex mtpcs-leaf-frame
7499 Generate a stack frame that is compliant with the Thumb Procedure Call
7500 Standard for all leaf functions. (A leaf function is one that does
7501 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7503 @item -mcallee-super-interworking
7504 @opindex mcallee-super-interworking
7505 Gives all externally visible functions in the file being compiled an ARM
7506 instruction set header which switches to Thumb mode before executing the
7507 rest of the function. This allows these functions to be called from
7508 non-interworking code.
7510 @item -mcaller-super-interworking
7511 @opindex mcaller-super-interworking
7512 Allows calls via function pointers (including virtual functions) to
7513 execute correctly regardless of whether the target code has been
7514 compiled for interworking or not. There is a small overhead in the cost
7515 of executing a function pointer if this option is enabled.
7517 @item -mtp=@var{name}
7519 Specify the access model for the thread local storage pointer. The valid
7520 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7521 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7522 (supported in the arm6k architecture), and @option{auto}, which uses the
7523 best available method for the selected processor. The default setting is
7529 @subsection AVR Options
7532 These options are defined for AVR implementations:
7535 @item -mmcu=@var{mcu}
7537 Specify ATMEL AVR instruction set or MCU type.
7539 Instruction set avr1 is for the minimal AVR core, not supported by the C
7540 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7541 attiny11, attiny12, attiny15, attiny28).
7543 Instruction set avr2 (default) is for the classic AVR core with up to
7544 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7545 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7546 at90c8534, at90s8535).
7548 Instruction set avr3 is for the classic AVR core with up to 128K program
7549 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7551 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7552 memory space (MCU types: atmega8, atmega83, atmega85).
7554 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7555 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7556 atmega64, atmega128, at43usb355, at94k).
7560 Output instruction sizes to the asm file.
7562 @item -minit-stack=@var{N}
7563 @opindex minit-stack
7564 Specify the initial stack address, which may be a symbol or numeric value,
7565 @samp{__stack} is the default.
7567 @item -mno-interrupts
7568 @opindex mno-interrupts
7569 Generated code is not compatible with hardware interrupts.
7570 Code size will be smaller.
7572 @item -mcall-prologues
7573 @opindex mcall-prologues
7574 Functions prologues/epilogues expanded as call to appropriate
7575 subroutines. Code size will be smaller.
7577 @item -mno-tablejump
7578 @opindex mno-tablejump
7579 Do not generate tablejump insns which sometimes increase code size.
7582 @opindex mtiny-stack
7583 Change only the low 8 bits of the stack pointer.
7587 Assume int to be 8 bit integer. This affects the sizes of all types: A
7588 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7589 and long long will be 4 bytes. Please note that this option does not
7590 comply to the C standards, but it will provide you with smaller code
7594 @node Blackfin Options
7595 @subsection Blackfin Options
7596 @cindex Blackfin Options
7599 @item -momit-leaf-frame-pointer
7600 @opindex momit-leaf-frame-pointer
7601 Don't keep the frame pointer in a register for leaf functions. This
7602 avoids the instructions to save, set up and restore frame pointers and
7603 makes an extra register available in leaf functions. The option
7604 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7605 which might make debugging harder.
7607 @item -mspecld-anomaly
7608 @opindex mspecld-anomaly
7609 When enabled, the compiler will ensure that the generated code does not
7610 contain speculative loads after jump instructions. This option is enabled
7613 @item -mno-specld-anomaly
7614 @opindex mno-specld-anomaly
7615 Don't generate extra code to prevent speculative loads from occurring.
7617 @item -mcsync-anomaly
7618 @opindex mcsync-anomaly
7619 When enabled, the compiler will ensure that the generated code does not
7620 contain CSYNC or SSYNC instructions too soon after conditional branches.
7621 This option is enabled by default.
7623 @item -mno-csync-anomaly
7624 @opindex mno-csync-anomaly
7625 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7626 occurring too soon after a conditional branch.
7630 When enabled, the compiler is free to take advantage of the knowledge that
7631 the entire program fits into the low 64k of memory.
7634 @opindex mno-low-64k
7635 Assume that the program is arbitrarily large. This is the default.
7637 @item -mid-shared-library
7638 @opindex mid-shared-library
7639 Generate code that supports shared libraries via the library ID method.
7640 This allows for execute in place and shared libraries in an environment
7641 without virtual memory management. This option implies @option{-fPIC}.
7643 @item -mno-id-shared-library
7644 @opindex mno-id-shared-library
7645 Generate code that doesn't assume ID based shared libraries are being used.
7646 This is the default.
7648 @item -mshared-library-id=n
7649 @opindex mshared-library-id
7650 Specified the identification number of the ID based shared library being
7651 compiled. Specifying a value of 0 will generate more compact code, specifying
7652 other values will force the allocation of that number to the current
7653 library but is no more space or time efficient than omitting this option.
7656 @itemx -mno-long-calls
7657 @opindex mlong-calls
7658 @opindex mno-long-calls
7659 Tells the compiler to perform function calls by first loading the
7660 address of the function into a register and then performing a subroutine
7661 call on this register. This switch is needed if the target function
7662 will lie outside of the 24 bit addressing range of the offset based
7663 version of subroutine call instruction.
7665 This feature is not enabled by default. Specifying
7666 @option{-mno-long-calls} will restore the default behavior. Note these
7667 switches have no effect on how the compiler generates code to handle
7668 function calls via function pointers.
7672 @subsection CRIS Options
7673 @cindex CRIS Options
7675 These options are defined specifically for the CRIS ports.
7678 @item -march=@var{architecture-type}
7679 @itemx -mcpu=@var{architecture-type}
7682 Generate code for the specified architecture. The choices for
7683 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7684 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7685 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7688 @item -mtune=@var{architecture-type}
7690 Tune to @var{architecture-type} everything applicable about the generated
7691 code, except for the ABI and the set of available instructions. The
7692 choices for @var{architecture-type} are the same as for
7693 @option{-march=@var{architecture-type}}.
7695 @item -mmax-stack-frame=@var{n}
7696 @opindex mmax-stack-frame
7697 Warn when the stack frame of a function exceeds @var{n} bytes.
7699 @item -melinux-stacksize=@var{n}
7700 @opindex melinux-stacksize
7701 Only available with the @samp{cris-axis-aout} target. Arranges for
7702 indications in the program to the kernel loader that the stack of the
7703 program should be set to @var{n} bytes.
7709 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7710 @option{-march=v3} and @option{-march=v8} respectively.
7712 @item -mmul-bug-workaround
7713 @itemx -mno-mul-bug-workaround
7714 @opindex mmul-bug-workaround
7715 @opindex mno-mul-bug-workaround
7716 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7717 models where it applies. This option is active by default.
7721 Enable CRIS-specific verbose debug-related information in the assembly
7722 code. This option also has the effect to turn off the @samp{#NO_APP}
7723 formatted-code indicator to the assembler at the beginning of the
7728 Do not use condition-code results from previous instruction; always emit
7729 compare and test instructions before use of condition codes.
7731 @item -mno-side-effects
7732 @opindex mno-side-effects
7733 Do not emit instructions with side-effects in addressing modes other than
7737 @itemx -mno-stack-align
7739 @itemx -mno-data-align
7740 @itemx -mconst-align
7741 @itemx -mno-const-align
7742 @opindex mstack-align
7743 @opindex mno-stack-align
7744 @opindex mdata-align
7745 @opindex mno-data-align
7746 @opindex mconst-align
7747 @opindex mno-const-align
7748 These options (no-options) arranges (eliminate arrangements) for the
7749 stack-frame, individual data and constants to be aligned for the maximum
7750 single data access size for the chosen CPU model. The default is to
7751 arrange for 32-bit alignment. ABI details such as structure layout are
7752 not affected by these options.
7760 Similar to the stack- data- and const-align options above, these options
7761 arrange for stack-frame, writable data and constants to all be 32-bit,
7762 16-bit or 8-bit aligned. The default is 32-bit alignment.
7764 @item -mno-prologue-epilogue
7765 @itemx -mprologue-epilogue
7766 @opindex mno-prologue-epilogue
7767 @opindex mprologue-epilogue
7768 With @option{-mno-prologue-epilogue}, the normal function prologue and
7769 epilogue that sets up the stack-frame are omitted and no return
7770 instructions or return sequences are generated in the code. Use this
7771 option only together with visual inspection of the compiled code: no
7772 warnings or errors are generated when call-saved registers must be saved,
7773 or storage for local variable needs to be allocated.
7779 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7780 instruction sequences that load addresses for functions from the PLT part
7781 of the GOT rather than (traditional on other architectures) calls to the
7782 PLT@. The default is @option{-mgotplt}.
7786 Legacy no-op option only recognized with the cris-axis-aout target.
7790 Legacy no-op option only recognized with the cris-axis-elf and
7791 cris-axis-linux-gnu targets.
7795 Only recognized with the cris-axis-aout target, where it selects a
7796 GNU/linux-like multilib, include files and instruction set for
7801 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7805 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7806 to link with input-output functions from a simulator library. Code,
7807 initialized data and zero-initialized data are allocated consecutively.
7811 Like @option{-sim}, but pass linker options to locate initialized data at
7812 0x40000000 and zero-initialized data at 0x80000000.
7816 @subsection CRX Options
7819 These options are defined specifically for the CRX ports.
7825 Enable the use of multiply-accumulate instructions. Disabled by default.
7829 Push instructions will be used to pass outgoing arguments when functions
7830 are called. Enabled by default.
7833 @node Darwin Options
7834 @subsection Darwin Options
7835 @cindex Darwin options
7837 These options are defined for all architectures running the Darwin operating
7840 FSF GCC on Darwin does not create ``fat'' object files; it will create
7841 an object file for the single architecture that it was built to
7842 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7843 @option{-arch} options are used; it does so by running the compiler or
7844 linker multiple times and joining the results together with
7847 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7848 @samp{i686}) is determined by the flags that specify the ISA
7849 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7850 @option{-force_cpusubtype_ALL} option can be used to override this.
7852 The Darwin tools vary in their behavior when presented with an ISA
7853 mismatch. The assembler, @file{as}, will only permit instructions to
7854 be used that are valid for the subtype of the file it is generating,
7855 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7856 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7857 and print an error if asked to create a shared library with a less
7858 restrictive subtype than its input files (for instance, trying to put
7859 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7860 for executables, @file{ld}, will quietly give the executable the most
7861 restrictive subtype of any of its input files.
7866 Add the framework directory @var{dir} to the head of the list of
7867 directories to be searched for header files. These directories are
7868 interleaved with those specified by @option{-I} options and are
7869 scanned in a left-to-right order.
7871 A framework directory is a directory with frameworks in it. A
7872 framework is a directory with a @samp{"Headers"} and/or
7873 @samp{"PrivateHeaders"} directory contained directly in it that ends
7874 in @samp{".framework"}. The name of a framework is the name of this
7875 directory excluding the @samp{".framework"}. Headers associated with
7876 the framework are found in one of those two directories, with
7877 @samp{"Headers"} being searched first. A subframework is a framework
7878 directory that is in a framework's @samp{"Frameworks"} directory.
7879 Includes of subframework headers can only appear in a header of a
7880 framework that contains the subframework, or in a sibling subframework
7881 header. Two subframeworks are siblings if they occur in the same
7882 framework. A subframework should not have the same name as a
7883 framework, a warning will be issued if this is violated. Currently a
7884 subframework cannot have subframeworks, in the future, the mechanism
7885 may be extended to support this. The standard frameworks can be found
7886 in @samp{"/System/Library/Frameworks"} and
7887 @samp{"/Library/Frameworks"}. An example include looks like
7888 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7889 the name of the framework and header.h is found in the
7890 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7894 Emit debugging information for symbols that are used. For STABS
7895 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7896 This is by default ON@.
7900 Emit debugging information for all symbols and types.
7902 @item -mmacosx-version-min=@var{version}
7903 The earliest version of MacOS X that this executable will run on
7904 is @var{version}. Typical values of @var{version} include @code{10.1},
7905 @code{10.2}, and @code{10.3.9}.
7907 The default for this option is to make choices that seem to be most
7910 @item -mone-byte-bool
7911 @opindex -mone-byte-bool
7912 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7913 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7914 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7915 option has no effect on x86.
7917 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7918 to generate code that is not binary compatible with code generated
7919 without that switch. Using this switch may require recompiling all
7920 other modules in a program, including system libraries. Use this
7921 switch to conform to a non-default data model.
7923 @item -mfix-and-continue
7924 @itemx -ffix-and-continue
7925 @itemx -findirect-data
7926 @opindex mfix-and-continue
7927 @opindex ffix-and-continue
7928 @opindex findirect-data
7929 Generate code suitable for fast turn around development. Needed to
7930 enable gdb to dynamically load @code{.o} files into already running
7931 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7932 are provided for backwards compatibility.
7936 Loads all members of static archive libraries.
7937 See man ld(1) for more information.
7939 @item -arch_errors_fatal
7940 @opindex arch_errors_fatal
7941 Cause the errors having to do with files that have the wrong architecture
7945 @opindex bind_at_load
7946 Causes the output file to be marked such that the dynamic linker will
7947 bind all undefined references when the file is loaded or launched.
7951 Produce a Mach-o bundle format file.
7952 See man ld(1) for more information.
7954 @item -bundle_loader @var{executable}
7955 @opindex bundle_loader
7956 This option specifies the @var{executable} that will be loading the build
7957 output file being linked. See man ld(1) for more information.
7960 @opindex -dynamiclib
7961 When passed this option, GCC will produce a dynamic library instead of
7962 an executable when linking, using the Darwin @file{libtool} command.
7964 @item -force_cpusubtype_ALL
7965 @opindex -force_cpusubtype_ALL
7966 This causes GCC's output file to have the @var{ALL} subtype, instead of
7967 one controlled by the @option{-mcpu} or @option{-march} option.
7969 @item -allowable_client @var{client_name}
7971 @itemx -compatibility_version
7972 @itemx -current_version
7974 @itemx -dependency-file
7976 @itemx -dylinker_install_name
7978 @itemx -exported_symbols_list
7980 @itemx -flat_namespace
7981 @itemx -force_flat_namespace
7982 @itemx -headerpad_max_install_names
7985 @itemx -install_name
7986 @itemx -keep_private_externs
7987 @itemx -multi_module
7988 @itemx -multiply_defined
7989 @itemx -multiply_defined_unused
7991 @itemx -no_dead_strip_inits_and_terms
7992 @itemx -nofixprebinding
7995 @itemx -noseglinkedit
7996 @itemx -pagezero_size
7998 @itemx -prebind_all_twolevel_modules
7999 @itemx -private_bundle
8000 @itemx -read_only_relocs
8002 @itemx -sectobjectsymbols
8006 @itemx -sectobjectsymbols
8009 @itemx -segs_read_only_addr
8010 @itemx -segs_read_write_addr
8011 @itemx -seg_addr_table
8012 @itemx -seg_addr_table_filename
8015 @itemx -segs_read_only_addr
8016 @itemx -segs_read_write_addr
8017 @itemx -single_module
8020 @itemx -sub_umbrella
8021 @itemx -twolevel_namespace
8024 @itemx -unexported_symbols_list
8025 @itemx -weak_reference_mismatches
8028 @opindex allowable_client
8029 @opindex client_name
8030 @opindex compatibility_version
8031 @opindex current_version
8033 @opindex dependency-file
8035 @opindex dylinker_install_name
8037 @opindex exported_symbols_list
8039 @opindex flat_namespace
8040 @opindex force_flat_namespace
8041 @opindex headerpad_max_install_names
8044 @opindex install_name
8045 @opindex keep_private_externs
8046 @opindex multi_module
8047 @opindex multiply_defined
8048 @opindex multiply_defined_unused
8050 @opindex no_dead_strip_inits_and_terms
8051 @opindex nofixprebinding
8052 @opindex nomultidefs
8054 @opindex noseglinkedit
8055 @opindex pagezero_size
8057 @opindex prebind_all_twolevel_modules
8058 @opindex private_bundle
8059 @opindex read_only_relocs
8061 @opindex sectobjectsymbols
8065 @opindex sectobjectsymbols
8068 @opindex segs_read_only_addr
8069 @opindex segs_read_write_addr
8070 @opindex seg_addr_table
8071 @opindex seg_addr_table_filename
8072 @opindex seglinkedit
8074 @opindex segs_read_only_addr
8075 @opindex segs_read_write_addr
8076 @opindex single_module
8078 @opindex sub_library
8079 @opindex sub_umbrella
8080 @opindex twolevel_namespace
8083 @opindex unexported_symbols_list
8084 @opindex weak_reference_mismatches
8085 @opindex whatsloaded
8087 These options are passed to the Darwin linker. The Darwin linker man page
8088 describes them in detail.
8091 @node DEC Alpha Options
8092 @subsection DEC Alpha Options
8094 These @samp{-m} options are defined for the DEC Alpha implementations:
8097 @item -mno-soft-float
8099 @opindex mno-soft-float
8100 @opindex msoft-float
8101 Use (do not use) the hardware floating-point instructions for
8102 floating-point operations. When @option{-msoft-float} is specified,
8103 functions in @file{libgcc.a} will be used to perform floating-point
8104 operations. Unless they are replaced by routines that emulate the
8105 floating-point operations, or compiled in such a way as to call such
8106 emulations routines, these routines will issue floating-point
8107 operations. If you are compiling for an Alpha without floating-point
8108 operations, you must ensure that the library is built so as not to call
8111 Note that Alpha implementations without floating-point operations are
8112 required to have floating-point registers.
8117 @opindex mno-fp-regs
8118 Generate code that uses (does not use) the floating-point register set.
8119 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8120 register set is not used, floating point operands are passed in integer
8121 registers as if they were integers and floating-point results are passed
8122 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8123 so any function with a floating-point argument or return value called by code
8124 compiled with @option{-mno-fp-regs} must also be compiled with that
8127 A typical use of this option is building a kernel that does not use,
8128 and hence need not save and restore, any floating-point registers.
8132 The Alpha architecture implements floating-point hardware optimized for
8133 maximum performance. It is mostly compliant with the IEEE floating
8134 point standard. However, for full compliance, software assistance is
8135 required. This option generates code fully IEEE compliant code
8136 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8137 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8138 defined during compilation. The resulting code is less efficient but is
8139 able to correctly support denormalized numbers and exceptional IEEE
8140 values such as not-a-number and plus/minus infinity. Other Alpha
8141 compilers call this option @option{-ieee_with_no_inexact}.
8143 @item -mieee-with-inexact
8144 @opindex mieee-with-inexact
8145 This is like @option{-mieee} except the generated code also maintains
8146 the IEEE @var{inexact-flag}. Turning on this option causes the
8147 generated code to implement fully-compliant IEEE math. In addition to
8148 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8149 macro. On some Alpha implementations the resulting code may execute
8150 significantly slower than the code generated by default. Since there is
8151 very little code that depends on the @var{inexact-flag}, you should
8152 normally not specify this option. Other Alpha compilers call this
8153 option @option{-ieee_with_inexact}.
8155 @item -mfp-trap-mode=@var{trap-mode}
8156 @opindex mfp-trap-mode
8157 This option controls what floating-point related traps are enabled.
8158 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8159 The trap mode can be set to one of four values:
8163 This is the default (normal) setting. The only traps that are enabled
8164 are the ones that cannot be disabled in software (e.g., division by zero
8168 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8172 Like @samp{su}, but the instructions are marked to be safe for software
8173 completion (see Alpha architecture manual for details).
8176 Like @samp{su}, but inexact traps are enabled as well.
8179 @item -mfp-rounding-mode=@var{rounding-mode}
8180 @opindex mfp-rounding-mode
8181 Selects the IEEE rounding mode. Other Alpha compilers call this option
8182 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8187 Normal IEEE rounding mode. Floating point numbers are rounded towards
8188 the nearest machine number or towards the even machine number in case
8192 Round towards minus infinity.
8195 Chopped rounding mode. Floating point numbers are rounded towards zero.
8198 Dynamic rounding mode. A field in the floating point control register
8199 (@var{fpcr}, see Alpha architecture reference manual) controls the
8200 rounding mode in effect. The C library initializes this register for
8201 rounding towards plus infinity. Thus, unless your program modifies the
8202 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8205 @item -mtrap-precision=@var{trap-precision}
8206 @opindex mtrap-precision
8207 In the Alpha architecture, floating point traps are imprecise. This
8208 means without software assistance it is impossible to recover from a
8209 floating trap and program execution normally needs to be terminated.
8210 GCC can generate code that can assist operating system trap handlers
8211 in determining the exact location that caused a floating point trap.
8212 Depending on the requirements of an application, different levels of
8213 precisions can be selected:
8217 Program precision. This option is the default and means a trap handler
8218 can only identify which program caused a floating point exception.
8221 Function precision. The trap handler can determine the function that
8222 caused a floating point exception.
8225 Instruction precision. The trap handler can determine the exact
8226 instruction that caused a floating point exception.
8229 Other Alpha compilers provide the equivalent options called
8230 @option{-scope_safe} and @option{-resumption_safe}.
8232 @item -mieee-conformant
8233 @opindex mieee-conformant
8234 This option marks the generated code as IEEE conformant. You must not
8235 use this option unless you also specify @option{-mtrap-precision=i} and either
8236 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8237 is to emit the line @samp{.eflag 48} in the function prologue of the
8238 generated assembly file. Under DEC Unix, this has the effect that
8239 IEEE-conformant math library routines will be linked in.
8241 @item -mbuild-constants
8242 @opindex mbuild-constants
8243 Normally GCC examines a 32- or 64-bit integer constant to
8244 see if it can construct it from smaller constants in two or three
8245 instructions. If it cannot, it will output the constant as a literal and
8246 generate code to load it from the data segment at runtime.
8248 Use this option to require GCC to construct @emph{all} integer constants
8249 using code, even if it takes more instructions (the maximum is six).
8251 You would typically use this option to build a shared library dynamic
8252 loader. Itself a shared library, it must relocate itself in memory
8253 before it can find the variables and constants in its own data segment.
