1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
16 Permission is granted to copy, distribute and/or modify this document
17 under the terms of the GNU Free Documentation License, Version 1.2 or
18 any later version published by the Free Software Foundation; with the
19 Invariant Sections being ``GNU General Public License'' and ``Funding
20 Free Software'', the Front-Cover texts being (a) (see below), and with
21 the Back-Cover Texts being (b) (see below). A copy of the license is
22 included in the gfdl(7) man page.
24 (a) The FSF's Front-Cover Text is:
28 (b) The FSF's Back-Cover Text is:
30 You have freedom to copy and modify this GNU Manual, like GNU
31 software. Copies published by the Free Software Foundation raise
32 funds for GNU development.
34 @c Set file name and title for the man page.
36 @settitle GNU project C and C++ compiler
38 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
39 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
40 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
41 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
42 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
43 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
44 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
46 Only the most useful options are listed here; see below for the
47 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
50 gpl(7), gfdl(7), fsf-funding(7),
51 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
52 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
53 @file{ld}, @file{binutils} and @file{gdb}.
56 For instructions on reporting bugs, see
57 @w{@uref{http://gcc.gnu.org/bugs.html}}.
60 See the Info entry for @command{gcc}, or
61 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
62 for contributors to GCC@.
67 @chapter GCC Command Options
68 @cindex GCC command options
69 @cindex command options
70 @cindex options, GCC command
72 @c man begin DESCRIPTION
73 When you invoke GCC, it normally does preprocessing, compilation,
74 assembly and linking. The ``overall options'' allow you to stop this
75 process at an intermediate stage. For example, the @option{-c} option
76 says not to run the linker. Then the output consists of object files
77 output by the assembler.
79 Other options are passed on to one stage of processing. Some options
80 control the preprocessor and others the compiler itself. Yet other
81 options control the assembler and linker; most of these are not
82 documented here, since you rarely need to use any of them.
84 @cindex C compilation options
85 Most of the command line options that you can use with GCC are useful
86 for C programs; when an option is only useful with another language
87 (usually C++), the explanation says so explicitly. If the description
88 for a particular option does not mention a source language, you can use
89 that option with all supported languages.
91 @cindex C++ compilation options
92 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
93 options for compiling C++ programs.
95 @cindex grouping options
96 @cindex options, grouping
97 The @command{gcc} program accepts options and file names as operands. Many
98 options have multi-letter names; therefore multiple single-letter options
99 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
102 @cindex order of options
103 @cindex options, order
104 You can mix options and other arguments. For the most part, the order
105 you use doesn't matter. Order does matter when you use several options
106 of the same kind; for example, if you specify @option{-L} more than once,
107 the directories are searched in the order specified.
109 Many options have long names starting with @samp{-f} or with
110 @samp{-W}---for example,
111 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
112 these have both positive and negative forms; the negative form of
113 @option{-ffoo} would be @option{-fno-foo}. This manual documents
114 only one of these two forms, whichever one is not the default.
118 @xref{Option Index}, for an index to GCC's options.
121 * Option Summary:: Brief list of all options, without explanations.
122 * Overall Options:: Controlling the kind of output:
123 an executable, object files, assembler files,
124 or preprocessed source.
125 * Invoking G++:: Compiling C++ programs.
126 * C Dialect Options:: Controlling the variant of C language compiled.
127 * C++ Dialect Options:: Variations on C++.
128 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
130 * Language Independent Options:: Controlling how diagnostics should be
132 * Warning Options:: How picky should the compiler be?
133 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
134 * Optimize Options:: How much optimization?
135 * Preprocessor Options:: Controlling header files and macro definitions.
136 Also, getting dependency information for Make.
137 * Assembler Options:: Passing options to the assembler.
138 * Link Options:: Specifying libraries and so on.
139 * Directory Options:: Where to find header files and libraries.
140 Where to find the compiler executable files.
141 * Spec Files:: How to pass switches to sub-processes.
142 * Target Options:: Running a cross-compiler, or an old version of GCC.
143 * Submodel Options:: Specifying minor hardware or convention variations,
144 such as 68010 vs 68020.
145 * Code Gen Options:: Specifying conventions for function calls, data layout
147 * Environment Variables:: Env vars that affect GCC.
148 * Precompiled Headers:: Compiling a header once, and using it many times.
149 * Running Protoize:: Automatically adding or removing function prototypes.
155 @section Option Summary
157 Here is a summary of all the options, grouped by type. Explanations are
158 in the following sections.