8259 Select whether to generate code to be assembled by the vendor-supplied
8260 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8278 Indicate whether GCC should generate code to use the optional BWX,
8279 CIX, FIX and MAX instruction sets. The default is to use the instruction
8280 sets supported by the CPU type specified via @option{-mcpu=} option or that
8281 of the CPU on which GCC was built if none was specified.
8286 @opindex mfloat-ieee
8287 Generate code that uses (does not use) VAX F and G floating point
8288 arithmetic instead of IEEE single and double precision.
8290 @item -mexplicit-relocs
8291 @itemx -mno-explicit-relocs
8292 @opindex mexplicit-relocs
8293 @opindex mno-explicit-relocs
8294 Older Alpha assemblers provided no way to generate symbol relocations
8295 except via assembler macros. Use of these macros does not allow
8296 optimal instruction scheduling. GNU binutils as of version 2.12
8297 supports a new syntax that allows the compiler to explicitly mark
8298 which relocations should apply to which instructions. This option
8299 is mostly useful for debugging, as GCC detects the capabilities of
8300 the assembler when it is built and sets the default accordingly.
8304 @opindex msmall-data
8305 @opindex mlarge-data
8306 When @option{-mexplicit-relocs} is in effect, static data is
8307 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8308 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8309 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8310 16-bit relocations off of the @code{$gp} register. This limits the
8311 size of the small data area to 64KB, but allows the variables to be
8312 directly accessed via a single instruction.
8314 The default is @option{-mlarge-data}. With this option the data area
8315 is limited to just below 2GB@. Programs that require more than 2GB of
8316 data must use @code{malloc} or @code{mmap} to allocate the data in the
8317 heap instead of in the program's data segment.
8319 When generating code for shared libraries, @option{-fpic} implies
8320 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8324 @opindex msmall-text
8325 @opindex mlarge-text
8326 When @option{-msmall-text} is used, the compiler assumes that the
8327 code of the entire program (or shared library) fits in 4MB, and is
8328 thus reachable with a branch instruction. When @option{-msmall-data}
8329 is used, the compiler can assume that all local symbols share the
8330 same @code{$gp} value, and thus reduce the number of instructions
8331 required for a function call from 4 to 1.
8333 The default is @option{-mlarge-text}.
8335 @item -mcpu=@var{cpu_type}
8337 Set the instruction set and instruction scheduling parameters for
8338 machine type @var{cpu_type}. You can specify either the @samp{EV}
8339 style name or the corresponding chip number. GCC supports scheduling
8340 parameters for the EV4, EV5 and EV6 family of processors and will
8341 choose the default values for the instruction set from the processor
8342 you specify. If you do not specify a processor type, GCC will default
8343 to the processor on which the compiler was built.
8345 Supported values for @var{cpu_type} are
8351 Schedules as an EV4 and has no instruction set extensions.
8355 Schedules as an EV5 and has no instruction set extensions.
8359 Schedules as an EV5 and supports the BWX extension.
8364 Schedules as an EV5 and supports the BWX and MAX extensions.
8368 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8372 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8375 @item -mtune=@var{cpu_type}
8377 Set only the instruction scheduling parameters for machine type
8378 @var{cpu_type}. The instruction set is not changed.
8380 @item -mmemory-latency=@var{time}
8381 @opindex mmemory-latency
8382 Sets the latency the scheduler should assume for typical memory
8383 references as seen by the application. This number is highly
8384 dependent on the memory access patterns used by the application
8385 and the size of the external cache on the machine.
8387 Valid options for @var{time} are
8391 A decimal number representing clock cycles.
8397 The compiler contains estimates of the number of clock cycles for
8398 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8399 (also called Dcache, Scache, and Bcache), as well as to main memory.
8400 Note that L3 is only valid for EV5.
8405 @node DEC Alpha/VMS Options
8406 @subsection DEC Alpha/VMS Options
8408 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8411 @item -mvms-return-codes
8412 @opindex mvms-return-codes
8413 Return VMS condition codes from main. The default is to return POSIX
8414 style condition (e.g.@ error) codes.
8418 @subsection FRV Options
8425 Only use the first 32 general purpose registers.
8430 Use all 64 general purpose registers.
8435 Use only the first 32 floating point registers.
8440 Use all 64 floating point registers
8443 @opindex mhard-float
8445 Use hardware instructions for floating point operations.
8448 @opindex msoft-float
8450 Use library routines for floating point operations.
8455 Dynamically allocate condition code registers.
8460 Do not try to dynamically allocate condition code registers, only
8461 use @code{icc0} and @code{fcc0}.
8466 Change ABI to use double word insns.
8471 Do not use double word instructions.
8476 Use floating point double instructions.
8481 Do not use floating point double instructions.
8486 Use media instructions.
8491 Do not use media instructions.
8496 Use multiply and add/subtract instructions.
8501 Do not use multiply and add/subtract instructions.
8506 Select the FDPIC ABI, that uses function descriptors to represent
8507 pointers to functions. Without any PIC/PIE-related options, it
8508 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8509 assumes GOT entries and small data are within a 12-bit range from the
8510 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8511 are computed with 32 bits.
8514 @opindex minline-plt
8516 Enable inlining of PLT entries in function calls to functions that are
8517 not known to bind locally. It has no effect without @option{-mfdpic}.
8518 It's enabled by default if optimizing for speed and compiling for
8519 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8520 optimization option such as @option{-O3} or above is present in the
8526 Assume a large TLS segment when generating thread-local code.
8531 Do not assume a large TLS segment when generating thread-local code.
8536 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8537 that is known to be in read-only sections. It's enabled by default,
8538 except for @option{-fpic} or @option{-fpie}: even though it may help
8539 make the global offset table smaller, it trades 1 instruction for 4.
8540 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8541 one of which may be shared by multiple symbols, and it avoids the need
8542 for a GOT entry for the referenced symbol, so it's more likely to be a
8543 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8545 @item -multilib-library-pic
8546 @opindex multilib-library-pic
8548 Link with the (library, not FD) pic libraries. It's implied by
8549 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8550 @option{-fpic} without @option{-mfdpic}. You should never have to use
8556 Follow the EABI requirement of always creating a frame pointer whenever
8557 a stack frame is allocated. This option is enabled by default and can
8558 be disabled with @option{-mno-linked-fp}.
8561 @opindex mlong-calls
8563 Use indirect addressing to call functions outside the current
8564 compilation unit. This allows the functions to be placed anywhere
8565 within the 32-bit address space.
8567 @item -malign-labels
8568 @opindex malign-labels
8570 Try to align labels to an 8-byte boundary by inserting nops into the
8571 previous packet. This option only has an effect when VLIW packing
8572 is enabled. It doesn't create new packets; it merely adds nops to
8576 @opindex mlibrary-pic
8578 Generate position-independent EABI code.
8583 Use only the first four media accumulator registers.
8588 Use all eight media accumulator registers.
8593 Pack VLIW instructions.
8598 Do not pack VLIW instructions.
8603 Do not mark ABI switches in e_flags.
8608 Enable the use of conditional-move instructions (default).
8610 This switch is mainly for debugging the compiler and will likely be removed
8611 in a future version.
8613 @item -mno-cond-move
8614 @opindex mno-cond-move
8616 Disable the use of conditional-move instructions.
8618 This switch is mainly for debugging the compiler and will likely be removed
8619 in a future version.
8624 Enable the use of conditional set instructions (default).
8626 This switch is mainly for debugging the compiler and will likely be removed
8627 in a future version.
8632 Disable the use of conditional set instructions.
8634 This switch is mainly for debugging the compiler and will likely be removed
8635 in a future version.
8640 Enable the use of conditional execution (default).
8642 This switch is mainly for debugging the compiler and will likely be removed
8643 in a future version.
8645 @item -mno-cond-exec
8646 @opindex mno-cond-exec
8648 Disable the use of conditional execution.
8650 This switch is mainly for debugging the compiler and will likely be removed
8651 in a future version.
8654 @opindex mvliw-branch
8656 Run a pass to pack branches into VLIW instructions (default).
8658 This switch is mainly for debugging the compiler and will likely be removed
8659 in a future version.
8661 @item -mno-vliw-branch
8662 @opindex mno-vliw-branch
8664 Do not run a pass to pack branches into VLIW instructions.
8666 This switch is mainly for debugging the compiler and will likely be removed
8667 in a future version.
8669 @item -mmulti-cond-exec
8670 @opindex mmulti-cond-exec
8672 Enable optimization of @code{&&} and @code{||} in conditional execution
8675 This switch is mainly for debugging the compiler and will likely be removed
8676 in a future version.
8678 @item -mno-multi-cond-exec
8679 @opindex mno-multi-cond-exec
8681 Disable optimization of @code{&&} and @code{||} in conditional execution.
8683 This switch is mainly for debugging the compiler and will likely be removed
8684 in a future version.
8686 @item -mnested-cond-exec
8687 @opindex mnested-cond-exec
8689 Enable nested conditional execution optimizations (default).
8691 This switch is mainly for debugging the compiler and will likely be removed
8692 in a future version.
8694 @item -mno-nested-cond-exec
8695 @opindex mno-nested-cond-exec
8697 Disable nested conditional execution optimizations.
8699 This switch is mainly for debugging the compiler and will likely be removed
8700 in a future version.
8702 @item -moptimize-membar
8703 @opindex moptimize-membar
8705 This switch removes redundant @code{membar} instructions from the
8706 compiler generated code. It is enabled by default.
8708 @item -mno-optimize-membar
8709 @opindex mno-optimize-membar
8711 This switch disables the automatic removal of redundant @code{membar}
8712 instructions from the generated code.
8714 @item -mtomcat-stats
8715 @opindex mtomcat-stats
8717 Cause gas to print out tomcat statistics.
8719 @item -mcpu=@var{cpu}
8722 Select the processor type for which to generate code. Possible values are
8723 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8724 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8728 @node H8/300 Options
8729 @subsection H8/300 Options
8731 These @samp{-m} options are defined for the H8/300 implementations:
8736 Shorten some address references at link time, when possible; uses the
8737 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8738 ld, Using ld}, for a fuller description.
8742 Generate code for the H8/300H@.
8746 Generate code for the H8S@.
8750 Generate code for the H8S and H8/300H in the normal mode. This switch
8751 must be used either with @option{-mh} or @option{-ms}.
8755 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8759 Make @code{int} data 32 bits by default.
8763 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8764 The default for the H8/300H and H8S is to align longs and floats on 4
8766 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8767 This option has no effect on the H8/300.
8771 @subsection HPPA Options
8772 @cindex HPPA Options
8774 These @samp{-m} options are defined for the HPPA family of computers:
8777 @item -march=@var{architecture-type}
8779 Generate code for the specified architecture. The choices for
8780 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8781 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8782 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8783 architecture option for your machine. Code compiled for lower numbered
8784 architectures will run on higher numbered architectures, but not the
8788 @itemx -mpa-risc-1-1
8789 @itemx -mpa-risc-2-0
8790 @opindex mpa-risc-1-0
8791 @opindex mpa-risc-1-1
8792 @opindex mpa-risc-2-0
8793 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8796 @opindex mbig-switch
8797 Generate code suitable for big switch tables. Use this option only if
8798 the assembler/linker complain about out of range branches within a switch
8801 @item -mjump-in-delay
8802 @opindex mjump-in-delay
8803 Fill delay slots of function calls with unconditional jump instructions
8804 by modifying the return pointer for the function call to be the target
8805 of the conditional jump.
8807 @item -mdisable-fpregs
8808 @opindex mdisable-fpregs
8809 Prevent floating point registers from being used in any manner. This is
8810 necessary for compiling kernels which perform lazy context switching of
8811 floating point registers. If you use this option and attempt to perform
8812 floating point operations, the compiler will abort.
8814 @item -mdisable-indexing
8815 @opindex mdisable-indexing
8816 Prevent the compiler from using indexing address modes. This avoids some
8817 rather obscure problems when compiling MIG generated code under MACH@.
8819 @item -mno-space-regs
8820 @opindex mno-space-regs
8821 Generate code that assumes the target has no space registers. This allows
8822 GCC to generate faster indirect calls and use unscaled index address modes.
8824 Such code is suitable for level 0 PA systems and kernels.
8826 @item -mfast-indirect-calls
8827 @opindex mfast-indirect-calls
8828 Generate code that assumes calls never cross space boundaries. This
8829 allows GCC to emit code which performs faster indirect calls.
8831 This option will not work in the presence of shared libraries or nested
8834 @item -mfixed-range=@var{register-range}
8835 @opindex mfixed-range
8836 Generate code treating the given register range as fixed registers.
8837 A fixed register is one that the register allocator can not use. This is
8838 useful when compiling kernel code. A register range is specified as
8839 two registers separated by a dash. Multiple register ranges can be
8840 specified separated by a comma.
8842 @item -mlong-load-store
8843 @opindex mlong-load-store
8844 Generate 3-instruction load and store sequences as sometimes required by
8845 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8848 @item -mportable-runtime
8849 @opindex mportable-runtime
8850 Use the portable calling conventions proposed by HP for ELF systems.
8854 Enable the use of assembler directives only GAS understands.
8856 @item -mschedule=@var{cpu-type}
8858 Schedule code according to the constraints for the machine type
8859 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8860 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8861 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8862 proper scheduling option for your machine. The default scheduling is
8866 @opindex mlinker-opt
8867 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8868 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8869 linkers in which they give bogus error messages when linking some programs.
8872 @opindex msoft-float
8873 Generate output containing library calls for floating point.
8874 @strong{Warning:} the requisite libraries are not available for all HPPA
8875 targets. Normally the facilities of the machine's usual C compiler are
8876 used, but this cannot be done directly in cross-compilation. You must make
8877 your own arrangements to provide suitable library functions for
8878 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8879 does provide software floating point support.
8881 @option{-msoft-float} changes the calling convention in the output file;
8882 therefore, it is only useful if you compile @emph{all} of a program with
8883 this option. In particular, you need to compile @file{libgcc.a}, the
8884 library that comes with GCC, with @option{-msoft-float} in order for
8889 Generate the predefine, @code{_SIO}, for server IO@. The default is
8890 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8891 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8892 options are available under HP-UX and HI-UX@.
8896 Use GNU ld specific options. This passes @option{-shared} to ld when
8897 building a shared library. It is the default when GCC is configured,
8898 explicitly or implicitly, with the GNU linker. This option does not
8899 have any affect on which ld is called, it only changes what parameters
8900 are passed to that ld. The ld that is called is determined by the
8901 @option{--with-ld} configure option, GCC's program search path, and
8902 finally by the user's @env{PATH}. The linker used by GCC can be printed
8903 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
8904 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8908 Use HP ld specific options. This passes @option{-b} to ld when building
8909 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8910 links. It is the default when GCC is configured, explicitly or
8911 implicitly, with the HP linker. This option does not have any affect on
8912 which ld is called, it only changes what parameters are passed to that
8913 ld. The ld that is called is determined by the @option{--with-ld}
8914 configure option, GCC's program search path, and finally by the user's
8915 @env{PATH}. The linker used by GCC can be printed using @samp{which
8916 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
8917 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8920 @opindex mno-long-calls
8921 Generate code that uses long call sequences. This ensures that a call
8922 is always able to reach linker generated stubs. The default is to generate
8923 long calls only when the distance from the call site to the beginning
8924 of the function or translation unit, as the case may be, exceeds a
8925 predefined limit set by the branch type being used. The limits for
8926 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8927 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8930 Distances are measured from the beginning of functions when using the
8931 @option{-ffunction-sections} option, or when using the @option{-mgas}
8932 and @option{-mno-portable-runtime} options together under HP-UX with
8935 It is normally not desirable to use this option as it will degrade
8936 performance. However, it may be useful in large applications,
8937 particularly when partial linking is used to build the application.
8939 The types of long calls used depends on the capabilities of the
8940 assembler and linker, and the type of code being generated. The
8941 impact on systems that support long absolute calls, and long pic
8942 symbol-difference or pc-relative calls should be relatively small.
8943 However, an indirect call is used on 32-bit ELF systems in pic code
8944 and it is quite long.
8946 @item -munix=@var{unix-std}
8948 Generate compiler predefines and select a startfile for the specified
8949 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8950 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8951 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8952 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8953 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8956 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8957 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8958 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8959 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8960 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8961 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8963 It is @emph{important} to note that this option changes the interfaces
8964 for various library routines. It also affects the operational behavior
8965 of the C library. Thus, @emph{extreme} care is needed in using this
8968 Library code that is intended to operate with more than one UNIX
8969 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8970 as appropriate. Most GNU software doesn't provide this capability.
8974 Suppress the generation of link options to search libdld.sl when the
8975 @option{-static} option is specified on HP-UX 10 and later.
8979 The HP-UX implementation of setlocale in libc has a dependency on
8980 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8981 when the @option{-static} option is specified, special link options
8982 are needed to resolve this dependency.
8984 On HP-UX 10 and later, the GCC driver adds the necessary options to
8985 link with libdld.sl when the @option{-static} option is specified.
8986 This causes the resulting binary to be dynamic. On the 64-bit port,
8987 the linkers generate dynamic binaries by default in any case. The
8988 @option{-nolibdld} option can be used to prevent the GCC driver from
8989 adding these link options.
8993 Add support for multithreading with the @dfn{dce thread} library
8994 under HP-UX@. This option sets flags for both the preprocessor and
8998 @node i386 and x86-64 Options
8999 @subsection Intel 386 and AMD x86-64 Options
9000 @cindex i386 Options
9001 @cindex x86-64 Options
9002 @cindex Intel 386 Options
9003 @cindex AMD x86-64 Options
9005 These @samp{-m} options are defined for the i386 and x86-64 family of
9009 @item -mtune=@var{cpu-type}
9011 Tune to @var{cpu-type} everything applicable about the generated code, except
9012 for the ABI and the set of available instructions. The choices for
9016 Original Intel's i386 CPU@.
9018 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9020 Intel Pentium CPU with no MMX support.
9022 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9023 @item i686, pentiumpro
9024 Intel PentiumPro CPU@.
9026 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9027 @item pentium3, pentium3m
9028 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9031 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9032 support. Used by Centrino notebooks.
9033 @item pentium4, pentium4m
9034 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9036 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9039 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9040 SSE2 and SSE3 instruction set support.
9042 AMD K6 CPU with MMX instruction set support.
9044 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9045 @item athlon, athlon-tbird
9046 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9048 @item athlon-4, athlon-xp, athlon-mp
9049 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9050 instruction set support.
9051 @item k8, opteron, athlon64, athlon-fx
9052 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9053 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9055 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9058 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9059 instruction set support.
9061 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9062 implemented for this chip.)
9064 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9065 implemented for this chip.)
9068 While picking a specific @var{cpu-type} will schedule things appropriately
9069 for that particular chip, the compiler will not generate any code that
9070 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9073 @item -march=@var{cpu-type}
9075 Generate instructions for the machine type @var{cpu-type}. The choices
9076 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9077 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9079 @item -mcpu=@var{cpu-type}
9081 A deprecated synonym for @option{-mtune}.
9090 @opindex mpentiumpro
9091 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9092 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9093 These synonyms are deprecated.
9095 @item -mfpmath=@var{unit}
9097 Generate floating point arithmetics for selected unit @var{unit}. The choices
9102 Use the standard 387 floating point coprocessor present majority of chips and
9103 emulated otherwise. Code compiled with this option will run almost everywhere.
9104 The temporary results are computed in 80bit precision instead of precision
9105 specified by the type resulting in slightly different results compared to most
9106 of other chips. See @option{-ffloat-store} for more detailed description.
9108 This is the default choice for i386 compiler.
9111 Use scalar floating point instructions present in the SSE instruction set.
9112 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9113 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9114 instruction set supports only single precision arithmetics, thus the double and
9115 extended precision arithmetics is still done using 387. Later version, present
9116 only in Pentium4 and the future AMD x86-64 chips supports double precision
9119 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9120 or @option{-msse2} switches to enable SSE extensions and make this option
9121 effective. For the x86-64 compiler, these extensions are enabled by default.
9123 The resulting code should be considerably faster in the majority of cases and avoid
9124 the numerical instability problems of 387 code, but may break some existing
9125 code that expects temporaries to be 80bit.
9127 This is the default choice for the x86-64 compiler.
9130 Attempt to utilize both instruction sets at once. This effectively double the
9131 amount of available registers and on chips with separate execution units for
9132 387 and SSE the execution resources too. Use this option with care, as it is
9133 still experimental, because the GCC register allocator does not model separate
9134 functional units well resulting in instable performance.
9137 @item -masm=@var{dialect}
9138 @opindex masm=@var{dialect}
9139 Output asm instructions using selected @var{dialect}. Supported
9140 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9141 not support @samp{intel}.
9146 @opindex mno-ieee-fp
9147 Control whether or not the compiler uses IEEE floating point
9148 comparisons. These handle correctly the case where the result of a
9149 comparison is unordered.
9152 @opindex msoft-float
9153 Generate output containing library calls for floating point.
9154 @strong{Warning:} the requisite libraries are not part of GCC@.
9155 Normally the facilities of the machine's usual C compiler are used, but
9156 this can't be done directly in cross-compilation. You must make your
9157 own arrangements to provide suitable library functions for
9160 On machines where a function returns floating point results in the 80387
9161 register stack, some floating point opcodes may be emitted even if
9162 @option{-msoft-float} is used.
9164 @item -mno-fp-ret-in-387
9165 @opindex mno-fp-ret-in-387
9166 Do not use the FPU registers for return values of functions.
9168 The usual calling convention has functions return values of types
9169 @code{float} and @code{double} in an FPU register, even if there
9170 is no FPU@. The idea is that the operating system should emulate
9173 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9174 in ordinary CPU registers instead.
9176 @item -mno-fancy-math-387
9177 @opindex mno-fancy-math-387
9178 Some 387 emulators do not support the @code{sin}, @code{cos} and
9179 @code{sqrt} instructions for the 387. Specify this option to avoid
9180 generating those instructions. This option is the default on FreeBSD,
9181 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9182 indicates that the target cpu will always have an FPU and so the
9183 instruction will not need emulation. As of revision 2.6.1, these
9184 instructions are not generated unless you also use the
9185 @option{-funsafe-math-optimizations} switch.