161 @item Overall Options
162 @xref{Overall Options,,Options Controlling the Kind of Output}.
163 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
164 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
165 --version @@@var{file}}
167 @item C Language Options
168 @xref{C Dialect Options,,Options Controlling C Dialect}.
169 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
170 -aux-info @var{filename} @gol
171 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
172 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
173 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
174 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
175 -fsigned-bitfields -fsigned-char @gol
176 -funsigned-bitfields -funsigned-char}
178 @item C++ Language Options
179 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
180 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
181 -fconserve-space -ffriend-injection @gol
182 -fno-elide-constructors @gol
183 -fno-enforce-eh-specs @gol
184 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
185 -fno-implicit-templates @gol
186 -fno-implicit-inline-templates @gol
187 -fno-implement-inlines -fms-extensions @gol
188 -fno-nonansi-builtins -fno-operator-names @gol
189 -fno-optional-diags -fpermissive @gol
190 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
191 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
192 -fno-default-inline -fvisibility-inlines-hidden @gol
193 -Wabi -Wctor-dtor-privacy @gol
194 -Wnon-virtual-dtor -Wreorder @gol
195 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
196 -Wno-non-template-friend -Wold-style-cast @gol
197 -Woverloaded-virtual -Wno-pmf-conversions @gol
200 @item Objective-C and Objective-C++ Language Options
201 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
202 Objective-C and Objective-C++ Dialects}.
203 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
204 -fgnu-runtime -fnext-runtime @gol
205 -fno-nil-receivers @gol
206 -fobjc-call-cxx-cdtors @gol
207 -fobjc-direct-dispatch @gol
208 -fobjc-exceptions @gol
210 -freplace-objc-classes @gol
213 -Wassign-intercept @gol
214 -Wno-protocol -Wselector @gol
215 -Wstrict-selector-match @gol
216 -Wundeclared-selector}
218 @item Language Independent Options
219 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
220 @gccoptlist{-fmessage-length=@var{n} @gol
221 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
222 -fdiagnostics-show-option}
224 @item Warning Options
225 @xref{Warning Options,,Options to Request or Suppress Warnings}.
226 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
227 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
228 -Wno-attributes -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
229 -Wchar-subscripts -Wclobbered -Wcomment @gol
230 -Wconversion -Wcoverage-mismatch -Wno-deprecated-declarations @gol
231 -Wdisabled-optimization -Wno-div-by-zero @gol
232 -Wempty-body -Wno-endif-labels @gol
233 -Werror -Werror=* @gol
234 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
235 -Wno-format-extra-args -Wformat-nonliteral @gol
236 -Wformat-security -Wformat-y2k @gol
237 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
238 -Wimport -Wno-import -Winit-self -Winline @gol
239 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
240 -Winvalid-pch -Wlarger-than-@var{len} -Wunsafe-loop-optimizations @gol
241 -Wlogical-op -Wlong-long @gol
242 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
243 -Wmissing-format-attribute -Wmissing-include-dirs @gol
244 -Wmissing-noreturn @gol
245 -Wno-multichar -Wnonnull -Wno-overflow @gol
246 -Woverlength-strings -Wpacked -Wpadded @gol
247 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
248 -Wredundant-decls @gol
249 -Wreturn-type -Wsequence-point -Wshadow @gol
250 -Wsign-compare -Wstack-protector @gol
251 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
252 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
253 -Wswitch -Wswitch-default -Wswitch-enum @gol
254 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
255 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
256 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
257 -Wunused-value -Wunused-variable @gol
258 -Wvariadic-macros -Wvla @gol
259 -Wvolatile-register-var -Wwrite-strings}
261 @item C-only Warning Options
262 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
263 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
264 -Wold-style-declaration -Wold-style-definition @gol
265 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
266 -Wdeclaration-after-statement -Wpointer-sign}
268 @item Debugging Options
269 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
270 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
271 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
272 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
273 -fdump-ipa-all -fdump-ipa-cgraph @gol
275 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
277 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
280 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
283 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-nrv -fdump-tree-vect @gol
289 -fdump-tree-sink @gol
290 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-salias @gol
292 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
293 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
294 -ftree-vectorizer-verbose=@var{n} @gol
295 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
296 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
297 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
298 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
299 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
300 -ftest-coverage -ftime-report -fvar-tracking @gol
301 -g -g@var{level} -gcoff -gdwarf-2 @gol
302 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
303 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
304 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
305 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
306 -print-multi-directory -print-multi-lib @gol
307 -print-prog-name=@var{program} -print-search-dirs -Q @gol
308 -print-sysroot-headers-suffix @gol
311 @item Optimization Options
312 @xref{Optimize Options,,Options that Control Optimization}.