9187 @item -malign-double
9188 @itemx -mno-align-double
9189 @opindex malign-double
9190 @opindex mno-align-double
9191 Control whether GCC aligns @code{double}, @code{long double}, and
9192 @code{long long} variables on a two word boundary or a one word
9193 boundary. Aligning @code{double} variables on a two word boundary will
9194 produce code that runs somewhat faster on a @samp{Pentium} at the
9195 expense of more memory.
9197 @strong{Warning:} if you use the @option{-malign-double} switch,
9198 structures containing the above types will be aligned differently than
9199 the published application binary interface specifications for the 386
9200 and will not be binary compatible with structures in code compiled
9201 without that switch.
9203 @item -m96bit-long-double
9204 @itemx -m128bit-long-double
9205 @opindex m96bit-long-double
9206 @opindex m128bit-long-double
9207 These switches control the size of @code{long double} type. The i386
9208 application binary interface specifies the size to be 96 bits,
9209 so @option{-m96bit-long-double} is the default in 32 bit mode.
9211 Modern architectures (Pentium and newer) would prefer @code{long double}
9212 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9213 conforming to the ABI, this would not be possible. So specifying a
9214 @option{-m128bit-long-double} will align @code{long double}
9215 to a 16 byte boundary by padding the @code{long double} with an additional
9218 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9219 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9221 Notice that neither of these options enable any extra precision over the x87
9222 standard of 80 bits for a @code{long double}.
9224 @strong{Warning:} if you override the default value for your target ABI, the
9225 structures and arrays containing @code{long double} variables will change
9226 their size as well as function calling convention for function taking
9227 @code{long double} will be modified. Hence they will not be binary
9228 compatible with arrays or structures in code compiled without that switch.
9230 @item -mmlarge-data-threshold=@var{number}
9231 @opindex mlarge-data-threshold=@var{number}
9232 When @option{-mcmodel=medium} is specified, the data greater than
9233 @var{threshold} are placed in large data section. This value must be the
9234 same across all object linked into the binary and defaults to 65535.
9237 @itemx -mno-svr3-shlib
9238 @opindex msvr3-shlib
9239 @opindex mno-svr3-shlib
9240 Control whether GCC places uninitialized local variables into the
9241 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9242 into @code{bss}. These options are meaningful only on System V Release 3.
9246 Use a different function-calling convention, in which functions that
9247 take a fixed number of arguments return with the @code{ret} @var{num}
9248 instruction, which pops their arguments while returning. This saves one
9249 instruction in the caller since there is no need to pop the arguments
9252 You can specify that an individual function is called with this calling
9253 sequence with the function attribute @samp{stdcall}. You can also
9254 override the @option{-mrtd} option by using the function attribute
9255 @samp{cdecl}. @xref{Function Attributes}.
9257 @strong{Warning:} this calling convention is incompatible with the one
9258 normally used on Unix, so you cannot use it if you need to call
9259 libraries compiled with the Unix compiler.
9261 Also, you must provide function prototypes for all functions that
9262 take variable numbers of arguments (including @code{printf});
9263 otherwise incorrect code will be generated for calls to those
9266 In addition, seriously incorrect code will result if you call a
9267 function with too many arguments. (Normally, extra arguments are
9268 harmlessly ignored.)
9270 @item -mregparm=@var{num}
9272 Control how many registers are used to pass integer arguments. By
9273 default, no registers are used to pass arguments, and at most 3
9274 registers can be used. You can control this behavior for a specific
9275 function by using the function attribute @samp{regparm}.
9276 @xref{Function Attributes}.
9278 @strong{Warning:} if you use this switch, and
9279 @var{num} is nonzero, then you must build all modules with the same
9280 value, including any libraries. This includes the system libraries and
9284 @opindex msseregparm
9285 Use SSE register passing conventions for float and double arguments
9286 and return values. You can control this behavior for a specific
9287 function by using the function attribute @samp{sseregparm}.
9288 @xref{Function Attributes}.
9290 @strong{Warning:} if you use this switch then you must build all
9291 modules with the same value, including any libraries. This includes
9292 the system libraries and startup modules.
9294 @item -mpreferred-stack-boundary=@var{num}
9295 @opindex mpreferred-stack-boundary
9296 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9297 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9298 the default is 4 (16 bytes or 128 bits), except when optimizing for code
9299 size (@option{-Os}), in which case the default is the minimum correct
9300 alignment (4 bytes for x86, and 8 bytes for x86-64).
9302 On Pentium and PentiumPro, @code{double} and @code{long double} values
9303 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9304 suffer significant run time performance penalties. On Pentium III, the
9305 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
9306 penalties if it is not 16 byte aligned.
9308 To ensure proper alignment of this values on the stack, the stack boundary
9309 must be as aligned as that required by any value stored on the stack.
9310 Further, every function must be generated such that it keeps the stack
9311 aligned. Thus calling a function compiled with a higher preferred
9312 stack boundary from a function compiled with a lower preferred stack
9313 boundary will most likely misalign the stack. It is recommended that
9314 libraries that use callbacks always use the default setting.
9316 This extra alignment does consume extra stack space, and generally
9317 increases code size. Code that is sensitive to stack space usage, such
9318 as embedded systems and operating system kernels, may want to reduce the
9319 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9337 These switches enable or disable the use of instructions in the MMX,
9338 SSE, SSE2 or 3DNow! extended instruction sets. These extensions are
9339 also available as built-in functions: see @ref{X86 Built-in Functions},
9340 for details of the functions enabled and disabled by these switches.
9342 To have SSE/SSE2 instructions generated automatically from floating-point
9343 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9345 These options will enable GCC to use these extended instructions in
9346 generated code, even without @option{-mfpmath=sse}. Applications which
9347 perform runtime CPU detection must compile separate files for each
9348 supported architecture, using the appropriate flags. In particular,
9349 the file containing the CPU detection code should be compiled without
9353 @itemx -mno-push-args
9355 @opindex mno-push-args
9356 Use PUSH operations to store outgoing parameters. This method is shorter
9357 and usually equally fast as method using SUB/MOV operations and is enabled
9358 by default. In some cases disabling it may improve performance because of
9359 improved scheduling and reduced dependencies.
9361 @item -maccumulate-outgoing-args
9362 @opindex maccumulate-outgoing-args
9363 If enabled, the maximum amount of space required for outgoing arguments will be
9364 computed in the function prologue. This is faster on most modern CPUs
9365 because of reduced dependencies, improved scheduling and reduced stack usage
9366 when preferred stack boundary is not equal to 2. The drawback is a notable
9367 increase in code size. This switch implies @option{-mno-push-args}.
9371 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9372 on thread-safe exception handling must compile and link all code with the
9373 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9374 @option{-D_MT}; when linking, it links in a special thread helper library
9375 @option{-lmingwthrd} which cleans up per thread exception handling data.
9377 @item -mno-align-stringops
9378 @opindex mno-align-stringops
9379 Do not align destination of inlined string operations. This switch reduces
9380 code size and improves performance in case the destination is already aligned,
9381 but GCC doesn't know about it.
9383 @item -minline-all-stringops
9384 @opindex minline-all-stringops
9385 By default GCC inlines string operations only when destination is known to be
9386 aligned at least to 4 byte boundary. This enables more inlining, increase code
9387 size, but may improve performance of code that depends on fast memcpy, strlen
9388 and memset for short lengths.
9390 @item -momit-leaf-frame-pointer
9391 @opindex momit-leaf-frame-pointer
9392 Don't keep the frame pointer in a register for leaf functions. This
9393 avoids the instructions to save, set up and restore frame pointers and
9394 makes an extra register available in leaf functions. The option
9395 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9396 which might make debugging harder.
9398 @item -mtls-direct-seg-refs
9399 @itemx -mno-tls-direct-seg-refs
9400 @opindex mtls-direct-seg-refs
9401 Controls whether TLS variables may be accessed with offsets from the
9402 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9403 or whether the thread base pointer must be added. Whether or not this
9404 is legal depends on the operating system, and whether it maps the
9405 segment to cover the entire TLS area.
9407 For systems that use GNU libc, the default is on.
9410 These @samp{-m} switches are supported in addition to the above
9411 on AMD x86-64 processors in 64-bit environments.
9418 Generate code for a 32-bit or 64-bit environment.
9419 The 32-bit environment sets int, long and pointer to 32 bits and
9420 generates code that runs on any i386 system.
9421 The 64-bit environment sets int to 32 bits and long and pointer
9422 to 64 bits and generates code for AMD's x86-64 architecture.
9425 @opindex no-red-zone
9426 Do not use a so called red zone for x86-64 code. The red zone is mandated
9427 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9428 stack pointer that will not be modified by signal or interrupt handlers
9429 and therefore can be used for temporary data without adjusting the stack
9430 pointer. The flag @option{-mno-red-zone} disables this red zone.
9432 @item -mcmodel=small
9433 @opindex mcmodel=small
9434 Generate code for the small code model: the program and its symbols must
9435 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9436 Programs can be statically or dynamically linked. This is the default
9439 @item -mcmodel=kernel
9440 @opindex mcmodel=kernel
9441 Generate code for the kernel code model. The kernel runs in the
9442 negative 2 GB of the address space.
9443 This model has to be used for Linux kernel code.
9445 @item -mcmodel=medium
9446 @opindex mcmodel=medium
9447 Generate code for the medium model: The program is linked in the lower 2
9448 GB of the address space but symbols can be located anywhere in the
9449 address space. Programs can be statically or dynamically linked, but
9450 building of shared libraries are not supported with the medium model.
9452 @item -mcmodel=large
9453 @opindex mcmodel=large
9454 Generate code for the large model: This model makes no assumptions
9455 about addresses and sizes of sections. Currently GCC does not implement
9460 @subsection IA-64 Options
9461 @cindex IA-64 Options
9463 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9467 @opindex mbig-endian
9468 Generate code for a big endian target. This is the default for HP-UX@.
9470 @item -mlittle-endian
9471 @opindex mlittle-endian
9472 Generate code for a little endian target. This is the default for AIX5
9479 Generate (or don't) code for the GNU assembler. This is the default.
9480 @c Also, this is the default if the configure option @option{--with-gnu-as}
9487 Generate (or don't) code for the GNU linker. This is the default.
9488 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9493 Generate code that does not use a global pointer register. The result
9494 is not position independent code, and violates the IA-64 ABI@.
9496 @item -mvolatile-asm-stop
9497 @itemx -mno-volatile-asm-stop
9498 @opindex mvolatile-asm-stop
9499 @opindex mno-volatile-asm-stop
9500 Generate (or don't) a stop bit immediately before and after volatile asm
9503 @item -mregister-names
9504 @itemx -mno-register-names
9505 @opindex mregister-names
9506 @opindex mno-register-names
9507 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9508 the stacked registers. This may make assembler output more readable.
9514 Disable (or enable) optimizations that use the small data section. This may
9515 be useful for working around optimizer bugs.
9518 @opindex mconstant-gp
9519 Generate code that uses a single constant global pointer value. This is
9520 useful when compiling kernel code.
9524 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9525 This is useful when compiling firmware code.
9527 @item -minline-float-divide-min-latency
9528 @opindex minline-float-divide-min-latency
9529 Generate code for inline divides of floating point values
9530 using the minimum latency algorithm.
9532 @item -minline-float-divide-max-throughput
9533 @opindex minline-float-divide-max-throughput
9534 Generate code for inline divides of floating point values
9535 using the maximum throughput algorithm.
9537 @item -minline-int-divide-min-latency
9538 @opindex minline-int-divide-min-latency
9539 Generate code for inline divides of integer values
9540 using the minimum latency algorithm.
9542 @item -minline-int-divide-max-throughput
9543 @opindex minline-int-divide-max-throughput
9544 Generate code for inline divides of integer values
9545 using the maximum throughput algorithm.
9547 @item -minline-sqrt-min-latency
9548 @opindex minline-sqrt-min-latency
9549 Generate code for inline square roots
9550 using the minimum latency algorithm.
9552 @item -minline-sqrt-max-throughput
9553 @opindex minline-sqrt-max-throughput
9554 Generate code for inline square roots
9555 using the maximum throughput algorithm.
9557 @item -mno-dwarf2-asm
9559 @opindex mno-dwarf2-asm
9560 @opindex mdwarf2-asm
9561 Don't (or do) generate assembler code for the DWARF2 line number debugging
9562 info. This may be useful when not using the GNU assembler.
9564 @item -mearly-stop-bits
9565 @itemx -mno-early-stop-bits
9566 @opindex mearly-stop-bits
9567 @opindex mno-early-stop-bits
9568 Allow stop bits to be placed earlier than immediately preceding the
9569 instruction that triggered the stop bit. This can improve instruction
9570 scheduling, but does not always do so.
9572 @item -mfixed-range=@var{register-range}
9573 @opindex mfixed-range
9574 Generate code treating the given register range as fixed registers.
9575 A fixed register is one that the register allocator can not use. This is
9576 useful when compiling kernel code. A register range is specified as
9577 two registers separated by a dash. Multiple register ranges can be
9578 specified separated by a comma.
9580 @item -mtls-size=@var{tls-size}
9582 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9585 @item -mtune=@var{cpu-type}
9587 Tune the instruction scheduling for a particular CPU, Valid values are
9588 itanium, itanium1, merced, itanium2, and mckinley.
9594 Add support for multithreading using the POSIX threads library. This
9595 option sets flags for both the preprocessor and linker. It does
9596 not affect the thread safety of object code produced by the compiler or
9597 that of libraries supplied with it. These are HP-UX specific flags.
9603 Generate code for a 32-bit or 64-bit environment.
9604 The 32-bit environment sets int, long and pointer to 32 bits.
9605 The 64-bit environment sets int to 32 bits and long and pointer
9606 to 64 bits. These are HP-UX specific flags.
9611 @subsection M32C Options
9612 @cindex M32C options
9615 @item -mcpu=@var{name}
9617 Select the CPU for which code is generated. @var{name} may be one of
9618 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9619 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9624 Specifies that the program will be run on the simulator. This causes
9625 an alternate runtime library to be linked in which supports, for
9626 example, file I/O. You must not use this option when generating
9627 programs that will run on real hardware; you must provide your own
9628 runtime library for whatever I/O functions are needed.
9630 @item -memregs=@var{number}
9632 Specifies the number of memory-based pseudo-registers GCC will use
9633 during code generation. These pseudo-registers will be used like real
9634 registers, so there is a tradeoff between GCC's ability to fit the
9635 code into available registers, and the performance penalty of using
9636 memory instead of registers. Note that all modules in a program must
9637 be compiled with the same value for this option. Because of that, you
9638 must not use this option with the default runtime libraries gcc
9643 @node M32R/D Options
9644 @subsection M32R/D Options
9645 @cindex M32R/D options
9647 These @option{-m} options are defined for Renesas M32R/D architectures:
9652 Generate code for the M32R/2@.
9656 Generate code for the M32R/X@.
9660 Generate code for the M32R@. This is the default.
9663 @opindex mmodel=small
9664 Assume all objects live in the lower 16MB of memory (so that their addresses
9665 can be loaded with the @code{ld24} instruction), and assume all subroutines
9666 are reachable with the @code{bl} instruction.
9667 This is the default.
9669 The addressability of a particular object can be set with the
9670 @code{model} attribute.
9672 @item -mmodel=medium
9673 @opindex mmodel=medium
9674 Assume objects may be anywhere in the 32-bit address space (the compiler
9675 will generate @code{seth/add3} instructions to load their addresses), and
9676 assume all subroutines are reachable with the @code{bl} instruction.
9679 @opindex mmodel=large
9680 Assume objects may be anywhere in the 32-bit address space (the compiler
9681 will generate @code{seth/add3} instructions to load their addresses), and
9682 assume subroutines may not be reachable with the @code{bl} instruction
9683 (the compiler will generate the much slower @code{seth/add3/jl}
9684 instruction sequence).
9687 @opindex msdata=none
9688 Disable use of the small data area. Variables will be put into
9689 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9690 @code{section} attribute has been specified).
9691 This is the default.
9693 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9694 Objects may be explicitly put in the small data area with the
9695 @code{section} attribute using one of these sections.
9698 @opindex msdata=sdata
9699 Put small global and static data in the small data area, but do not
9700 generate special code to reference them.
9704 Put small global and static data in the small data area, and generate
9705 special instructions to reference them.
9709 @cindex smaller data references
9710 Put global and static objects less than or equal to @var{num} bytes
9711 into the small data or bss sections instead of the normal data or bss
9712 sections. The default value of @var{num} is 8.
9713 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9714 for this option to have any effect.
9716 All modules should be compiled with the same @option{-G @var{num}} value.
9717 Compiling with different values of @var{num} may or may not work; if it
9718 doesn't the linker will give an error message---incorrect code will not be
9723 Makes the M32R specific code in the compiler display some statistics
9724 that might help in debugging programs.
9727 @opindex malign-loops
9728 Align all loops to a 32-byte boundary.
9730 @item -mno-align-loops
9731 @opindex mno-align-loops
9732 Do not enforce a 32-byte alignment for loops. This is the default.
9734 @item -missue-rate=@var{number}
9735 @opindex missue-rate=@var{number}
9736 Issue @var{number} instructions per cycle. @var{number} can only be 1
9739 @item -mbranch-cost=@var{number}
9740 @opindex mbranch-cost=@var{number}
9741 @var{number} can only be 1 or 2. If it is 1 then branches will be
9742 preferred over conditional code, if it is 2, then the opposite will
9745 @item -mflush-trap=@var{number}
9746 @opindex mflush-trap=@var{number}
9747 Specifies the trap number to use to flush the cache. The default is
9748 12. Valid numbers are between 0 and 15 inclusive.
9750 @item -mno-flush-trap
9751 @opindex mno-flush-trap
9752 Specifies that the cache cannot be flushed by using a trap.
9754 @item -mflush-func=@var{name}
9755 @opindex mflush-func=@var{name}
9756 Specifies the name of the operating system function to call to flush
9757 the cache. The default is @emph{_flush_cache}, but a function call
9758 will only be used if a trap is not available.
9760 @item -mno-flush-func
9761 @opindex mno-flush-func
9762 Indicates that there is no OS function for flushing the cache.
9766 @node M680x0 Options
9767 @subsection M680x0 Options
9768 @cindex M680x0 options
9770 These are the @samp{-m} options defined for the 68000 series. The default
9771 values for these options depends on which style of 68000 was selected when
9772 the compiler was configured; the defaults for the most common choices are
9780 Generate output for a 68000. This is the default
9781 when the compiler is configured for 68000-based systems.
9783 Use this option for microcontrollers with a 68000 or EC000 core,
9784 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9790 Generate output for a 68020. This is the default
9791 when the compiler is configured for 68020-based systems.
9795 Generate output containing 68881 instructions for floating point.
9796 This is the default for most 68020 systems unless @option{--nfp} was
9797 specified when the compiler was configured.
9801 Generate output for a 68030. This is the default when the compiler is
9802 configured for 68030-based systems.
9806 Generate output for a 68040. This is the default when the compiler is
9807 configured for 68040-based systems.
9809 This option inhibits the use of 68881/68882 instructions that have to be
9810 emulated by software on the 68040. Use this option if your 68040 does not
9811 have code to emulate those instructions.
9815 Generate output for a 68060. This is the default when the compiler is
9816 configured for 68060-based systems.
9818 This option inhibits the use of 68020 and 68881/68882 instructions that
9819 have to be emulated by software on the 68060. Use this option if your 68060
9820 does not have code to emulate those instructions.
9824 Generate output for a CPU32. This is the default
9825 when the compiler is configured for CPU32-based systems.
9827 Use this option for microcontrollers with a
9828 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9829 68336, 68340, 68341, 68349 and 68360.
9833 Generate output for a 520X ``coldfire'' family cpu. This is the default
9834 when the compiler is configured for 520X-based systems.
9836 Use this option for microcontroller with a 5200 core, including
9837 the MCF5202, MCF5203, MCF5204 and MCF5202.
9842 Generate output for a 68040, without using any of the new instructions.
9843 This results in code which can run relatively efficiently on either a
9844 68020/68881 or a 68030 or a 68040. The generated code does use the
9845 68881 instructions that are emulated on the 68040.
9849 Generate output for a 68060, without using any of the new instructions.
9850 This results in code which can run relatively efficiently on either a
9851 68020/68881 or a 68030 or a 68040. The generated code does use the
9852 68881 instructions that are emulated on the 68060.
9855 @opindex msoft-float
9856 Generate output containing library calls for floating point.
9857 @strong{Warning:} the requisite libraries are not available for all m68k
9858 targets. Normally the facilities of the machine's usual C compiler are
9859 used, but this can't be done directly in cross-compilation. You must
9860 make your own arrangements to provide suitable library functions for
9861 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9862 @samp{m68k-*-coff} do provide software floating point support.
9866 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9867 Additionally, parameters passed on the stack are also aligned to a
9868 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9871 @opindex mnobitfield
9872 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9873 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9877 Do use the bit-field instructions. The @option{-m68020} option implies
9878 @option{-mbitfield}. This is the default if you use a configuration
9879 designed for a 68020.
9883 Use a different function-calling convention, in which functions
9884 that take a fixed number of arguments return with the @code{rtd}
9885 instruction, which pops their arguments while returning. This
9886 saves one instruction in the caller since there is no need to pop
9887 the arguments there.
9889 This calling convention is incompatible with the one normally
9890 used on Unix, so you cannot use it if you need to call libraries
9891 compiled with the Unix compiler.
9893 Also, you must provide function prototypes for all functions that
9894 take variable numbers of arguments (including @code{printf});
9895 otherwise incorrect code will be generated for calls to those
9898 In addition, seriously incorrect code will result if you call a
9899 function with too many arguments. (Normally, extra arguments are
9900 harmlessly ignored.)
9902 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9903 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9906 @itemx -mno-align-int
9908 @opindex mno-align-int
9909 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9910 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9911 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9912 Aligning variables on 32-bit boundaries produces code that runs somewhat
9913 faster on processors with 32-bit busses at the expense of more memory.
9915 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9916 align structures containing the above types differently than
9917 most published application binary interface specifications for the m68k.
9921 Use the pc-relative addressing mode of the 68000 directly, instead of
9922 using a global offset table. At present, this option implies @option{-fpic},
9923 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9924 not presently supported with @option{-mpcrel}, though this could be supported for
9925 68020 and higher processors.
9927 @item -mno-strict-align
9928 @itemx -mstrict-align
9929 @opindex mno-strict-align
9930 @opindex mstrict-align
9931 Do not (do) assume that unaligned memory references will be handled by
9935 Generate code that allows the data segment to be located in a different
9936 area of memory from the text segment. This allows for execute in place in
9937 an environment without virtual memory management. This option implies
9941 Generate code that assumes that the data segment follows the text segment.
9942 This is the default.
9944 @item -mid-shared-library
9945 Generate code that supports shared libraries via the library ID method.
9946 This allows for execute in place and shared libraries in an environment
9947 without virtual memory management. This option implies @option{-fPIC}.
9949 @item -mno-id-shared-library
9950 Generate code that doesn't assume ID based shared libraries are being used.
9951 This is the default.
9953 @item -mshared-library-id=n
9954 Specified the identification number of the ID based shared library being
9955 compiled. Specifying a value of 0 will generate more compact code, specifying
9956 other values will force the allocation of that number to the current
9957 library but is no more space or time efficient than omitting this option.
9961 @node M68hc1x Options
9962 @subsection M68hc1x Options
9963 @cindex M68hc1x options
9965 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9966 microcontrollers. The default values for these options depends on
9967 which style of microcontroller was selected when the compiler was configured;
9968 the defaults for the most common choices are given below.
9975 Generate output for a 68HC11. This is the default
9976 when the compiler is configured for 68HC11-based systems.
9982 Generate output for a 68HC12. This is the default
9983 when the compiler is configured for 68HC12-based systems.
9989 Generate output for a 68HCS12.
9992 @opindex mauto-incdec
9993 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10000 Enable the use of 68HC12 min and max instructions.
10003 @itemx -mno-long-calls
10004 @opindex mlong-calls
10005 @opindex mno-long-calls
10006 Treat all calls as being far away (near). If calls are assumed to be
10007 far away, the compiler will use the @code{call} instruction to
10008 call a function and the @code{rtc} instruction for returning.
10012 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10014 @item -msoft-reg-count=@var{count}
10015 @opindex msoft-reg-count
10016 Specify the number of pseudo-soft registers which are used for the
10017 code generation. The maximum number is 32. Using more pseudo-soft
10018 register may or may not result in better code depending on the program.
10019 The default is 4 for 68HC11 and 2 for 68HC12.
10023 @node MCore Options
10024 @subsection MCore Options
10025 @cindex MCore options
10027 These are the @samp{-m} options defined for the Motorola M*Core
10033 @itemx -mno-hardlit
10035 @opindex mno-hardlit
10036 Inline constants into the code stream if it can be done in two
10037 instructions or less.
10043 Use the divide instruction. (Enabled by default).
10045 @item -mrelax-immediate
10046 @itemx -mno-relax-immediate
10047 @opindex mrelax-immediate
10048 @opindex mno-relax-immediate
10049 Allow arbitrary sized immediates in bit operations.
10051 @item -mwide-bitfields
10052 @itemx -mno-wide-bitfields
10053 @opindex mwide-bitfields
10054 @opindex mno-wide-bitfields
10055 Always treat bit-fields as int-sized.
10057 @item -m4byte-functions
10058 @itemx -mno-4byte-functions
10059 @opindex m4byte-functions
10060 @opindex mno-4byte-functions
10061 Force all functions to be aligned to a four byte boundary.
10063 @item -mcallgraph-data
10064 @itemx -mno-callgraph-data
10065 @opindex mcallgraph-data
10066 @opindex mno-callgraph-data
10067 Emit callgraph information.
10070 @itemx -mno-slow-bytes
10071 @opindex mslow-bytes
10072 @opindex mno-slow-bytes
10073 Prefer word access when reading byte quantities.
10075 @item -mlittle-endian
10076 @itemx -mbig-endian
10077 @opindex mlittle-endian
10078 @opindex mbig-endian
10079 Generate code for a little endian target.
10085 Generate code for the 210 processor.
10089 @subsection MIPS Options
10090 @cindex MIPS options
10096 Generate big-endian code.
10100 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10103 @item -march=@var{arch}
10105 Generate code that will run on @var{arch}, which can be the name of a
10106 generic MIPS ISA, or the name of a particular processor.
10108 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10109 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10110 The processor names are:
10111 @samp{4kc}, @samp{4km}, @samp{4kp},
10112 @samp{5kc}, @samp{5kf},
10114 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10117 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10118 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10119 @samp{rm7000}, @samp{rm9000},
10122 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10123 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10124 The special value @samp{from-abi} selects the
10125 most compatible architecture for the selected ABI (that is,
10126 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10128 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10129 (for example, @samp{-march=r2k}). Prefixes are optional, and
10130 @samp{vr} may be written @samp{r}.
10132 GCC defines two macros based on the value of this option. The first
10133 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10134 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10135 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10136 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10137 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10139 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10140 above. In other words, it will have the full prefix and will not
10141 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10142 the macro names the resolved architecture (either @samp{"mips1"} or
10143 @samp{"mips3"}). It names the default architecture when no
10144 @option{-march} option is given.
10146 @item -mtune=@var{arch}
10148 Optimize for @var{arch}. Among other things, this option controls
10149 the way instructions are scheduled, and the perceived cost of arithmetic
10150 operations. The list of @var{arch} values is the same as for
10153 When this option is not used, GCC will optimize for the processor
10154 specified by @option{-march}. By using @option{-march} and
10155 @option{-mtune} together, it is possible to generate code that will
10156 run on a family of processors, but optimize the code for one
10157 particular member of that family.
10159 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10160 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10161 @samp{-march} ones described above.
10165 Equivalent to @samp{-march=mips1}.
10169 Equivalent to @samp{-march=mips2}.
10173 Equivalent to @samp{-march=mips3}.
10177 Equivalent to @samp{-march=mips4}.
10181 Equivalent to @samp{-march=mips32}.
10185 Equivalent to @samp{-march=mips32r2}.
10189 Equivalent to @samp{-march=mips64}.
10194 @opindex mno-mips16
10195 Generate (do not generate) MIPS16 code. If GCC is targetting a
10196 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10208 Generate code for the given ABI@.
10210 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10211 generates 64-bit code when you select a 64-bit architecture, but you
10212 can use @option{-mgp32} to get 32-bit code instead.
10214 For information about the O64 ABI, see
10215 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10218 @itemx -mno-abicalls
10220 @opindex mno-abicalls
10221 Generate (do not generate) SVR4-style position-independent code.
10222 @option{-mabicalls} is the default for SVR4-based systems.
10228 Lift (do not lift) the usual restrictions on the size of the global
10231 GCC normally uses a single instruction to load values from the GOT@.
10232 While this is relatively efficient, it will only work if the GOT
10233 is smaller than about 64k. Anything larger will cause the linker
10234 to report an error such as:
10236 @cindex relocation truncated to fit (MIPS)
10238 relocation truncated to fit: R_MIPS_GOT16 foobar
10241 If this happens, you should recompile your code with @option{-mxgot}.
10242 It should then work with very large GOTs, although it will also be
10243 less efficient, since it will take three instructions to fetch the
10244 value of a global symbol.
10246 Note that some linkers can create multiple GOTs. If you have such a
10247 linker, you should only need to use @option{-mxgot} when a single object
10248 file accesses more than 64k's worth of GOT entries. Very few do.
10250 These options have no effect unless GCC is generating position
10255 Assume that general-purpose registers are 32 bits wide.
10259 Assume that general-purpose registers are 64 bits wide.
10263 Assume that floating-point registers are 32 bits wide.
10267 Assume that floating-point registers are 64 bits wide.
10270 @opindex mhard-float
10271 Use floating-point coprocessor instructions.
10274 @opindex msoft-float
10275 Do not use floating-point coprocessor instructions. Implement
10276 floating-point calculations using library calls instead.
10278 @item -msingle-float
10279 @opindex msingle-float
10280 Assume that the floating-point coprocessor only supports single-precision
10283 @itemx -mdouble-float
10284 @opindex mdouble-float
10285 Assume that the floating-point coprocessor supports double-precision
10286 operations. This is the default.
10292 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10294 @itemx -mpaired-single
10295 @itemx -mno-paired-single
10296 @opindex mpaired-single
10297 @opindex mno-paired-single
10298 Use (do not use) paired-single floating-point instructions.
10299 @xref{MIPS Paired-Single Support}. This option can only be used
10300 when generating 64-bit code and requires hardware floating-point
10301 support to be enabled.
10306 @opindex mno-mips3d
10307 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10308 The option @option{-mips3d} implies @option{-mpaired-single}.
10312 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10313 an explanation of the default and the way that the pointer size is
10318 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10320 The default size of @code{int}s, @code{long}s and pointers depends on
10321 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10322 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10323 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10324 or the same size as integer registers, whichever is smaller.
10330 Assume (do not assume) that all symbols have 32-bit values, regardless
10331 of the selected ABI@. This option is useful in combination with
10332 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10333 to generate shorter and faster references to symbolic addresses.
10337 @cindex smaller data references (MIPS)
10338 @cindex gp-relative references (MIPS)
10339 Put global and static items less than or equal to @var{num} bytes into
10340 the small data or bss section instead of the normal data or bss section.
10341 This allows the data to be accessed using a single instruction.
10343 All modules should be compiled with the same @option{-G @var{num}}
10346 @item -membedded-data
10347 @itemx -mno-embedded-data
10348 @opindex membedded-data
10349 @opindex mno-embedded-data
10350 Allocate variables to the read-only data section first if possible, then
10351 next in the small data section if possible, otherwise in data. This gives
10352 slightly slower code than the default, but reduces the amount of RAM required
10353 when executing, and thus may be preferred for some embedded systems.
10355 @item -muninit-const-in-rodata
10356 @itemx -mno-uninit-const-in-rodata
10357 @opindex muninit-const-in-rodata
10358 @opindex mno-uninit-const-in-rodata
10359 Put uninitialized @code{const} variables in the read-only data section.
10360 This option is only meaningful in conjunction with @option{-membedded-data}.
10362 @item -msplit-addresses
10363 @itemx -mno-split-addresses
10364 @opindex msplit-addresses
10365 @opindex mno-split-addresses
10366 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10367 relocation operators. This option has been superseded by
10368 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10370 @item -mexplicit-relocs
10371 @itemx -mno-explicit-relocs
10372 @opindex mexplicit-relocs
10373 @opindex mno-explicit-relocs
10374 Use (do not use) assembler relocation operators when dealing with symbolic
10375 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10376 is to use assembler macros instead.
10378 @option{-mexplicit-relocs} is the default if GCC was configured
10379 to use an assembler that supports relocation operators.
10381 @item -mcheck-zero-division
10382 @itemx -mno-check-zero-division
10383 @opindex mcheck-zero-division
10384 @opindex mno-check-zero-division
10385 Trap (do not trap) on integer division by zero. The default is
10386 @option{-mcheck-zero-division}.
10388 @item -mdivide-traps
10389 @itemx -mdivide-breaks
10390 @opindex mdivide-traps
10391 @opindex mdivide-breaks
10392 MIPS systems check for division by zero by generating either a
10393 conditional trap or a break instruction. Using traps results in
10394 smaller code, but is only supported on MIPS II and later. Also, some
10395 versions of the Linux kernel have a bug that prevents trap from
10396 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10397 allow conditional traps on architectures that support them and
10398 @option{-mdivide-breaks} to force the use of breaks.
10400 The default is usually @option{-mdivide-traps}, but this can be
10401 overridden at configure time using @option{--with-divide=breaks}.
10402 Divide-by-zero checks can be completely disabled using
10403 @option{-mno-check-zero-division}.
10408 @opindex mno-memcpy
10409 Force (do not force) the use of @code{memcpy()} for non-trivial block
10410 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10411 most constant-sized copies.
10414 @itemx -mno-long-calls
10415 @opindex mlong-calls
10416 @opindex mno-long-calls
10417 Disable (do not disable) use of the @code{jal} instruction. Calling
10418 functions using @code{jal} is more efficient but requires the caller
10419 and callee to be in the same 256 megabyte segment.
10421 This option has no effect on abicalls code. The default is
10422 @option{-mno-long-calls}.
10428 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10429 instructions, as provided by the R4650 ISA@.
10432 @itemx -mno-fused-madd
10433 @opindex mfused-madd
10434 @opindex mno-fused-madd
10435 Enable (disable) use of the floating point multiply-accumulate
10436 instructions, when they are available. The default is
10437 @option{-mfused-madd}.
10439 When multiply-accumulate instructions are used, the intermediate
10440 product is calculated to infinite precision and is not subject to
10441 the FCSR Flush to Zero bit. This may be undesirable in some
10446 Tell the MIPS assembler to not run its preprocessor over user
10447 assembler files (with a @samp{.s} suffix) when assembling them.
10450 @itemx -mno-fix-r4000
10451 @opindex mfix-r4000
10452 @opindex mno-fix-r4000
10453 Work around certain R4000 CPU errata:
10456 A double-word or a variable shift may give an incorrect result if executed
10457 immediately after starting an integer division.
10459 A double-word or a variable shift may give an incorrect result if executed
10460 while an integer multiplication is in progress.
10462 An integer division may give an incorrect result if started in a delay slot
10463 of a taken branch or a jump.
10467 @itemx -mno-fix-r4400
10468 @opindex mfix-r4400
10469 @opindex mno-fix-r4400
10470 Work around certain R4400 CPU errata:
10473 A double-word or a variable shift may give an incorrect result if executed
10474 immediately after starting an integer division.
10478 @itemx -mno-fix-vr4120
10479 @opindex mfix-vr4120
10480 Work around certain VR4120 errata:
10483 @code{dmultu} does not always produce the correct result.
10485 @code{div} and @code{ddiv} do not always produce the correct result if one
10486 of the operands is negative.
10488 The workarounds for the division errata rely on special functions in
10489 @file{libgcc.a}. At present, these functions are only provided by
10490 the @code{mips64vr*-elf} configurations.
10492 Other VR4120 errata require a nop to be inserted between certain pairs of
10493 instructions. These errata are handled by the assembler, not by GCC itself.
10496 @opindex mfix-vr4130
10497 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10498 workarounds are implemented by the assembler rather than by GCC,
10499 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10500 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10501 instructions are available instead.
10504 @itemx -mno-fix-sb1
10506 Work around certain SB-1 CPU core errata.
10507 (This flag currently works around the SB-1 revision 2
10508 ``F1'' and ``F2'' floating point errata.)
10510 @item -mflush-func=@var{func}
10511 @itemx -mno-flush-func
10512 @opindex mflush-func
10513 Specifies the function to call to flush the I and D caches, or to not
10514 call any such function. If called, the function must take the same
10515 arguments as the common @code{_flush_func()}, that is, the address of the
10516 memory range for which the cache is being flushed, the size of the
10517 memory range, and the number 3 (to flush both caches). The default
10518 depends on the target GCC was configured for, but commonly is either
10519 @samp{_flush_func} or @samp{__cpu_flush}.
10521 @item -mbranch-likely
10522 @itemx -mno-branch-likely
10523 @opindex mbranch-likely
10524 @opindex mno-branch-likely
10525 Enable or disable use of Branch Likely instructions, regardless of the
10526 default for the selected architecture. By default, Branch Likely
10527 instructions may be generated if they are supported by the selected
10528 architecture. An exception is for the MIPS32 and MIPS64 architectures
10529 and processors which implement those architectures; for those, Branch
10530 Likely instructions will not be generated by default because the MIPS32
10531 and MIPS64 architectures specifically deprecate their use.
10533 @item -mfp-exceptions
10534 @itemx -mno-fp-exceptions
10535 @opindex mfp-exceptions
10536 Specifies whether FP exceptions are enabled. This affects how we schedule
10537 FP instructions for some processors. The default is that FP exceptions are
10540 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10541 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10544 @item -mvr4130-align
10545 @itemx -mno-vr4130-align
10546 @opindex mvr4130-align
10547 The VR4130 pipeline is two-way superscalar, but can only issue two
10548 instructions together if the first one is 8-byte aligned. When this
10549 option is enabled, GCC will align pairs of instructions that it
10550 thinks should execute in parallel.
10552 This option only has an effect when optimizing for the VR4130.
10553 It normally makes code faster, but at the expense of making it bigger.
10554 It is enabled by default at optimization level @option{-O3}.
10558 @subsection MMIX Options
10559 @cindex MMIX Options
10561 These options are defined for the MMIX:
10565 @itemx -mno-libfuncs
10567 @opindex mno-libfuncs
10568 Specify that intrinsic library functions are being compiled, passing all
10569 values in registers, no matter the size.
10572 @itemx -mno-epsilon
10574 @opindex mno-epsilon
10575 Generate floating-point comparison instructions that compare with respect
10576 to the @code{rE} epsilon register.
10578 @item -mabi=mmixware
10580 @opindex mabi-mmixware
10582 Generate code that passes function parameters and return values that (in
10583 the called function) are seen as registers @code{$0} and up, as opposed to
10584 the GNU ABI which uses global registers @code{$231} and up.
10586 @item -mzero-extend
10587 @itemx -mno-zero-extend
10588 @opindex mzero-extend
10589 @opindex mno-zero-extend
10590 When reading data from memory in sizes shorter than 64 bits, use (do not
10591 use) zero-extending load instructions by default, rather than
10592 sign-extending ones.
10595 @itemx -mno-knuthdiv
10597 @opindex mno-knuthdiv
10598 Make the result of a division yielding a remainder have the same sign as
10599 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10600 remainder follows the sign of the dividend. Both methods are
10601 arithmetically valid, the latter being almost exclusively used.
10603 @item -mtoplevel-symbols
10604 @itemx -mno-toplevel-symbols
10605 @opindex mtoplevel-symbols
10606 @opindex mno-toplevel-symbols
10607 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10608 code can be used with the @code{PREFIX} assembly directive.
10612 Generate an executable in the ELF format, rather than the default
10613 @samp{mmo} format used by the @command{mmix} simulator.
10615 @item -mbranch-predict
10616 @itemx -mno-branch-predict
10617 @opindex mbranch-predict
10618 @opindex mno-branch-predict
10619 Use (do not use) the probable-branch instructions, when static branch
10620 prediction indicates a probable branch.
10622 @item -mbase-addresses
10623 @itemx -mno-base-addresses
10624 @opindex mbase-addresses
10625 @opindex mno-base-addresses
10626 Generate (do not generate) code that uses @emph{base addresses}. Using a
10627 base address automatically generates a request (handled by the assembler
10628 and the linker) for a constant to be set up in a global register. The
10629 register is used for one or more base address requests within the range 0
10630 to 255 from the value held in the register. The generally leads to short
10631 and fast code, but the number of different data items that can be
10632 addressed is limited. This means that a program that uses lots of static
10633 data may require @option{-mno-base-addresses}.
10635 @item -msingle-exit
10636 @itemx -mno-single-exit
10637 @opindex msingle-exit
10638 @opindex mno-single-exit
10639 Force (do not force) generated code to have a single exit point in each
10643 @node MN10300 Options
10644 @subsection MN10300 Options
10645 @cindex MN10300 options
10647 These @option{-m} options are defined for Matsushita MN10300 architectures:
10652 Generate code to avoid bugs in the multiply instructions for the MN10300
10653 processors. This is the default.
10655 @item -mno-mult-bug
10656 @opindex mno-mult-bug
10657 Do not generate code to avoid bugs in the multiply instructions for the
10658 MN10300 processors.
10662 Generate code which uses features specific to the AM33 processor.
10666 Do not generate code which uses features specific to the AM33 processor. This
10669 @item -mreturn-pointer-on-d0
10670 @opindex mreturn-pointer-on-d0
10671 When generating a function which returns a pointer, return the pointer
10672 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
10673 only in a0, and attempts to call such functions without a prototype
10674 would result in errors. Note that this option is on by default; use
10675 @option{-mno-return-pointer-on-d0} to disable it.
10679 Do not link in the C run-time initialization object file.
10683 Indicate to the linker that it should perform a relaxation optimization pass
10684 to shorten branches, calls and absolute memory addresses. This option only
10685 has an effect when used on the command line for the final link step.
10687 This option makes symbolic debugging impossible.
10691 @subsection MT Options
10694 These @option{-m} options are defined for Morpho MT architectures:
10698 @item -march=@var{cpu-type}
10700 Generate code that will run on @var{cpu-type}, which is the name of a system
10701 representing a certain processor type. Possible values for
10702 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
10703 @samp{ms1-16-003} and @samp{ms2}.
10705 When this option is not used, the default is @option{-march=ms1-16-002}.
10709 Use byte loads and stores when generating code.
10713 Do not use byte loads and stores when generating code.
10717 Use simulator runtime
10721 Do not link in the C run-time initialization object file
10722 @file{crti.o}. Other run-time initialization and termination files
10723 such as @file{startup.o} and @file{exit.o} are still included on the
10724 linker command line.
10728 @node PDP-11 Options
10729 @subsection PDP-11 Options
10730 @cindex PDP-11 Options
10732 These options are defined for the PDP-11:
10737 Use hardware FPP floating point. This is the default. (FIS floating
10738 point on the PDP-11/40 is not supported.)
10741 @opindex msoft-float
10742 Do not use hardware floating point.
10746 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10750 Return floating-point results in memory. This is the default.
10754 Generate code for a PDP-11/40.
10758 Generate code for a PDP-11/45. This is the default.
10762 Generate code for a PDP-11/10.
10764 @item -mbcopy-builtin
10765 @opindex bcopy-builtin
10766 Use inline @code{movmemhi} patterns for copying memory. This is the
10771 Do not use inline @code{movmemhi} patterns for copying memory.
10777 Use 16-bit @code{int}. This is the default.
10783 Use 32-bit @code{int}.
10786 @itemx -mno-float32
10788 @opindex mno-float32
10789 Use 64-bit @code{float}. This is the default.
10792 @itemx -mno-float64
10794 @opindex mno-float64
10795 Use 32-bit @code{float}.
10799 Use @code{abshi2} pattern. This is the default.
10803 Do not use @code{abshi2} pattern.
10805 @item -mbranch-expensive
10806 @opindex mbranch-expensive
10807 Pretend that branches are expensive. This is for experimenting with
10808 code generation only.