313 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
314 -falign-labels=@var{n} -falign-loops=@var{n} @gol
315 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
316 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
317 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
318 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
319 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
320 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
321 -fexpensive-optimizations -ffast-math -ffloat-store @gol
322 -fforce-addr -fforward-propagate -ffunction-sections @gol
323 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
324 -fcrossjumping -fif-conversion -fif-conversion2 @gol
325 -finline-functions -finline-functions-called-once @gol
326 -finline-limit=@var{n} -fkeep-inline-functions @gol
327 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
328 -fmodulo-sched -fno-branch-count-reg @gol
329 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
330 -fno-function-cse -fno-guess-branch-probability @gol
331 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
332 -funsafe-math-optimizations -funsafe-loop-optimizations @gol
333 -ffinite-math-only -fno-signed-zeros @gol
334 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
335 -fomit-frame-pointer -foptimize-register-move @gol
336 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
337 -fprofile-generate -fprofile-use @gol
338 -fregmove -frename-registers @gol
339 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
340 -frerun-cse-after-loop @gol
341 -frounding-math -frtl-abstract-sequences @gol
342 -fschedule-insns -fschedule-insns2 @gol
343 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
344 -fsched-spec-load-dangerous @gol
345 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
346 -fsched2-use-superblocks @gol
347 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
348 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
349 -fno-split-wide-types -fstack-protector -fstack-protector-all @gol
350 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
351 -funroll-all-loops -funroll-loops -fpeel-loops @gol
352 -fsplit-ivs-in-unroller -funswitch-loops @gol
353 -fvariable-expansion-in-unroller @gol
354 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
355 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
356 -fcheck-data-deps @gol
357 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
358 -ftree-ch -ftree-sra -ftree-ter -ftree-fre -ftree-vectorize @gol
359 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
360 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
361 --param @var{name}=@var{value}
362 -O -O0 -O1 -O2 -O3 -Os}
364 @item Preprocessor Options
365 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
366 @gccoptlist{-A@var{question}=@var{answer} @gol
367 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
368 -C -dD -dI -dM -dN @gol
369 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
370 -idirafter @var{dir} @gol
371 -include @var{file} -imacros @var{file} @gol
372 -iprefix @var{file} -iwithprefix @var{dir} @gol
373 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
374 -imultilib @var{dir} -isysroot @var{dir} @gol
375 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
376 -P -fworking-directory -remap @gol
377 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
378 -Xpreprocessor @var{option}}
380 @item Assembler Option
381 @xref{Assembler Options,,Passing Options to the Assembler}.
382 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
385 @xref{Link Options,,Options for Linking}.
386 @gccoptlist{@var{object-file-name} -l@var{library} @gol
387 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
388 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
389 -Wl,@var{option} -Xlinker @var{option} @gol
392 @item Directory Options
393 @xref{Directory Options,,Options for Directory Search}.
394 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
395 -specs=@var{file} -I- --sysroot=@var{dir}}
398 @c I wrote this xref this way to avoid overfull hbox. -- rms
399 @xref{Target Options}.
400 @gccoptlist{-V @var{version} -b @var{machine}}
402 @item Machine Dependent Options
403 @xref{Submodel Options,,Hardware Models and Configurations}.
404 @c This list is ordered alphanumerically by subsection name.
405 @c Try and put the significant identifier (CPU or system) first,
406 @c so users have a clue at guessing where the ones they want will be.