10810 @item -mbranch-cheap
10811 @opindex mbranch-cheap
10812 Do not pretend that branches are expensive. This is the default.
10816 Generate code for a system with split I&D@.
10820 Generate code for a system without split I&D@. This is the default.
10824 Use Unix assembler syntax. This is the default when configured for
10825 @samp{pdp11-*-bsd}.
10829 Use DEC assembler syntax. This is the default when configured for any
10830 PDP-11 target other than @samp{pdp11-*-bsd}.
10833 @node PowerPC Options
10834 @subsection PowerPC Options
10835 @cindex PowerPC options
10837 These are listed under @xref{RS/6000 and PowerPC Options}.
10839 @node RS/6000 and PowerPC Options
10840 @subsection IBM RS/6000 and PowerPC Options
10841 @cindex RS/6000 and PowerPC Options
10842 @cindex IBM RS/6000 and PowerPC Options
10844 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10851 @itemx -mno-powerpc
10852 @itemx -mpowerpc-gpopt
10853 @itemx -mno-powerpc-gpopt
10854 @itemx -mpowerpc-gfxopt
10855 @itemx -mno-powerpc-gfxopt
10857 @itemx -mno-powerpc64
10861 @itemx -mno-popcntb
10867 @opindex mno-power2
10869 @opindex mno-powerpc
10870 @opindex mpowerpc-gpopt
10871 @opindex mno-powerpc-gpopt
10872 @opindex mpowerpc-gfxopt
10873 @opindex mno-powerpc-gfxopt
10874 @opindex mpowerpc64
10875 @opindex mno-powerpc64
10879 @opindex mno-popcntb
10882 GCC supports two related instruction set architectures for the
10883 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10884 instructions supported by the @samp{rios} chip set used in the original
10885 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10886 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
10887 the IBM 4xx, 6xx, and follow-on microprocessors.
10889 Neither architecture is a subset of the other. However there is a
10890 large common subset of instructions supported by both. An MQ
10891 register is included in processors supporting the POWER architecture.
10893 You use these options to specify which instructions are available on the
10894 processor you are using. The default value of these options is
10895 determined when configuring GCC@. Specifying the
10896 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10897 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10898 rather than the options listed above.
10900 The @option{-mpower} option allows GCC to generate instructions that
10901 are found only in the POWER architecture and to use the MQ register.
10902 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10903 to generate instructions that are present in the POWER2 architecture but
10904 not the original POWER architecture.
10906 The @option{-mpowerpc} option allows GCC to generate instructions that
10907 are found only in the 32-bit subset of the PowerPC architecture.
10908 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10909 GCC to use the optional PowerPC architecture instructions in the
10910 General Purpose group, including floating-point square root. Specifying
10911 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10912 use the optional PowerPC architecture instructions in the Graphics
10913 group, including floating-point select.
10915 The @option{-mmfcrf} option allows GCC to generate the move from
10916 condition register field instruction implemented on the POWER4
10917 processor and other processors that support the PowerPC V2.01
10919 The @option{-mpopcntb} option allows GCC to generate the popcount and
10920 double precision FP reciprocal estimate instruction implemented on the
10921 POWER5 processor and other processors that support the PowerPC V2.02
10923 The @option{-mfprnd} option allows GCC to generate the FP round to
10924 integer instructions implemented on the POWER5+ processor and other
10925 processors that support the PowerPC V2.03 architecture.
10927 The @option{-mpowerpc64} option allows GCC to generate the additional
10928 64-bit instructions that are found in the full PowerPC64 architecture
10929 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10930 @option{-mno-powerpc64}.
10932 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10933 will use only the instructions in the common subset of both
10934 architectures plus some special AIX common-mode calls, and will not use
10935 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10936 permits GCC to use any instruction from either architecture and to
10937 allow use of the MQ register; specify this for the Motorola MPC601.
10939 @item -mnew-mnemonics
10940 @itemx -mold-mnemonics
10941 @opindex mnew-mnemonics
10942 @opindex mold-mnemonics
10943 Select which mnemonics to use in the generated assembler code. With
10944 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10945 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10946 assembler mnemonics defined for the POWER architecture. Instructions
10947 defined in only one architecture have only one mnemonic; GCC uses that
10948 mnemonic irrespective of which of these options is specified.
10950 GCC defaults to the mnemonics appropriate for the architecture in
10951 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10952 value of these option. Unless you are building a cross-compiler, you
10953 should normally not specify either @option{-mnew-mnemonics} or
10954 @option{-mold-mnemonics}, but should instead accept the default.
10956 @item -mcpu=@var{cpu_type}
10958 Set architecture type, register usage, choice of mnemonics, and
10959 instruction scheduling parameters for machine type @var{cpu_type}.
10960 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10961 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10962 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10963 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10964 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10965 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
10966 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10967 @samp{power4}, @samp{power5}, @samp{power5+},
10968 @samp{common}, @samp{powerpc}, @samp{powerpc64},
10969 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10971 @option{-mcpu=common} selects a completely generic processor. Code
10972 generated under this option will run on any POWER or PowerPC processor.
10973 GCC will use only the instructions in the common subset of both
10974 architectures, and will not use the MQ register. GCC assumes a generic
10975 processor model for scheduling purposes.
10977 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10978 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10979 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10980 types, with an appropriate, generic processor model assumed for
10981 scheduling purposes.
10983 The other options specify a specific processor. Code generated under
10984 those options will run best on that processor, and may not run at all on
10987 The @option{-mcpu} options automatically enable or disable the
10988 following options: @option{-maltivec}, @option{-mfprnd},
10989 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
10990 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
10991 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
10992 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw}.
10993 The particular options
10994 set for any particular CPU will vary between compiler versions,
10995 depending on what setting seems to produce optimal code for that CPU;
10996 it doesn't necessarily reflect the actual hardware's capabilities. If
10997 you wish to set an individual option to a particular value, you may
10998 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
11001 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11002 not enabled or disabled by the @option{-mcpu} option at present because
11003 AIX does not have full support for these options. You may still
11004 enable or disable them individually if you're sure it'll work in your
11007 @item -mtune=@var{cpu_type}
11009 Set the instruction scheduling parameters for machine type
11010 @var{cpu_type}, but do not set the architecture type, register usage, or
11011 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11012 values for @var{cpu_type} are used for @option{-mtune} as for
11013 @option{-mcpu}. If both are specified, the code generated will use the
11014 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11015 scheduling parameters set by @option{-mtune}.
11021 Generate code to compute division as reciprocal estimate and iterative
11022 refinement, creating opportunities for increased throughput. This
11023 feature requires: optional PowerPC Graphics instruction set for single
11024 precision and FRE instruction for double precision, assuming divides
11025 cannot generate user-visible traps, and the domain values not include
11026 Infinities, denormals or zero denominator.
11029 @itemx -mno-altivec
11031 @opindex mno-altivec
11032 Generate code that uses (does not use) AltiVec instructions, and also
11033 enable the use of built-in functions that allow more direct access to
11034 the AltiVec instruction set. You may also need to set
11035 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11041 @opindex mno-vrsave
11042 Generate VRSAVE instructions when generating AltiVec code.
11046 Extend the current ABI with SPE ABI extensions. This does not change
11047 the default ABI, instead it adds the SPE ABI extensions to the current
11051 @opindex mabi=no-spe
11052 Disable Booke SPE ABI extensions for the current ABI@.
11055 @opindex msecure-plt
11056 Generate code that allows ld and ld.so to build executables and shared
11057 libraries with non-exec .plt and .got sections. This is a PowerPC
11058 32-bit SYSV ABI option.
11062 Generate code that uses a BSS .plt section that ld.so fills in, and
11063 requires .plt and .got sections that are both writable and executable.
11064 This is a PowerPC 32-bit SYSV ABI option.
11070 This switch enables or disables the generation of ISEL instructions.
11072 @item -misel=@var{yes/no}
11073 This switch has been deprecated. Use @option{-misel} and
11074 @option{-mno-isel} instead.
11080 This switch enables or disables the generation of SPE simd
11083 @item -mspe=@var{yes/no}
11084 This option has been deprecated. Use @option{-mspe} and
11085 @option{-mno-spe} instead.
11087 @item -mfloat-gprs=@var{yes/single/double/no}
11088 @itemx -mfloat-gprs
11089 @opindex mfloat-gprs
11090 This switch enables or disables the generation of floating point
11091 operations on the general purpose registers for architectures that
11094 The argument @var{yes} or @var{single} enables the use of
11095 single-precision floating point operations.
11097 The argument @var{double} enables the use of single and
11098 double-precision floating point operations.
11100 The argument @var{no} disables floating point operations on the
11101 general purpose registers.
11103 This option is currently only available on the MPC854x.
11109 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11110 targets (including GNU/Linux). The 32-bit environment sets int, long
11111 and pointer to 32 bits and generates code that runs on any PowerPC
11112 variant. The 64-bit environment sets int to 32 bits and long and
11113 pointer to 64 bits, and generates code for PowerPC64, as for
11114 @option{-mpowerpc64}.
11117 @itemx -mno-fp-in-toc
11118 @itemx -mno-sum-in-toc
11119 @itemx -mminimal-toc
11121 @opindex mno-fp-in-toc
11122 @opindex mno-sum-in-toc
11123 @opindex mminimal-toc
11124 Modify generation of the TOC (Table Of Contents), which is created for
11125 every executable file. The @option{-mfull-toc} option is selected by
11126 default. In that case, GCC will allocate at least one TOC entry for
11127 each unique non-automatic variable reference in your program. GCC
11128 will also place floating-point constants in the TOC@. However, only
11129 16,384 entries are available in the TOC@.
11131 If you receive a linker error message that saying you have overflowed
11132 the available TOC space, you can reduce the amount of TOC space used
11133 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11134 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11135 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11136 generate code to calculate the sum of an address and a constant at
11137 run-time instead of putting that sum into the TOC@. You may specify one
11138 or both of these options. Each causes GCC to produce very slightly
11139 slower and larger code at the expense of conserving TOC space.
11141 If you still run out of space in the TOC even when you specify both of
11142 these options, specify @option{-mminimal-toc} instead. This option causes
11143 GCC to make only one TOC entry for every file. When you specify this
11144 option, GCC will produce code that is slower and larger but which
11145 uses extremely little TOC space. You may wish to use this option
11146 only on files that contain less frequently executed code.
11152 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11153 @code{long} type, and the infrastructure needed to support them.
11154 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11155 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11156 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11159 @itemx -mno-xl-compat
11160 @opindex mxl-compat
11161 @opindex mno-xl-compat
11162 Produce code that conforms more closely to IBM XLC semantics when using
11163 AIX-compatible ABI. Pass floating-point arguments to prototyped
11164 functions beyond the register save area (RSA) on the stack in addition
11165 to argument FPRs. Do not assume that most significant double in 128
11166 bit long double value is properly rounded when comparing values.
11168 The AIX calling convention was extended but not initially documented to
11169 handle an obscure K&R C case of calling a function that takes the
11170 address of its arguments with fewer arguments than declared. AIX XL
11171 compilers access floating point arguments which do not fit in the
11172 RSA from the stack when a subroutine is compiled without
11173 optimization. Because always storing floating-point arguments on the
11174 stack is inefficient and rarely needed, this option is not enabled by
11175 default and only is necessary when calling subroutines compiled by AIX
11176 XL compilers without optimization.
11180 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11181 application written to use message passing with special startup code to
11182 enable the application to run. The system must have PE installed in the
11183 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11184 must be overridden with the @option{-specs=} option to specify the
11185 appropriate directory location. The Parallel Environment does not
11186 support threads, so the @option{-mpe} option and the @option{-pthread}
11187 option are incompatible.
11189 @item -malign-natural
11190 @itemx -malign-power
11191 @opindex malign-natural
11192 @opindex malign-power
11193 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11194 @option{-malign-natural} overrides the ABI-defined alignment of larger
11195 types, such as floating-point doubles, on their natural size-based boundary.
11196 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11197 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11199 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11203 @itemx -mhard-float
11204 @opindex msoft-float
11205 @opindex mhard-float
11206 Generate code that does not use (uses) the floating-point register set.
11207 Software floating point emulation is provided if you use the
11208 @option{-msoft-float} option, and pass the option to GCC when linking.
11211 @itemx -mno-multiple
11213 @opindex mno-multiple
11214 Generate code that uses (does not use) the load multiple word
11215 instructions and the store multiple word instructions. These
11216 instructions are generated by default on POWER systems, and not
11217 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11218 endian PowerPC systems, since those instructions do not work when the
11219 processor is in little endian mode. The exceptions are PPC740 and
11220 PPC750 which permit the instructions usage in little endian mode.
11225 @opindex mno-string
11226 Generate code that uses (does not use) the load string instructions
11227 and the store string word instructions to save multiple registers and
11228 do small block moves. These instructions are generated by default on
11229 POWER systems, and not generated on PowerPC systems. Do not use
11230 @option{-mstring} on little endian PowerPC systems, since those
11231 instructions do not work when the processor is in little endian mode.
11232 The exceptions are PPC740 and PPC750 which permit the instructions
11233 usage in little endian mode.
11238 @opindex mno-update
11239 Generate code that uses (does not use) the load or store instructions
11240 that update the base register to the address of the calculated memory
11241 location. These instructions are generated by default. If you use
11242 @option{-mno-update}, there is a small window between the time that the
11243 stack pointer is updated and the address of the previous frame is
11244 stored, which means code that walks the stack frame across interrupts or
11245 signals may get corrupted data.
11248 @itemx -mno-fused-madd
11249 @opindex mfused-madd
11250 @opindex mno-fused-madd
11251 Generate code that uses (does not use) the floating point multiply and
11252 accumulate instructions. These instructions are generated by default if
11253 hardware floating is used.
11259 Generate code that uses (does not use) the half-word multiply and
11260 multiply-accumulate instructions on the IBM 405 and 440 processors.
11261 These instructions are generated by default when targetting those
11264 @item -mno-bit-align
11266 @opindex mno-bit-align
11267 @opindex mbit-align
11268 On System V.4 and embedded PowerPC systems do not (do) force structures
11269 and unions that contain bit-fields to be aligned to the base type of the
11272 For example, by default a structure containing nothing but 8
11273 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11274 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11275 the structure would be aligned to a 1 byte boundary and be one byte in
11278 @item -mno-strict-align
11279 @itemx -mstrict-align
11280 @opindex mno-strict-align
11281 @opindex mstrict-align
11282 On System V.4 and embedded PowerPC systems do not (do) assume that
11283 unaligned memory references will be handled by the system.
11285 @item -mrelocatable
11286 @itemx -mno-relocatable
11287 @opindex mrelocatable
11288 @opindex mno-relocatable
11289 On embedded PowerPC systems generate code that allows (does not allow)
11290 the program to be relocated to a different address at runtime. If you
11291 use @option{-mrelocatable} on any module, all objects linked together must
11292 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11294 @item -mrelocatable-lib
11295 @itemx -mno-relocatable-lib
11296 @opindex mrelocatable-lib
11297 @opindex mno-relocatable-lib
11298 On embedded PowerPC systems generate code that allows (does not allow)
11299 the program to be relocated to a different address at runtime. Modules
11300 compiled with @option{-mrelocatable-lib} can be linked with either modules
11301 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11302 with modules compiled with the @option{-mrelocatable} options.
11308 On System V.4 and embedded PowerPC systems do not (do) assume that
11309 register 2 contains a pointer to a global area pointing to the addresses
11310 used in the program.
11313 @itemx -mlittle-endian
11315 @opindex mlittle-endian
11316 On System V.4 and embedded PowerPC systems compile code for the
11317 processor in little endian mode. The @option{-mlittle-endian} option is
11318 the same as @option{-mlittle}.
11321 @itemx -mbig-endian
11323 @opindex mbig-endian
11324 On System V.4 and embedded PowerPC systems compile code for the
11325 processor in big endian mode. The @option{-mbig-endian} option is
11326 the same as @option{-mbig}.
11328 @item -mdynamic-no-pic
11329 @opindex mdynamic-no-pic
11330 On Darwin and Mac OS X systems, compile code so that it is not
11331 relocatable, but that its external references are relocatable. The
11332 resulting code is suitable for applications, but not shared
11335 @item -mprioritize-restricted-insns=@var{priority}
11336 @opindex mprioritize-restricted-insns
11337 This option controls the priority that is assigned to
11338 dispatch-slot restricted instructions during the second scheduling
11339 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11340 @var{no/highest/second-highest} priority to dispatch slot restricted
11343 @item -msched-costly-dep=@var{dependence_type}
11344 @opindex msched-costly-dep
11345 This option controls which dependences are considered costly
11346 by the target during instruction scheduling. The argument
11347 @var{dependence_type} takes one of the following values:
11348 @var{no}: no dependence is costly,
11349 @var{all}: all dependences are costly,
11350 @var{true_store_to_load}: a true dependence from store to load is costly,
11351 @var{store_to_load}: any dependence from store to load is costly,
11352 @var{number}: any dependence which latency >= @var{number} is costly.
11354 @item -minsert-sched-nops=@var{scheme}
11355 @opindex minsert-sched-nops
11356 This option controls which nop insertion scheme will be used during
11357 the second scheduling pass. The argument @var{scheme} takes one of the
11359 @var{no}: Don't insert nops.
11360 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11361 according to the scheduler's grouping.
11362 @var{regroup_exact}: Insert nops to force costly dependent insns into
11363 separate groups. Insert exactly as many nops as needed to force an insn
11364 to a new group, according to the estimated processor grouping.
11365 @var{number}: Insert nops to force costly dependent insns into
11366 separate groups. Insert @var{number} nops to force an insn to a new group.
11369 @opindex mcall-sysv
11370 On System V.4 and embedded PowerPC systems compile code using calling
11371 conventions that adheres to the March 1995 draft of the System V
11372 Application Binary Interface, PowerPC processor supplement. This is the
11373 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11375 @item -mcall-sysv-eabi
11376 @opindex mcall-sysv-eabi
11377 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11379 @item -mcall-sysv-noeabi
11380 @opindex mcall-sysv-noeabi
11381 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11383 @item -mcall-solaris
11384 @opindex mcall-solaris
11385 On System V.4 and embedded PowerPC systems compile code for the Solaris
11389 @opindex mcall-linux
11390 On System V.4 and embedded PowerPC systems compile code for the
11391 Linux-based GNU system.
11395 On System V.4 and embedded PowerPC systems compile code for the
11396 Hurd-based GNU system.
11398 @item -mcall-netbsd
11399 @opindex mcall-netbsd
11400 On System V.4 and embedded PowerPC systems compile code for the
11401 NetBSD operating system.
11403 @item -maix-struct-return
11404 @opindex maix-struct-return
11405 Return all structures in memory (as specified by the AIX ABI)@.
11407 @item -msvr4-struct-return
11408 @opindex msvr4-struct-return
11409 Return structures smaller than 8 bytes in registers (as specified by the
11412 @item -mabi=@var{abi-type}
11414 Extend the current ABI with a particular extension, or remove such extension.
11415 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11419 @itemx -mno-prototype
11420 @opindex mprototype
11421 @opindex mno-prototype
11422 On System V.4 and embedded PowerPC systems assume that all calls to
11423 variable argument functions are properly prototyped. Otherwise, the
11424 compiler must insert an instruction before every non prototyped call to
11425 set or clear bit 6 of the condition code register (@var{CR}) to
11426 indicate whether floating point values were passed in the floating point
11427 registers in case the function takes a variable arguments. With
11428 @option{-mprototype}, only calls to prototyped variable argument functions
11429 will set or clear the bit.
11433 On embedded PowerPC systems, assume that the startup module is called
11434 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11435 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11440 On embedded PowerPC systems, assume that the startup module is called
11441 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11446 On embedded PowerPC systems, assume that the startup module is called
11447 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11450 @item -myellowknife
11451 @opindex myellowknife
11452 On embedded PowerPC systems, assume that the startup module is called
11453 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11458 On System V.4 and embedded PowerPC systems, specify that you are
11459 compiling for a VxWorks system.
11463 Specify that you are compiling for the WindISS simulation environment.
11467 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11468 header to indicate that @samp{eabi} extended relocations are used.
11474 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11475 Embedded Applications Binary Interface (eabi) which is a set of
11476 modifications to the System V.4 specifications. Selecting @option{-meabi}
11477 means that the stack is aligned to an 8 byte boundary, a function
11478 @code{__eabi} is called to from @code{main} to set up the eabi
11479 environment, and the @option{-msdata} option can use both @code{r2} and
11480 @code{r13} to point to two separate small data areas. Selecting
11481 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11482 do not call an initialization function from @code{main}, and the
11483 @option{-msdata} option will only use @code{r13} to point to a single
11484 small data area. The @option{-meabi} option is on by default if you
11485 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11488 @opindex msdata=eabi
11489 On System V.4 and embedded PowerPC systems, put small initialized
11490 @code{const} global and static data in the @samp{.sdata2} section, which
11491 is pointed to by register @code{r2}. Put small initialized
11492 non-@code{const} global and static data in the @samp{.sdata} section,
11493 which is pointed to by register @code{r13}. Put small uninitialized
11494 global and static data in the @samp{.sbss} section, which is adjacent to
11495 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11496 incompatible with the @option{-mrelocatable} option. The
11497 @option{-msdata=eabi} option also sets the @option{-memb} option.
11500 @opindex msdata=sysv
11501 On System V.4 and embedded PowerPC systems, put small global and static
11502 data in the @samp{.sdata} section, which is pointed to by register
11503 @code{r13}. Put small uninitialized global and static data in the
11504 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11505 The @option{-msdata=sysv} option is incompatible with the
11506 @option{-mrelocatable} option.
11508 @item -msdata=default
11510 @opindex msdata=default
11512 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11513 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11514 same as @option{-msdata=sysv}.
11517 @opindex msdata-data
11518 On System V.4 and embedded PowerPC systems, put small global
11519 data in the @samp{.sdata} section. Put small uninitialized global
11520 data in the @samp{.sbss} section. Do not use register @code{r13}
11521 to address small data however. This is the default behavior unless
11522 other @option{-msdata} options are used.
11526 @opindex msdata=none
11528 On embedded PowerPC systems, put all initialized global and static data
11529 in the @samp{.data} section, and all uninitialized data in the
11530 @samp{.bss} section.
11534 @cindex smaller data references (PowerPC)
11535 @cindex .sdata/.sdata2 references (PowerPC)
11536 On embedded PowerPC systems, put global and static items less than or
11537 equal to @var{num} bytes into the small data or bss sections instead of
11538 the normal data or bss section. By default, @var{num} is 8. The
11539 @option{-G @var{num}} switch is also passed to the linker.
11540 All modules should be compiled with the same @option{-G @var{num}} value.
11543 @itemx -mno-regnames
11545 @opindex mno-regnames
11546 On System V.4 and embedded PowerPC systems do (do not) emit register
11547 names in the assembly language output using symbolic forms.
11550 @itemx -mno-longcall
11552 @opindex mno-longcall
11553 Default to making all function calls indirectly, using a register, so
11554 that functions which reside further than 32 megabytes (33,554,432
11555 bytes) from the current location can be called. This setting can be
11556 overridden by the @code{shortcall} function attribute, or by
11557 @code{#pragma longcall(0)}.
11559 Some linkers are capable of detecting out-of-range calls and generating
11560 glue code on the fly. On these systems, long calls are unnecessary and
11561 generate slower code. As of this writing, the AIX linker can do this,
11562 as can the GNU linker for PowerPC/64. It is planned to add this feature
11563 to the GNU linker for 32-bit PowerPC systems as well.
11565 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11566 callee, L42'', plus a ``branch island'' (glue code). The two target
11567 addresses represent the callee and the ``branch island''. The
11568 Darwin/PPC linker will prefer the first address and generate a ``bl
11569 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11570 otherwise, the linker will generate ``bl L42'' to call the ``branch
11571 island''. The ``branch island'' is appended to the body of the
11572 calling function; it computes the full 32-bit address of the callee
11575 On Mach-O (Darwin) systems, this option directs the compiler emit to
11576 the glue for every direct call, and the Darwin linker decides whether
11577 to use or discard it.
11579 In the future, we may cause GCC to ignore all longcall specifications
11580 when the linker is known to generate glue.
11584 Adds support for multithreading with the @dfn{pthreads} library.
11585 This option sets flags for both the preprocessor and linker.
11589 @node S/390 and zSeries Options
11590 @subsection S/390 and zSeries Options
11591 @cindex S/390 and zSeries Options
11593 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11597 @itemx -msoft-float
11598 @opindex mhard-float
11599 @opindex msoft-float
11600 Use (do not use) the hardware floating-point instructions and registers
11601 for floating-point operations. When @option{-msoft-float} is specified,
11602 functions in @file{libgcc.a} will be used to perform floating-point
11603 operations. When @option{-mhard-float} is specified, the compiler
11604 generates IEEE floating-point instructions. This is the default.
11607 @itemx -mno-backchain
11608 @opindex mbackchain
11609 @opindex mno-backchain
11610 Store (do not store) the address of the caller's frame as backchain pointer
11611 into the callee's stack frame.
11612 A backchain may be needed to allow debugging using tools that do not understand
11613 DWARF-2 call frame information.
11614 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11615 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11616 the backchain is placed into the topmost word of the 96/160 byte register
11619 In general, code compiled with @option{-mbackchain} is call-compatible with
11620 code compiled with @option{-mmo-backchain}; however, use of the backchain
11621 for debugging purposes usually requires that the whole binary is built with
11622 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11623 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11624 to build a linux kernel use @option{-msoft-float}.
11626 The default is to not maintain the backchain.
11628 @item -mpacked-stack
11629 @item -mno-packed-stack
11630 @opindex mpacked-stack
11631 @opindex mno-packed-stack
11632 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11633 specified, the compiler uses the all fields of the 96/160 byte register save
11634 area only for their default purpose; unused fields still take up stack space.
11635 When @option{-mpacked-stack} is specified, register save slots are densely
11636 packed at the top of the register save area; unused space is reused for other
11637 purposes, allowing for more efficient use of the available stack space.
11638 However, when @option{-mbackchain} is also in effect, the topmost word of
11639 the save area is always used to store the backchain, and the return address
11640 register is always saved two words below the backchain.
11642 As long as the stack frame backchain is not used, code generated with
11643 @option{-mpacked-stack} is call-compatible with code generated with
11644 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11645 S/390 or zSeries generated code that uses the stack frame backchain at run
11646 time, not just for debugging purposes. Such code is not call-compatible
11647 with code compiled with @option{-mpacked-stack}. Also, note that the
11648 combination of @option{-mbackchain},
11649 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11650 to build a linux kernel use @option{-msoft-float}.
11652 The default is to not use the packed stack layout.
11655 @itemx -mno-small-exec
11656 @opindex msmall-exec
11657 @opindex mno-small-exec
11658 Generate (or do not generate) code using the @code{bras} instruction
11659 to do subroutine calls.
11660 This only works reliably if the total executable size does not
11661 exceed 64k. The default is to use the @code{basr} instruction instead,
11662 which does not have this limitation.
11668 When @option{-m31} is specified, generate code compliant to the
11669 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11670 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11671 particular to generate 64-bit instructions. For the @samp{s390}
11672 targets, the default is @option{-m31}, while the @samp{s390x}
11673 targets default to @option{-m64}.
11679 When @option{-mzarch} is specified, generate code using the
11680 instructions available on z/Architecture.
11681 When @option{-mesa} is specified, generate code using the
11682 instructions available on ESA/390. Note that @option{-mesa} is
11683 not possible with @option{-m64}.
11684 When generating code compliant to the GNU/Linux for S/390 ABI,
11685 the default is @option{-mesa}. When generating code compliant
11686 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11692 Generate (or do not generate) code using the @code{mvcle} instruction
11693 to perform block moves. When @option{-mno-mvcle} is specified,
11694 use a @code{mvc} loop instead. This is the default unless optimizing for
11701 Print (or do not print) additional debug information when compiling.
11702 The default is to not print debug information.
11704 @item -march=@var{cpu-type}
11706 Generate code that will run on @var{cpu-type}, which is the name of a system
11707 representing a certain processor type. Possible values for
11708 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11709 When generating code using the instructions available on z/Architecture,
11710 the default is @option{-march=z900}. Otherwise, the default is
11711 @option{-march=g5}.
11713 @item -mtune=@var{cpu-type}
11715 Tune to @var{cpu-type} everything applicable about the generated code,
11716 except for the ABI and the set of available instructions.
11717 The list of @var{cpu-type} values is the same as for @option{-march}.
11718 The default is the value used for @option{-march}.
11721 @itemx -mno-tpf-trace
11722 @opindex mtpf-trace
11723 @opindex mno-tpf-trace
11724 Generate code that adds (does not add) in TPF OS specific branches to trace
11725 routines in the operating system. This option is off by default, even
11726 when compiling for the TPF OS@.
11729 @itemx -mno-fused-madd
11730 @opindex mfused-madd
11731 @opindex mno-fused-madd
11732 Generate code that uses (does not use) the floating point multiply and
11733 accumulate instructions. These instructions are generated by default if
11734 hardware floating point is used.
11736 @item -mwarn-framesize=@var{framesize}
11737 @opindex mwarn-framesize
11738 Emit a warning if the current function exceeds the given frame size. Because
11739 this is a compile time check it doesn't need to be a real problem when the program
11740 runs. It is intended to identify functions which most probably cause
11741 a stack overflow. It is useful to be used in an environment with limited stack
11742 size e.g.@: the linux kernel.
11744 @item -mwarn-dynamicstack
11745 @opindex mwarn-dynamicstack
11746 Emit a warning if the function calls alloca or uses dynamically
11747 sized arrays. This is generally a bad idea with a limited stack size.
11749 @item -mstack-guard=@var{stack-guard}
11750 @item -mstack-size=@var{stack-size}
11751 @opindex mstack-guard
11752 @opindex mstack-size
11753 These arguments always have to be used in conjunction. If they are present the s390
11754 back end emits additional instructions in the function prologue which trigger a trap
11755 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11756 (remember that the stack on s390 grows downward). These options are intended to
11757 be used to help debugging stack overflow problems. The additionally emitted code
11758 causes only little overhead and hence can also be used in production like systems
11759 without greater performance degradation. The given values have to be exact
11760 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
11762 In order to be efficient the extra code makes the assumption that the stack starts
11763 at an address aligned to the value given by @var{stack-size}.
11767 @subsection SH Options
11769 These @samp{-m} options are defined for the SH implementations:
11774 Generate code for the SH1.
11778 Generate code for the SH2.
11781 Generate code for the SH2e.
11785 Generate code for the SH3.
11789 Generate code for the SH3e.
11793 Generate code for the SH4 without a floating-point unit.
11795 @item -m4-single-only
11796 @opindex m4-single-only
11797 Generate code for the SH4 with a floating-point unit that only
11798 supports single-precision arithmetic.
11802 Generate code for the SH4 assuming the floating-point unit is in
11803 single-precision mode by default.
11807 Generate code for the SH4.
11811 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11812 floating-point unit is not used.
11814 @item -m4a-single-only
11815 @opindex m4a-single-only
11816 Generate code for the SH4a, in such a way that no double-precision
11817 floating point operations are used.
11820 @opindex m4a-single
11821 Generate code for the SH4a assuming the floating-point unit is in
11822 single-precision mode by default.
11826 Generate code for the SH4a.
11830 Same as @option{-m4a-nofpu}, except that it implicitly passes
11831 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11832 instructions at the moment.
11836 Compile code for the processor in big endian mode.
11840 Compile code for the processor in little endian mode.
11844 Align doubles at 64-bit boundaries. Note that this changes the calling
11845 conventions, and thus some functions from the standard C library will
11846 not work unless you recompile it first with @option{-mdalign}.
11850 Shorten some address references at link time, when possible; uses the
11851 linker option @option{-relax}.
11855 Use 32-bit offsets in @code{switch} tables. The default is to use
11860 Enable the use of the instruction @code{fmovd}.
11864 Comply with the calling conventions defined by Renesas.
11868 Comply with the calling conventions defined by Renesas.
11872 Comply with the calling conventions defined for GCC before the Renesas
11873 conventions were available. This option is the default for all
11874 targets of the SH toolchain except for @samp{sh-symbianelf}.
11877 @opindex mnomacsave
11878 Mark the @code{MAC} register as call-clobbered, even if
11879 @option{-mhitachi} is given.
11883 Increase IEEE-compliance of floating-point code.
11884 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11885 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11886 comparisons of NANs / infinities incurs extra overhead in every
11887 floating point comparison, therefore the default is set to
11888 @option{-ffinite-math-only}.
11892 Dump instruction size and location in the assembly code.
11895 @opindex mpadstruct
11896 This option is deprecated. It pads structures to multiple of 4 bytes,
11897 which is incompatible with the SH ABI@.
11901 Optimize for space instead of speed. Implied by @option{-Os}.
11904 @opindex mprefergot
11905 When generating position-independent code, emit function calls using
11906 the Global Offset Table instead of the Procedure Linkage Table.
11910 Generate a library function call to invalidate instruction cache
11911 entries, after fixing up a trampoline. This library function call
11912 doesn't assume it can write to the whole memory address space. This
11913 is the default when the target is @code{sh-*-linux*}.
11915 @item -multcost=@var{number}
11916 @opindex multcost=@var{number}
11917 Set the cost to assume for a multiply insn.
11919 @item -mdiv=@var{strategy}
11920 @opindex mdiv=@var{strategy}
11921 Set the division strategy to use for SHmedia code. @var{strategy} must be
11922 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11923 inv:call2, inv:fp .
11924 "fp" performs the operation in floating point. This has a very high latency,
11925 but needs only a few instructions, so it might be a good choice if
11926 your code has enough easily exploitable ILP to allow the compiler to
11927 schedule the floating point instructions together with other instructions.
11928 Division by zero causes a floating point exception.
11929 "inv" uses integer operations to calculate the inverse of the divisor,
11930 and then multiplies the dividend with the inverse. This strategy allows
11931 cse and hoisting of the inverse calculation. Division by zero calculates
11932 an unspecified result, but does not trap.
11933 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11934 have been found, or if the entire operation has been hoisted to the same
11935 place, the last stages of the inverse calculation are intertwined with the
11936 final multiply to reduce the overall latency, at the expense of using a few
11937 more instructions, and thus offering fewer scheduling opportunities with
11939 "call" calls a library function that usually implements the inv:minlat
11941 This gives high code density for m5-*media-nofpu compilations.
11942 "call2" uses a different entry point of the same library function, where it
11943 assumes that a pointer to a lookup table has already been set up, which
11944 exposes the pointer load to cse / code hoisting optimizations.
11945 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11946 code generation, but if the code stays unoptimized, revert to the "call",
11947 "call2", or "fp" strategies, respectively. Note that the
11948 potentially-trapping side effect of division by zero is carried by a
11949 separate instruction, so it is possible that all the integer instructions
11950 are hoisted out, but the marker for the side effect stays where it is.
11951 A recombination to fp operations or a call is not possible in that case.
11952 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11953 that the inverse calculation was nor separated from the multiply, they speed
11954 up division where the dividend fits into 20 bits (plus sign where applicable),
11955 by inserting a test to skip a number of operations in this case; this test
11956 slows down the case of larger dividends. inv20u assumes the case of a such
11957 a small dividend to be unlikely, and inv20l assumes it to be likely.
11959 @item -mdivsi3_libfunc=@var{name}
11960 @opindex mdivsi3_libfunc=@var{name}
11961 Set the name of the library function used for 32 bit signed division to
11962 @var{name}. This only affect the name used in the call and inv:call
11963 division strategies, and the compiler will still expect the same
11964 sets of input/output/clobbered registers as if this option was not present.
11966 @item -madjust-unroll
11967 @opindex madjust-unroll
11968 Throttle unrolling to avoid thrashing target registers.
11969 This option only has an effect if the gcc code base supports the
11970 TARGET_ADJUST_UNROLL_MAX target hook.
11972 @item -mindexed-addressing
11973 @opindex mindexed-addressing
11974 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11975 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11976 semantics for the indexed addressing mode. The architecture allows the
11977 implementation of processors with 64 bit MMU, which the OS could use to
11978 get 32 bit addressing, but since no current hardware implementation supports
11979 this or any other way to make the indexed addressing mode safe to use in
11980 the 32 bit ABI, the default is -mno-indexed-addressing.
11982 @item -mgettrcost=@var{number}
11983 @opindex mgettrcost=@var{number}
11984 Set the cost assumed for the gettr instruction to @var{number}.
11985 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11989 Assume pt* instructions won't trap. This will generally generate better
11990 scheduled code, but is unsafe on current hardware. The current architecture
11991 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11992 This has the unintentional effect of making it unsafe to schedule ptabs /
11993 ptrel before a branch, or hoist it out of a loop. For example,
11994 __do_global_ctors, a part of libgcc that runs constructors at program
11995 startup, calls functions in a list which is delimited by -1. With the
11996 -mpt-fixed option, the ptabs will be done before testing against -1.
11997 That means that all the constructors will be run a bit quicker, but when
11998 the loop comes to the end of the list, the program crashes because ptabs
11999 loads -1 into a target register. Since this option is unsafe for any
12000 hardware implementing the current architecture specification, the default
12001 is -mno-pt-fixed. Unless the user specifies a specific cost with
12002 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12003 this deters register allocation using target registers for storing
12006 @item -minvalid-symbols
12007 @opindex minvalid-symbols
12008 Assume symbols might be invalid. Ordinary function symbols generated by
12009 the compiler will always be valid to load with movi/shori/ptabs or
12010 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12011 to generate symbols that will cause ptabs / ptrel to trap.
12012 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12013 It will then prevent cross-basic-block cse, hoisting and most scheduling
12014 of symbol loads. The default is @option{-mno-invalid-symbols}.
12017 @node SPARC Options
12018 @subsection SPARC Options
12019 @cindex SPARC options
12021 These @samp{-m} options are supported on the SPARC:
12024 @item -mno-app-regs
12026 @opindex mno-app-regs
12028 Specify @option{-mapp-regs} to generate output using the global registers
12029 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12032 To be fully SVR4 ABI compliant at the cost of some performance loss,
12033 specify @option{-mno-app-regs}. You should compile libraries and system
12034 software with this option.
12037 @itemx -mhard-float
12039 @opindex mhard-float
12040 Generate output containing floating point instructions. This is the
12044 @itemx -msoft-float
12046 @opindex msoft-float
12047 Generate output containing library calls for floating point.
12048 @strong{Warning:} the requisite libraries are not available for all SPARC
12049 targets. Normally the facilities of the machine's usual C compiler are
12050 used, but this cannot be done directly in cross-compilation. You must make
12051 your own arrangements to provide suitable library functions for
12052 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12053 @samp{sparclite-*-*} do provide software floating point support.
12055 @option{-msoft-float} changes the calling convention in the output file;
12056 therefore, it is only useful if you compile @emph{all} of a program with
12057 this option. In particular, you need to compile @file{libgcc.a}, the
12058 library that comes with GCC, with @option{-msoft-float} in order for
12061 @item -mhard-quad-float
12062 @opindex mhard-quad-float
12063 Generate output containing quad-word (long double) floating point
12066 @item -msoft-quad-float
12067 @opindex msoft-quad-float
12068 Generate output containing library calls for quad-word (long double)
12069 floating point instructions. The functions called are those specified
12070 in the SPARC ABI@. This is the default.
12072 As of this writing, there are no SPARC implementations that have hardware
12073 support for the quad-word floating point instructions. They all invoke
12074 a trap handler for one of these instructions, and then the trap handler
12075 emulates the effect of the instruction. Because of the trap handler overhead,
12076 this is much slower than calling the ABI library routines. Thus the
12077 @option{-msoft-quad-float} option is the default.
12079 @item -mno-unaligned-doubles
12080 @itemx -munaligned-doubles
12081 @opindex mno-unaligned-doubles
12082 @opindex munaligned-doubles
12083 Assume that doubles have 8 byte alignment. This is the default.
12085 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12086 alignment only if they are contained in another type, or if they have an
12087 absolute address. Otherwise, it assumes they have 4 byte alignment.
12088 Specifying this option avoids some rare compatibility problems with code
12089 generated by other compilers. It is not the default because it results
12090 in a performance loss, especially for floating point code.
12092 @item -mno-faster-structs
12093 @itemx -mfaster-structs
12094 @opindex mno-faster-structs
12095 @opindex mfaster-structs
12096 With @option{-mfaster-structs}, the compiler assumes that structures
12097 should have 8 byte alignment. This enables the use of pairs of
12098 @code{ldd} and @code{std} instructions for copies in structure
12099 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12100 However, the use of this changed alignment directly violates the SPARC
12101 ABI@. Thus, it's intended only for use on targets where the developer
12102 acknowledges that their resulting code will not be directly in line with
12103 the rules of the ABI@.
12105 @item -mimpure-text
12106 @opindex mimpure-text
12107 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12108 the compiler to not pass @option{-z text} to the linker when linking a
12109 shared object. Using this option, you can link position-dependent
12110 code into a shared object.
12112 @option{-mimpure-text} suppresses the ``relocations remain against
12113 allocatable but non-writable sections'' linker error message.
12114 However, the necessary relocations will trigger copy-on-write, and the
12115 shared object is not actually shared across processes. Instead of
12116 using @option{-mimpure-text}, you should compile all source code with
12117 @option{-fpic} or @option{-fPIC}.
12119 This option is only available on SunOS and Solaris.
12121 @item -mcpu=@var{cpu_type}
12123 Set the instruction set, register set, and instruction scheduling parameters
12124 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12125 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12126 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12127 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
12128 @samp{ultrasparc3}.
12130 Default instruction scheduling parameters are used for values that select
12131 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12132 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12134 Here is a list of each supported architecture and their supported
12139 v8: supersparc, hypersparc
12140 sparclite: f930, f934, sparclite86x
12142 v9: ultrasparc, ultrasparc3
12145 By default (unless configured otherwise), GCC generates code for the V7
12146 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12147 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12148 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12149 SPARCStation 1, 2, IPX etc.
12151 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12152 architecture. The only difference from V7 code is that the compiler emits
12153 the integer multiply and integer divide instructions which exist in SPARC-V8
12154 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12155 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12158 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12159 the SPARC architecture. This adds the integer multiply, integer divide step
12160 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12161 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12162 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12163 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12164 MB86934 chip, which is the more recent SPARClite with FPU@.
12166 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12167 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12168 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12169 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12170 optimizes it for the TEMIC SPARClet chip.
12172 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12173 architecture. This adds 64-bit integer and floating-point move instructions,
12174 3 additional floating-point condition code registers and conditional move
12175 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12176 optimizes it for the Sun UltraSPARC I/II chips. With
12177 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12178 Sun UltraSPARC III chip.
12180 @item -mtune=@var{cpu_type}
12182 Set the instruction scheduling parameters for machine type
12183 @var{cpu_type}, but do not set the instruction set or register set that the
12184 option @option{-mcpu=@var{cpu_type}} would.
12186 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12187 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12188 that select a particular cpu implementation. Those are @samp{cypress},
12189 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12190 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
12191 @samp{ultrasparc3}.
12196 @opindex mno-v8plus
12197 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12198 difference from the V8 ABI is that the global and out registers are
12199 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12200 mode for all SPARC-V9 processors.
12206 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12207 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12210 These @samp{-m} options are supported in addition to the above
12211 on SPARC-V9 processors in 64-bit environments:
12214 @item -mlittle-endian
12215 @opindex mlittle-endian
12216 Generate code for a processor running in little-endian mode. It is only
12217 available for a few configurations and most notably not on Solaris and Linux.
12223 Generate code for a 32-bit or 64-bit environment.
12224 The 32-bit environment sets int, long and pointer to 32 bits.
12225 The 64-bit environment sets int to 32 bits and long and pointer
12228 @item -mcmodel=medlow
12229 @opindex mcmodel=medlow
12230 Generate code for the Medium/Low code model: 64-bit addresses, programs
12231 must be linked in the low 32 bits of memory. Programs can be statically
12232 or dynamically linked.
12234 @item -mcmodel=medmid
12235 @opindex mcmodel=medmid
12236 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12237 must be linked in the low 44 bits of memory, the text and data segments must
12238 be less than 2GB in size and the data segment must be located within 2GB of
12241 @item -mcmodel=medany
12242 @opindex mcmodel=medany
12243 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12244 may be linked anywhere in memory, the text and data segments must be less
12245 than 2GB in size and the data segment must be located within 2GB of the
12248 @item -mcmodel=embmedany
12249 @opindex mcmodel=embmedany
12250 Generate code for the Medium/Anywhere code model for embedded systems:
12251 64-bit addresses, the text and data segments must be less than 2GB in
12252 size, both starting anywhere in memory (determined at link time). The
12253 global register %g4 points to the base of the data segment. Programs
12254 are statically linked and PIC is not supported.
12257 @itemx -mno-stack-bias
12258 @opindex mstack-bias
12259 @opindex mno-stack-bias
12260 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12261 frame pointer if present, are offset by @minus{}2047 which must be added back
12262 when making stack frame references. This is the default in 64-bit mode.
12263 Otherwise, assume no such offset is present.
12266 These switches are supported in addition to the above on Solaris:
12271 Add support for multithreading using the Solaris threads library. This
12272 option sets flags for both the preprocessor and linker. This option does
12273 not affect the thread safety of object code produced by the compiler or
12274 that of libraries supplied with it.
12278 Add support for multithreading using the POSIX threads library. This
12279 option sets flags for both the preprocessor and linker. This option does
12280 not affect the thread safety of object code produced by the compiler or
12281 that of libraries supplied with it.
12284 @node System V Options
12285 @subsection Options for System V
12287 These additional options are available on System V Release 4 for
12288 compatibility with other compilers on those systems:
12293 Create a shared object.
12294 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12298 Identify the versions of each tool used by the compiler, in a
12299 @code{.ident} assembler directive in the output.
12303 Refrain from adding @code{.ident} directives to the output file (this is
12306 @item -YP,@var{dirs}
12308 Search the directories @var{dirs}, and no others, for libraries
12309 specified with @option{-l}.
12311 @item -Ym,@var{dir}
12313 Look in the directory @var{dir} to find the M4 preprocessor.
12314 The assembler uses this option.
12315 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12316 @c the generic assembler that comes with Solaris takes just -Ym.
12319 @node TMS320C3x/C4x Options
12320 @subsection TMS320C3x/C4x Options
12321 @cindex TMS320C3x/C4x Options
12323 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12327 @item -mcpu=@var{cpu_type}
12329 Set the instruction set, register set, and instruction scheduling
12330 parameters for machine type @var{cpu_type}. Supported values for
12331 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12332 @samp{c44}. The default is @samp{c40} to generate code for the
12337 @itemx -msmall-memory
12339 @opindex mbig-memory
12341 @opindex msmall-memory
12343 Generates code for the big or small memory model. The small memory
12344 model assumed that all data fits into one 64K word page. At run-time
12345 the data page (DP) register must be set to point to the 64K page
12346 containing the .bss and .data program sections. The big memory model is
12347 the default and requires reloading of the DP register for every direct
12354 Allow (disallow) allocation of general integer operands into the block
12355 count register BK@.
12361 Enable (disable) generation of code using decrement and branch,
12362 DBcond(D), instructions. This is enabled by default for the C4x. To be
12363 on the safe side, this is disabled for the C3x, since the maximum
12364 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12365 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12366 that it can utilize the decrement and branch instruction, but will give
12367 up if there is more than one memory reference in the loop. Thus a loop
12368 where the loop counter is decremented can generate slightly more
12369 efficient code, in cases where the RPTB instruction cannot be utilized.
12371 @item -mdp-isr-reload
12373 @opindex mdp-isr-reload
12375 Force the DP register to be saved on entry to an interrupt service
12376 routine (ISR), reloaded to point to the data section, and restored on
12377 exit from the ISR@. This should not be required unless someone has
12378 violated the small memory model by modifying the DP register, say within
12385 For the C3x use the 24-bit MPYI instruction for integer multiplies
12386 instead of a library call to guarantee 32-bit results. Note that if one
12387 of the operands is a constant, then the multiplication will be performed
12388 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12389 then squaring operations are performed inline instead of a library call.
12392 @itemx -mno-fast-fix
12394 @opindex mno-fast-fix
12395 The C3x/C4x FIX instruction to convert a floating point value to an
12396 integer value chooses the nearest integer less than or equal to the
12397 floating point value rather than to the nearest integer. Thus if the
12398 floating point number is negative, the result will be incorrectly
12399 truncated an additional code is necessary to detect and correct this
12400 case. This option can be used to disable generation of the additional
12401 code required to correct the result.
12407 Enable (disable) generation of repeat block sequences using the RPTB
12408 instruction for zero overhead looping. The RPTB construct is only used
12409 for innermost loops that do not call functions or jump across the loop
12410 boundaries. There is no advantage having nested RPTB loops due to the
12411 overhead required to save and restore the RC, RS, and RE registers.
12412 This is enabled by default with @option{-O2}.
12414 @item -mrpts=@var{count}
12418 Enable (disable) the use of the single instruction repeat instruction
12419 RPTS@. If a repeat block contains a single instruction, and the loop
12420 count can be guaranteed to be less than the value @var{count}, GCC will
12421 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12422 then a RPTS will be emitted even if the loop count cannot be determined
12423 at compile time. Note that the repeated instruction following RPTS does
12424 not have to be reloaded from memory each iteration, thus freeing up the
12425 CPU buses for operands. However, since interrupts are blocked by this
12426 instruction, it is disabled by default.
12428 @item -mloop-unsigned
12429 @itemx -mno-loop-unsigned
12430 @opindex mloop-unsigned
12431 @opindex mno-loop-unsigned
12432 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12433 is @math{2^{31} + 1} since these instructions test if the iteration count is
12434 negative to terminate the loop. If the iteration count is unsigned
12435 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12436 exceeded. This switch allows an unsigned iteration count.
12440 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12441 with. This also enforces compatibility with the API employed by the TI
12442 C3x C compiler. For example, long doubles are passed as structures
12443 rather than in floating point registers.
12449 Generate code that uses registers (stack) for passing arguments to functions.
12450 By default, arguments are passed in registers where possible rather
12451 than by pushing arguments on to the stack.
12453 @item -mparallel-insns
12454 @itemx -mno-parallel-insns
12455 @opindex mparallel-insns
12456 @opindex mno-parallel-insns
12457 Allow the generation of parallel instructions. This is enabled by
12458 default with @option{-O2}.
12460 @item -mparallel-mpy
12461 @itemx -mno-parallel-mpy
12462 @opindex mparallel-mpy
12463 @opindex mno-parallel-mpy
12464 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12465 provided @option{-mparallel-insns} is also specified. These instructions have
12466 tight register constraints which can pessimize the code generation
12467 of large functions.
12472 @subsection V850 Options
12473 @cindex V850 Options
12475 These @samp{-m} options are defined for V850 implementations:
12479 @itemx -mno-long-calls
12480 @opindex mlong-calls
12481 @opindex mno-long-calls
12482 Treat all calls as being far away (near). If calls are assumed to be
12483 far away, the compiler will always load the functions address up into a
12484 register, and call indirect through the pointer.
12490 Do not optimize (do optimize) basic blocks that use the same index
12491 pointer 4 or more times to copy pointer into the @code{ep} register, and
12492 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12493 option is on by default if you optimize.
12495 @item -mno-prolog-function
12496 @itemx -mprolog-function
12497 @opindex mno-prolog-function
12498 @opindex mprolog-function
12499 Do not use (do use) external functions to save and restore registers
12500 at the prologue and epilogue of a function. The external functions
12501 are slower, but use less code space if more than one function saves
12502 the same number of registers. The @option{-mprolog-function} option
12503 is on by default if you optimize.
12507 Try to make the code as small as possible. At present, this just turns
12508 on the @option{-mep} and @option{-mprolog-function} options.
12510 @item -mtda=@var{n}
12512 Put static or global variables whose size is @var{n} bytes or less into
12513 the tiny data area that register @code{ep} points to. The tiny data
12514 area can hold up to 256 bytes in total (128 bytes for byte references).
12516 @item -msda=@var{n}
12518 Put static or global variables whose size is @var{n} bytes or less into
12519 the small data area that register @code{gp} points to. The small data
12520 area can hold up to 64 kilobytes.
12522 @item -mzda=@var{n}
12524 Put static or global variables whose size is @var{n} bytes or less into
12525 the first 32 kilobytes of memory.
12529 Specify that the target processor is the V850.
12532 @opindex mbig-switch
12533 Generate code suitable for big switch tables. Use this option only if
12534 the assembler/linker complain about out of range branches within a switch
12539 This option will cause r2 and r5 to be used in the code generated by
12540 the compiler. This setting is the default.
12542 @item -mno-app-regs
12543 @opindex mno-app-regs
12544 This option will cause r2 and r5 to be treated as fixed registers.
12548 Specify that the target processor is the V850E1. The preprocessor
12549 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12550 this option is used.
12554 Specify that the target processor is the V850E@. The preprocessor
12555 constant @samp{__v850e__} will be defined if this option is used.
12557 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12558 are defined then a default target processor will be chosen and the
12559 relevant @samp{__v850*__} preprocessor constant will be defined.
12561 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12562 defined, regardless of which processor variant is the target.
12564 @item -mdisable-callt
12565 @opindex mdisable-callt
12566 This option will suppress generation of the CALLT instruction for the
12567 v850e and v850e1 flavors of the v850 architecture. The default is
12568 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12573 @subsection VAX Options
12574 @cindex VAX options
12576 These @samp{-m} options are defined for the VAX:
12581 Do not output certain jump instructions (@code{aobleq} and so on)
12582 that the Unix assembler for the VAX cannot handle across long
12587 Do output those jump instructions, on the assumption that you
12588 will assemble with the GNU assembler.
12592 Output code for g-format floating point numbers instead of d-format.
12595 @node x86-64 Options
12596 @subsection x86-64 Options
12597 @cindex x86-64 options
12599 These are listed under @xref{i386 and x86-64 Options}.
12601 @node Xstormy16 Options
12602 @subsection Xstormy16 Options
12603 @cindex Xstormy16 Options
12605 These options are defined for Xstormy16:
12610 Choose startup files and linker script suitable for the simulator.
12613 @node Xtensa Options
12614 @subsection Xtensa Options
12615 @cindex Xtensa Options
12617 These options are supported for Xtensa targets:
12621 @itemx -mno-const16
12623 @opindex mno-const16
12624 Enable or disable use of @code{CONST16} instructions for loading
12625 constant values. The @code{CONST16} instruction is currently not a
12626 standard option from Tensilica. When enabled, @code{CONST16}
12627 instructions are always used in place of the standard @code{L32R}
12628 instructions. The use of @code{CONST16} is enabled by default only if
12629 the @code{L32R} instruction is not available.
12632 @itemx -mno-fused-madd
12633 @opindex mfused-madd
12634 @opindex mno-fused-madd
12635 Enable or disable use of fused multiply/add and multiply/subtract
12636 instructions in the floating-point option. This has no effect if the
12637 floating-point option is not also enabled. Disabling fused multiply/add
12638 and multiply/subtract instructions forces the compiler to use separate
12639 instructions for the multiply and add/subtract operations. This may be
12640 desirable in some cases where strict IEEE 754-compliant results are
12641 required: the fused multiply add/subtract instructions do not round the
12642 intermediate result, thereby producing results with @emph{more} bits of
12643 precision than specified by the IEEE standard. Disabling fused multiply
12644 add/subtract instructions also ensures that the program output is not
12645 sensitive to the compiler's ability to combine multiply and add/subtract
12648 @item -mtext-section-literals
12649 @itemx -mno-text-section-literals
12650 @opindex mtext-section-literals
12651 @opindex mno-text-section-literals
12652 Control the treatment of literal pools. The default is
12653 @option{-mno-text-section-literals}, which places literals in a separate
12654 section in the output file. This allows the literal pool to be placed
12655 in a data RAM/ROM, and it also allows the linker to combine literal
12656 pools from separate object files to remove redundant literals and
12657 improve code size. With @option{-mtext-section-literals}, the literals
12658 are interspersed in the text section in order to keep them as close as
12659 possible to their references. This may be necessary for large assembly
12662 @item -mtarget-align
12663 @itemx -mno-target-align
12664 @opindex mtarget-align
12665 @opindex mno-target-align
12666 When this option is enabled, GCC instructs the assembler to
12667 automatically align instructions to reduce branch penalties at the
12668 expense of some code density. The assembler attempts to widen density
12669 instructions to align branch targets and the instructions following call
12670 instructions. If there are not enough preceding safe density
12671 instructions to align a target, no widening will be performed. The
12672 default is @option{-mtarget-align}. These options do not affect the
12673 treatment of auto-aligned instructions like @code{LOOP}, which the
12674 assembler will always align, either by widening density instructions or
12675 by inserting no-op instructions.
12678 @itemx -mno-longcalls
12679 @opindex mlongcalls
12680 @opindex mno-longcalls
12681 When this option is enabled, GCC instructs the assembler to translate
12682 direct calls to indirect calls unless it can determine that the target
12683 of a direct call is in the range allowed by the call instruction. This
12684 translation typically occurs for calls to functions in other source
12685 files. Specifically, the assembler translates a direct @code{CALL}
12686 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12687 The default is @option{-mno-longcalls}. This option should be used in
12688 programs where the call target can potentially be out of range. This
12689 option is implemented in the assembler, not the compiler, so the
12690 assembly code generated by GCC will still show direct call
12691 instructions---look at the disassembled object code to see the actual
12692 instructions. Note that the assembler will use an indirect call for
12693 every cross-file call, not just those that really will be out of range.
12696 @node zSeries Options
12697 @subsection zSeries Options
12698 @cindex zSeries options
12700 These are listed under @xref{S/390 and zSeries Options}.
12702 @node Code Gen Options
12703 @section Options for Code Generation Conventions
12704 @cindex code generation conventions
12705 @cindex options, code generation
12706 @cindex run-time options
12708 These machine-independent options control the interface conventions
12709 used in code generation.
12711 Most of them have both positive and negative forms; the negative form
12712 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12713 one of the forms is listed---the one which is not the default. You
12714 can figure out the other form by either removing @samp{no-} or adding
12718 @item -fbounds-check
12719 @opindex fbounds-check
12720 For front-ends that support it, generate additional code to check that
12721 indices used to access arrays are within the declared range. This is
12722 currently only supported by the Java and Fortran 77 front-ends, where
12723 this option defaults to true and false respectively.
12727 This option generates traps for signed overflow on addition, subtraction,
12728 multiplication operations.
12732 This option instructs the compiler to assume that signed arithmetic
12733 overflow of addition, subtraction and multiplication wraps around
12734 using twos-complement representation. This flag enables some optimizations
12735 and disables others. This option is enabled by default for the Java
12736 front-end, as required by the Java language specification.
12739 @opindex fexceptions
12740 Enable exception handling. Generates extra code needed to propagate
12741 exceptions. For some targets, this implies GCC will generate frame
12742 unwind information for all functions, which can produce significant data
12743 size overhead, although it does not affect execution. If you do not
12744 specify this option, GCC will enable it by default for languages like
12745 C++ which normally require exception handling, and disable it for
12746 languages like C that do not normally require it. However, you may need
12747 to enable this option when compiling C code that needs to interoperate
12748 properly with exception handlers written in C++. You may also wish to
12749 disable this option if you are compiling older C++ programs that don't
12750 use exception handling.
12752 @item -fnon-call-exceptions
12753 @opindex fnon-call-exceptions
12754 Generate code that allows trapping instructions to throw exceptions.
12755 Note that this requires platform-specific runtime support that does
12756 not exist everywhere. Moreover, it only allows @emph{trapping}
12757 instructions to throw exceptions, i.e.@: memory references or floating
12758 point instructions. It does not allow exceptions to be thrown from
12759 arbitrary signal handlers such as @code{SIGALRM}.
12761 @item -funwind-tables
12762 @opindex funwind-tables
12763 Similar to @option{-fexceptions}, except that it will just generate any needed
12764 static data, but will not affect the generated code in any other way.
12765 You will normally not enable this option; instead, a language processor
12766 that needs this handling would enable it on your behalf.
12768 @item -fasynchronous-unwind-tables
12769 @opindex fasynchronous-unwind-tables
12770 Generate unwind table in dwarf2 format, if supported by target machine. The
12771 table is exact at each instruction boundary, so it can be used for stack
12772 unwinding from asynchronous events (such as debugger or garbage collector).
12774 @item -fpcc-struct-return
12775 @opindex fpcc-struct-return
12776 Return ``short'' @code{struct} and @code{union} values in memory like
12777 longer ones, rather than in registers. This convention is less
12778 efficient, but it has the advantage of allowing intercallability between
12779 GCC-compiled files and files compiled with other compilers, particularly
12780 the Portable C Compiler (pcc).
12782 The precise convention for returning structures in memory depends
12783 on the target configuration macros.
12785 Short structures and unions are those whose size and alignment match
12786 that of some integer type.
12788 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12789 switch is not binary compatible with code compiled with the
12790 @option{-freg-struct-return} switch.
12791 Use it to conform to a non-default application binary interface.
12793 @item -freg-struct-return
12794 @opindex freg-struct-return
12795 Return @code{struct} and @code{union} values in registers when possible.
12796 This is more efficient for small structures than
12797 @option{-fpcc-struct-return}.
12799 If you specify neither @option{-fpcc-struct-return} nor
12800 @option{-freg-struct-return}, GCC defaults to whichever convention is
12801 standard for the target. If there is no standard convention, GCC
12802 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12803 the principal compiler. In those cases, we can choose the standard, and
12804 we chose the more efficient register return alternative.
12806 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12807 switch is not binary compatible with code compiled with the
12808 @option{-fpcc-struct-return} switch.
12809 Use it to conform to a non-default application binary interface.
12811 @item -fshort-enums
12812 @opindex fshort-enums
12813 Allocate to an @code{enum} type only as many bytes as it needs for the
12814 declared range of possible values. Specifically, the @code{enum} type
12815 will be equivalent to the smallest integer type which has enough room.
12817 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12818 code that is not binary compatible with code generated without that switch.
12819 Use it to conform to a non-default application binary interface.
12821 @item -fshort-double
12822 @opindex fshort-double
12823 Use the same size for @code{double} as for @code{float}.
12825 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12826 code that is not binary compatible with code generated without that switch.
12827 Use it to conform to a non-default application binary interface.
12829 @item -fshort-wchar
12830 @opindex fshort-wchar
12831 Override the underlying type for @samp{wchar_t} to be @samp{short
12832 unsigned int} instead of the default for the target. This option is
12833 useful for building programs to run under WINE@.
12835 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12836 code that is not binary compatible with code generated without that switch.
12837 Use it to conform to a non-default application binary interface.
12839 @item -fshared-data
12840 @opindex fshared-data
12841 Requests that the data and non-@code{const} variables of this
12842 compilation be shared data rather than private data. The distinction
12843 makes sense only on certain operating systems, where shared data is
12844 shared between processes running the same program, while private data
12845 exists in one copy per process.
12848 @opindex fno-common
12849 In C, allocate even uninitialized global variables in the data section of the
12850 object file, rather than generating them as common blocks. This has the
12851 effect that if the same variable is declared (without @code{extern}) in
12852 two different compilations, you will get an error when you link them.
12853 The only reason this might be useful is if you wish to verify that the
12854 program will work on other systems which always work this way.
12858 Ignore the @samp{#ident} directive.
12860 @item -finhibit-size-directive
12861 @opindex finhibit-size-directive
12862 Don't output a @code{.size} assembler directive, or anything else that
12863 would cause trouble if the function is split in the middle, and the
12864 two halves are placed at locations far apart in memory. This option is
12865 used when compiling @file{crtstuff.c}; you should not need to use it
12868 @item -fverbose-asm
12869 @opindex fverbose-asm
12870 Put extra commentary information in the generated assembly code to
12871 make it more readable. This option is generally only of use to those
12872 who actually need to read the generated assembly code (perhaps while
12873 debugging the compiler itself).
12875 @option{-fno-verbose-asm}, the default, causes the
12876 extra information to be omitted and is useful when comparing two assembler
12881 @cindex global offset table
12883 Generate position-independent code (PIC) suitable for use in a shared
12884 library, if supported for the target machine. Such code accesses all
12885 constant addresses through a global offset table (GOT)@. The dynamic
12886 loader resolves the GOT entries when the program starts (the dynamic
12887 loader is not part of GCC; it is part of the operating system). If
12888 the GOT size for the linked executable exceeds a machine-specific
12889 maximum size, you get an error message from the linker indicating that
12890 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12891 instead. (These maximums are 8k on the SPARC and 32k
12892 on the m68k and RS/6000. The 386 has no such limit.)
12894 Position-independent code requires special support, and therefore works
12895 only on certain machines. For the 386, GCC supports PIC for System V
12896 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12897 position-independent.
12899 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12904 If supported for the target machine, emit position-independent code,
12905 suitable for dynamic linking and avoiding any limit on the size of the
12906 global offset table. This option makes a difference on the m68k,
12907 PowerPC and SPARC@.
12909 Position-independent code requires special support, and therefore works
12910 only on certain machines.
12912 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
12919 These options are similar to @option{-fpic} and @option{-fPIC}, but
12920 generated position independent code can be only linked into executables.
12921 Usually these options are used when @option{-pie} GCC option will be
12922 used during linking.
12924 @item -fno-jump-tables
12925 @opindex fno-jump-tables
12926 Do not use jump tables for switch statements even where it would be
12927 more efficient than other code generation strategies. This option is
12928 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12929 building code which forms part of a dynamic linker and cannot
12930 reference the address of a jump table. On some targets, jump tables
12931 do not require a GOT and this option is not needed.
12933 @item -ffixed-@var{reg}
12935 Treat the register named @var{reg} as a fixed register; generated code
12936 should never refer to it (except perhaps as a stack pointer, frame
12937 pointer or in some other fixed role).
12939 @var{reg} must be the name of a register. The register names accepted
12940 are machine-specific and are defined in the @code{REGISTER_NAMES}
12941 macro in the machine description macro file.
12943 This flag does not have a negative form, because it specifies a
12946 @item -fcall-used-@var{reg}
12947 @opindex fcall-used
12948 Treat the register named @var{reg} as an allocable register that is
12949 clobbered by function calls. It may be allocated for temporaries or
12950 variables that do not live across a call. Functions compiled this way
12951 will not save and restore the register @var{reg}.
12953 It is an error to used this flag with the frame pointer or stack pointer.
12954 Use of this flag for other registers that have fixed pervasive roles in
12955 the machine's execution model will produce disastrous results.
12957 This flag does not have a negative form, because it specifies a
12960 @item -fcall-saved-@var{reg}
12961 @opindex fcall-saved
12962 Treat the register named @var{reg} as an allocable register saved by
12963 functions. It may be allocated even for temporaries or variables that
12964 live across a call. Functions compiled this way will save and restore
12965 the register @var{reg} if they use it.
12967 It is an error to used this flag with the frame pointer or stack pointer.
12968 Use of this flag for other registers that have fixed pervasive roles in
12969 the machine's execution model will produce disastrous results.
12971 A different sort of disaster will result from the use of this flag for
12972 a register in which function values may be returned.
12974 This flag does not have a negative form, because it specifies a
12977 @item -fpack-struct[=@var{n}]
12978 @opindex fpack-struct
12979 Without a value specified, pack all structure members together without
12980 holes. When a value is specified (which must be a small power of two), pack
12981 structure members according to this value, representing the maximum
12982 alignment (that is, objects with default alignment requirements larger than
12983 this will be output potentially unaligned at the next fitting location.
12985 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12986 code that is not binary compatible with code generated without that switch.
12987 Additionally, it makes the code suboptimal.
12988 Use it to conform to a non-default application binary interface.
12990 @item -finstrument-functions
12991 @opindex finstrument-functions
12992 Generate instrumentation calls for entry and exit to functions. Just
12993 after function entry and just before function exit, the following
12994 profiling functions will be called with the address of the current
12995 function and its call site. (On some platforms,
12996 @code{__builtin_return_address} does not work beyond the current
12997 function, so the call site information may not be available to the
12998 profiling functions otherwise.)
13001 void __cyg_profile_func_enter (void *this_fn,
13003 void __cyg_profile_func_exit (void *this_fn,
13007 The first argument is the address of the start of the current function,
13008 which may be looked up exactly in the symbol table.
13010 This instrumentation is also done for functions expanded inline in other
13011 functions. The profiling calls will indicate where, conceptually, the
13012 inline function is entered and exited. This means that addressable
13013 versions of such functions must be available. If all your uses of a
13014 function are expanded inline, this may mean an additional expansion of
13015 code size. If you use @samp{extern inline} in your C code, an
13016 addressable version of such functions must be provided. (This is
13017 normally the case anyways, but if you get lucky and the optimizer always
13018 expands the functions inline, you might have gotten away without
13019 providing static copies.)
13021 A function may be given the attribute @code{no_instrument_function}, in
13022 which case this instrumentation will not be done. This can be used, for
13023 example, for the profiling functions listed above, high-priority
13024 interrupt routines, and any functions from which the profiling functions
13025 cannot safely be called (perhaps signal handlers, if the profiling
13026 routines generate output or allocate memory).
13028 @item -fstack-check
13029 @opindex fstack-check
13030 Generate code to verify that you do not go beyond the boundary of the
13031 stack. You should specify this flag if you are running in an
13032 environment with multiple threads, but only rarely need to specify it in
13033 a single-threaded environment since stack overflow is automatically
13034 detected on nearly all systems if there is only one stack.
13036 Note that this switch does not actually cause checking to be done; the
13037 operating system must do that. The switch causes generation of code
13038 to ensure that the operating system sees the stack being extended.
13040 @item -fstack-limit-register=@var{reg}
13041 @itemx -fstack-limit-symbol=@var{sym}
13042 @itemx -fno-stack-limit
13043 @opindex fstack-limit-register
13044 @opindex fstack-limit-symbol
13045 @opindex fno-stack-limit
13046 Generate code to ensure that the stack does not grow beyond a certain value,
13047 either the value of a register or the address of a symbol. If the stack
13048 would grow beyond the value, a signal is raised. For most targets,
13049 the signal is raised before the stack overruns the boundary, so
13050 it is possible to catch the signal without taking special precautions.
13052 For instance, if the stack starts at absolute address @samp{0x80000000}
13053 and grows downwards, you can use the flags
13054 @option{-fstack-limit-symbol=__stack_limit} and
13055 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13056 of 128KB@. Note that this may only work with the GNU linker.
13058 @cindex aliasing of parameters
13059 @cindex parameters, aliased
13060 @item -fargument-alias
13061 @itemx -fargument-noalias
13062 @itemx -fargument-noalias-global
13063 @opindex fargument-alias
13064 @opindex fargument-noalias
13065 @opindex fargument-noalias-global
13066 Specify the possible relationships among parameters and between
13067 parameters and global data.
13069 @option{-fargument-alias} specifies that arguments (parameters) may
13070 alias each other and may alias global storage.@*
13071 @option{-fargument-noalias} specifies that arguments do not alias
13072 each other, but may alias global storage.@*
13073 @option{-fargument-noalias-global} specifies that arguments do not
13074 alias each other and do not alias global storage.
13076 Each language will automatically use whatever option is required by
13077 the language standard. You should not need to use these options yourself.
13079 @item -fleading-underscore
13080 @opindex fleading-underscore
13081 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13082 change the way C symbols are represented in the object file. One use
13083 is to help link with legacy assembly code.
13085 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13086 generate code that is not binary compatible with code generated without that
13087 switch. Use it to conform to a non-default application binary interface.
13088 Not all targets provide complete support for this switch.
13090 @item -ftls-model=@var{model}
13091 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13092 The @var{model} argument should be one of @code{global-dynamic},
13093 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13095 The default without @option{-fpic} is @code{initial-exec}; with
13096 @option{-fpic} the default is @code{global-dynamic}.
13098 @item -fvisibility=@var{default|internal|hidden|protected}
13099 @opindex fvisibility
13100 Set the default ELF image symbol visibility to the specified option---all
13101 symbols will be marked with this unless overridden within the code.
13102 Using this feature can very substantially improve linking and
13103 load times of shared object libraries, produce more optimized
13104 code, provide near-perfect API export and prevent symbol clashes.
13105 It is @strong{strongly} recommended that you use this in any shared objects
13108 Despite the nomenclature, @code{default} always means public ie;
13109 available to be linked against from outside the shared object.
13110 @code{protected} and @code{internal} are pretty useless in real-world
13111 usage so the only other commonly used option will be @code{hidden}.
13112 The default if @option{-fvisibility} isn't specified is
13113 @code{default}, i.e., make every
13114 symbol public---this causes the same behavior as previous versions of
13117 A good explanation of the benefits offered by ensuring ELF
13118 symbols have the correct visibility is given by ``How To Write
13119 Shared Libraries'' by Ulrich Drepper (which can be found at
13120 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13121 solution made possible by this option to marking things hidden when
13122 the default is public is to make the default hidden and mark things
13123 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13124 and @code{__attribute__ ((visibility("default")))} instead of
13125 @code{__declspec(dllexport)} you get almost identical semantics with
13126 identical syntax. This is a great boon to those working with
13127 cross-platform projects.
13129 For those adding visibility support to existing code, you may find
13130 @samp{#pragma GCC visibility} of use. This works by you enclosing
13131 the declarations you wish to set visibility for with (for example)
13132 @samp{#pragma GCC visibility push(hidden)} and
13133 @samp{#pragma GCC visibility pop}.
13134 Bear in mind that symbol visibility should be viewed @strong{as
13135 part of the API interface contract} and thus all new code should
13136 always specify visibility when it is not the default ie; declarations
13137 only for use within the local DSO should @strong{always} be marked explicitly
13138 as hidden as so to avoid PLT indirection overheads---making this
13139 abundantly clear also aids readability and self-documentation of the code.
13140 Note that due to ISO C++ specification requirements, operator new and
13141 operator delete must always be of default visibility.
13143 An overview of these techniques, their benefits and how to use them
13144 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13150 @node Environment Variables
13151 @section Environment Variables Affecting GCC
13152 @cindex environment variables
13154 @c man begin ENVIRONMENT
13155 This section describes several environment variables that affect how GCC
13156 operates. Some of them work by specifying directories or prefixes to use
13157 when searching for various kinds of files. Some are used to specify other
13158 aspects of the compilation environment.
13160 Note that you can also specify places to search using options such as
13161 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13162 take precedence over places specified using environment variables, which
13163 in turn take precedence over those specified by the configuration of GCC@.
13164 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13165 GNU Compiler Collection (GCC) Internals}.
13170 @c @itemx LC_COLLATE
13172 @c @itemx LC_MONETARY
13173 @c @itemx LC_NUMERIC
13178 @c @findex LC_COLLATE
13179 @findex LC_MESSAGES
13180 @c @findex LC_MONETARY
13181 @c @findex LC_NUMERIC
13185 These environment variables control the way that GCC uses
13186 localization information that allow GCC to work with different
13187 national conventions. GCC inspects the locale categories
13188 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13189 so. These locale categories can be set to any value supported by your
13190 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13191 Kingdom encoded in UTF-8.
13193 The @env{LC_CTYPE} environment variable specifies character
13194 classification. GCC uses it to determine the character boundaries in
13195 a string; this is needed for some multibyte encodings that contain quote
13196 and escape characters that would otherwise be interpreted as a string
13199 The @env{LC_MESSAGES} environment variable specifies the language to
13200 use in diagnostic messages.
13202 If the @env{LC_ALL} environment variable is set, it overrides the value
13203 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13204 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13205 environment variable. If none of these variables are set, GCC
13206 defaults to traditional C English behavior.
13210 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13211 files. GCC uses temporary files to hold the output of one stage of
13212 compilation which is to be used as input to the next stage: for example,
13213 the output of the preprocessor, which is the input to the compiler
13216 @item GCC_EXEC_PREFIX
13217 @findex GCC_EXEC_PREFIX
13218 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13219 names of the subprograms executed by the compiler. No slash is added
13220 when this prefix is combined with the name of a subprogram, but you can
13221 specify a prefix that ends with a slash if you wish.
13223 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13224 an appropriate prefix to use based on the pathname it was invoked with.
13226 If GCC cannot find the subprogram using the specified prefix, it
13227 tries looking in the usual places for the subprogram.
13229 The default value of @env{GCC_EXEC_PREFIX} is
13230 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13231 of @code{prefix} when you ran the @file{configure} script.
13233 Other prefixes specified with @option{-B} take precedence over this prefix.
13235 This prefix is also used for finding files such as @file{crt0.o} that are
13238 In addition, the prefix is used in an unusual way in finding the
13239 directories to search for header files. For each of the standard
13240 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13241 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13242 replacing that beginning with the specified prefix to produce an
13243 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13244 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13245 These alternate directories are searched first; the standard directories
13248 @item COMPILER_PATH
13249 @findex COMPILER_PATH
13250 The value of @env{COMPILER_PATH} is a colon-separated list of
13251 directories, much like @env{PATH}. GCC tries the directories thus
13252 specified when searching for subprograms, if it can't find the
13253 subprograms using @env{GCC_EXEC_PREFIX}.
13256 @findex LIBRARY_PATH
13257 The value of @env{LIBRARY_PATH} is a colon-separated list of
13258 directories, much like @env{PATH}. When configured as a native compiler,
13259 GCC tries the directories thus specified when searching for special
13260 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13261 using GCC also uses these directories when searching for ordinary
13262 libraries for the @option{-l} option (but directories specified with
13263 @option{-L} come first).
13267 @cindex locale definition
13268 This variable is used to pass locale information to the compiler. One way in
13269 which this information is used is to determine the character set to be used
13270 when character literals, string literals and comments are parsed in C and C++.
13271 When the compiler is configured to allow multibyte characters,
13272 the following values for @env{LANG} are recognized:
13276 Recognize JIS characters.
13278 Recognize SJIS characters.
13280 Recognize EUCJP characters.
13283 If @env{LANG} is not defined, or if it has some other value, then the
13284 compiler will use mblen and mbtowc as defined by the default locale to
13285 recognize and translate multibyte characters.
13289 Some additional environments variables affect the behavior of the
13292 @include cppenv.texi
13296 @node Precompiled Headers
13297 @section Using Precompiled Headers
13298 @cindex precompiled headers
13299 @cindex speed of compilation
13301 Often large projects have many header files that are included in every
13302 source file. The time the compiler takes to process these header files
13303 over and over again can account for nearly all of the time required to
13304 build the project. To make builds faster, GCC allows users to
13305 `precompile' a header file; then, if builds can use the precompiled
13306 header file they will be much faster.
13308 To create a precompiled header file, simply compile it as you would any
13309 other file, if necessary using the @option{-x} option to make the driver
13310 treat it as a C or C++ header file. You will probably want to use a
13311 tool like @command{make} to keep the precompiled header up-to-date when
13312 the headers it contains change.
13314 A precompiled header file will be searched for when @code{#include} is
13315 seen in the compilation. As it searches for the included file
13316 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13317 compiler looks for a precompiled header in each directory just before it
13318 looks for the include file in that directory. The name searched for is
13319 the name specified in the @code{#include} with @samp{.gch} appended. If
13320 the precompiled header file can't be used, it is ignored.
13322 For instance, if you have @code{#include "all.h"}, and you have
13323 @file{all.h.gch} in the same directory as @file{all.h}, then the
13324 precompiled header file will be used if possible, and the original
13325 header will be used otherwise.
13327 Alternatively, you might decide to put the precompiled header file in a
13328 directory and use @option{-I} to ensure that directory is searched
13329 before (or instead of) the directory containing the original header.
13330 Then, if you want to check that the precompiled header file is always
13331 used, you can put a file of the same name as the original header in this
13332 directory containing an @code{#error} command.
13334 This also works with @option{-include}. So yet another way to use
13335 precompiled headers, good for projects not designed with precompiled
13336 header files in mind, is to simply take most of the header files used by
13337 a project, include them from another header file, precompile that header
13338 file, and @option{-include} the precompiled header. If the header files
13339 have guards against multiple inclusion, they will be skipped because
13340 they've already been included (in the precompiled header).
13342 If you need to precompile the same header file for different
13343 languages, targets, or compiler options, you can instead make a
13344 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13345 header in the directory, perhaps using @option{-o}. It doesn't matter
13346 what you call the files in the directory, every precompiled header in
13347 the directory will be considered. The first precompiled header
13348 encountered in the directory that is valid for this compilation will
13349 be used; they're searched in no particular order.
13351 There are many other possibilities, limited only by your imagination,
13352 good sense, and the constraints of your build system.
13354 A precompiled header file can be used only when these conditions apply:
13358 Only one precompiled header can be used in a particular compilation.
13361 A precompiled header can't be used once the first C token is seen. You
13362 can have preprocessor directives before a precompiled header; you can
13363 even include a precompiled header from inside another header, so long as
13364 there are no C tokens before the @code{#include}.
13367 The precompiled header file must be produced for the same language as
13368 the current compilation. You can't use a C precompiled header for a C++
13372 The precompiled header file must have been produced by the same compiler
13373 binary as the current compilation is using.
13376 Any macros defined before the precompiled header is included must
13377 either be defined in the same way as when the precompiled header was
13378 generated, or must not affect the precompiled header, which usually
13379 means that they don't appear in the precompiled header at all.
13381 The @option{-D} option is one way to define a macro before a
13382 precompiled header is included; using a @code{#define} can also do it.
13383 There are also some options that define macros implicitly, like
13384 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13387 @item If debugging information is output when using the precompiled
13388 header, using @option{-g} or similar, the same kind of debugging information
13389 must have been output when building the precompiled header. However,
13390 a precompiled header built using @option{-g} can be used in a compilation
13391 when no debugging information is being output.
13393 @item The same @option{-m} options must generally be used when building
13394 and using the precompiled header. @xref{Submodel Options},
13395 for any cases where this rule is relaxed.
13397 @item Each of the following options must be the same when building and using
13398 the precompiled header:
13400 @gccoptlist{-fexceptions -funit-at-a-time}
13403 Some other command-line options starting with @option{-f},
13404 @option{-p}, or @option{-O} must be defined in the same way as when
13405 the precompiled header was generated. At present, it's not clear
13406 which options are safe to change and which are not; the safest choice
13407 is to use exactly the same options when generating and using the
13408 precompiled header. The following are known to be safe:
13410 @gccoptlist{-fmessage-length= -fpreprocessed
13411 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13412 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13417 For all of these except the last, the compiler will automatically
13418 ignore the precompiled header if the conditions aren't met. If you
13419 find an option combination that doesn't work and doesn't cause the
13420 precompiled header to be ignored, please consider filing a bug report,
13423 If you do use differing options when generating and using the
13424 precompiled header, the actual behavior will be a mixture of the
13425 behavior for the options. For instance, if you use @option{-g} to
13426 generate the precompiled header but not when using it, you may or may
13427 not get debugging information for routines in the precompiled header.
13429 @node Running Protoize
13430 @section Running Protoize
13432 The program @code{protoize} is an optional part of GCC@. You can use
13433 it to add prototypes to a program, thus converting the program to ISO
13434 C in one respect. The companion program @code{unprotoize} does the
13435 reverse: it removes argument types from any prototypes that are found.
13437 When you run these programs, you must specify a set of source files as
13438 command line arguments. The conversion programs start out by compiling
13439 these files to see what functions they define. The information gathered
13440 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13442 After scanning comes actual conversion. The specified files are all
13443 eligible to be converted; any files they include (whether sources or
13444 just headers) are eligible as well.
13446 But not all the eligible files are converted. By default,
13447 @code{protoize} and @code{unprotoize} convert only source and header
13448 files in the current directory. You can specify additional directories
13449 whose files should be converted with the @option{-d @var{directory}}
13450 option. You can also specify particular files to exclude with the
13451 @option{-x @var{file}} option. A file is converted if it is eligible, its
13452 directory name matches one of the specified directory names, and its
13453 name within the directory has not been excluded.
13455 Basic conversion with @code{protoize} consists of rewriting most
13456 function definitions and function declarations to specify the types of
13457 the arguments. The only ones not rewritten are those for varargs
13460 @code{protoize} optionally inserts prototype declarations at the
13461 beginning of the source file, to make them available for any calls that
13462 precede the function's definition. Or it can insert prototype
13463 declarations with block scope in the blocks where undeclared functions
13466 Basic conversion with @code{unprotoize} consists of rewriting most
13467 function declarations to remove any argument types, and rewriting
13468 function definitions to the old-style pre-ISO form.
13470 Both conversion programs print a warning for any function declaration or
13471 definition that they can't convert. You can suppress these warnings
13474 The output from @code{protoize} or @code{unprotoize} replaces the
13475 original source file. The original file is renamed to a name ending
13476 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13477 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13478 for DOS) file already exists, then the source file is simply discarded.
13480 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13481 scan the program and collect information about the functions it uses.
13482 So neither of these programs will work until GCC is installed.
13484 Here is a table of the options you can use with @code{protoize} and
13485 @code{unprotoize}. Each option works with both programs unless
13489 @item -B @var{directory}
13490 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13491 usual directory (normally @file{/usr/local/lib}). This file contains
13492 prototype information about standard system functions. This option
13493 applies only to @code{protoize}.
13495 @item -c @var{compilation-options}
13496 Use @var{compilation-options} as the options when running @command{gcc} to
13497 produce the @samp{.X} files. The special option @option{-aux-info} is
13498 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13500 Note that the compilation options must be given as a single argument to
13501 @code{protoize} or @code{unprotoize}. If you want to specify several
13502 @command{gcc} options, you must quote the entire set of compilation options
13503 to make them a single word in the shell.
13505 There are certain @command{gcc} arguments that you cannot use, because they
13506 would produce the wrong kind of output. These include @option{-g},
13507 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13508 the @var{compilation-options}, they are ignored.
13511 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13512 systems) instead of @samp{.c}. This is convenient if you are converting
13513 a C program to C++. This option applies only to @code{protoize}.
13516 Add explicit global declarations. This means inserting explicit
13517 declarations at the beginning of each source file for each function
13518 that is called in the file and was not declared. These declarations
13519 precede the first function definition that contains a call to an
13520 undeclared function. This option applies only to @code{protoize}.
13522 @item -i @var{string}
13523 Indent old-style parameter declarations with the string @var{string}.
13524 This option applies only to @code{protoize}.
13526 @code{unprotoize} converts prototyped function definitions to old-style
13527 function definitions, where the arguments are declared between the
13528 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13529 uses five spaces as the indentation. If you want to indent with just
13530 one space instead, use @option{-i " "}.
13533 Keep the @samp{.X} files. Normally, they are deleted after conversion
13537 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13538 a prototype declaration for each function in each block which calls the
13539 function without any declaration. This option applies only to
13543 Make no real changes. This mode just prints information about the conversions
13544 that would have been done without @option{-n}.
13547 Make no @samp{.save} files. The original files are simply deleted.
13548 Use this option with caution.
13550 @item -p @var{program}
13551 Use the program @var{program} as the compiler. Normally, the name
13552 @file{gcc} is used.
13555 Work quietly. Most warnings are suppressed.
13558 Print the version number, just like @option{-v} for @command{gcc}.
13561 If you need special compiler options to compile one of your program's
13562 source files, then you should generate that file's @samp{.X} file
13563 specially, by running @command{gcc} on that source file with the
13564 appropriate options and the option @option{-aux-info}. Then run
13565 @code{protoize} on the entire set of files. @code{protoize} will use
13566 the existing @samp{.X} file because it is newer than the source file.
13570 gcc -Dfoo=bar file1.c -aux-info file1.X
13575 You need to include the special files along with the rest in the
13576 @code{protoize} command, even though their @samp{.X} files already
13577 exist, because otherwise they won't get converted.
13579 @xref{Protoize Caveats}, for more information on how to use
13580 @code{protoize} successfully.