409 @gccoptlist{-EB -EL @gol
410 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
411 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
414 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
415 -mabi=@var{name} @gol
416 -mapcs-stack-check -mno-apcs-stack-check @gol
417 -mapcs-float -mno-apcs-float @gol
418 -mapcs-reentrant -mno-apcs-reentrant @gol
419 -msched-prolog -mno-sched-prolog @gol
420 -mlittle-endian -mbig-endian -mwords-little-endian @gol
421 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
422 -mthumb-interwork -mno-thumb-interwork @gol
423 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
424 -mstructure-size-boundary=@var{n} @gol
425 -mabort-on-noreturn @gol
426 -mlong-calls -mno-long-calls @gol
427 -msingle-pic-base -mno-single-pic-base @gol
428 -mpic-register=@var{reg} @gol
429 -mnop-fun-dllimport @gol
430 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
431 -mpoke-function-name @gol
433 -mtpcs-frame -mtpcs-leaf-frame @gol
434 -mcaller-super-interworking -mcallee-super-interworking @gol
438 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
439 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
441 @emph{Blackfin Options}
442 @gccoptlist{-mcpu=@var{cpu} -msim -momit-leaf-frame-pointer @gol
443 -mno-omit-leaf-frame-pointer -mspecld-anomaly -mno-specld-anomaly @gol
444 -mcsync-anomaly -mno-csync-anomaly -mlow-64k -mno-low64k @gol
445 -mstack-check-l1 -mid-shared-library -mno-id-shared-library @gol
446 -mshared-library-id=@var{n} -mleaf-id-shared-library @gol
447 -mno-leaf-id-shared-library -msep-data -mno-sep-data -mlong-calls @gol
451 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
452 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
453 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
454 -mstack-align -mdata-align -mconst-align @gol
455 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
456 -melf -maout -melinux -mlinux -sim -sim2 @gol
457 -mmul-bug-workaround -mno-mul-bug-workaround}
460 @gccoptlist{-mmac -mpush-args}
462 @emph{Darwin Options}
463 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
464 -arch_only -bind_at_load -bundle -bundle_loader @gol
465 -client_name -compatibility_version -current_version @gol
467 -dependency-file -dylib_file -dylinker_install_name @gol
468 -dynamic -dynamiclib -exported_symbols_list @gol
469 -filelist -flat_namespace -force_cpusubtype_ALL @gol
470 -force_flat_namespace -headerpad_max_install_names @gol
472 -image_base -init -install_name -keep_private_externs @gol
473 -multi_module -multiply_defined -multiply_defined_unused @gol
474 -noall_load -no_dead_strip_inits_and_terms @gol
475 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
476 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
477 -private_bundle -read_only_relocs -sectalign @gol
478 -sectobjectsymbols -whyload -seg1addr @gol
479 -sectcreate -sectobjectsymbols -sectorder @gol
480 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
481 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
482 -segprot -segs_read_only_addr -segs_read_write_addr @gol
483 -single_module -static -sub_library -sub_umbrella @gol
484 -twolevel_namespace -umbrella -undefined @gol
485 -unexported_symbols_list -weak_reference_mismatches @gol
486 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
487 -mkernel -mone-byte-bool}
489 @emph{DEC Alpha Options}
490 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
491 -mieee -mieee-with-inexact -mieee-conformant @gol
492 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
493 -mtrap-precision=@var{mode} -mbuild-constants @gol
494 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
495 -mbwx -mmax -mfix -mcix @gol
496 -mfloat-vax -mfloat-ieee @gol
497 -mexplicit-relocs -msmall-data -mlarge-data @gol
498 -msmall-text -mlarge-text @gol
499 -mmemory-latency=@var{time}}
501 @emph{DEC Alpha/VMS Options}
502 @gccoptlist{-mvms-return-codes}
505 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
506 -mhard-float -msoft-float @gol
507 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
508 -mdouble -mno-double @gol
509 -mmedia -mno-media -mmuladd -mno-muladd @gol
510 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
511 -mlinked-fp -mlong-calls -malign-labels @gol
512 -mlibrary-pic -macc-4 -macc-8 @gol
513 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
514 -moptimize-membar -mno-optimize-membar @gol
515 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
516 -mvliw-branch -mno-vliw-branch @gol
517 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
518 -mno-nested-cond-exec -mtomcat-stats @gol
522 @emph{GNU/Linux Options}
523 @gccoptlist{-muclibc}
525 @emph{H8/300 Options}
526 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
529 @gccoptlist{-march=@var{architecture-type} @gol
530 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
531 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
532 -mfixed-range=@var{register-range} @gol
533 -mjump-in-delay -mlinker-opt -mlong-calls @gol
534 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
535 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
536 -mno-jump-in-delay -mno-long-load-store @gol
537 -mno-portable-runtime -mno-soft-float @gol
538 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
539 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
540 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
541 -munix=@var{unix-std} -nolibdld -static -threads}
543 @emph{i386 and x86-64 Options}
544 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
545 -mfpmath=@var{unit} @gol
546 -masm=@var{dialect} -mno-fancy-math-387 @gol
547 -mno-fp-ret-in-387 -msoft-float @gol
548 -mno-wide-multiply -mrtd -malign-double @gol
549 -mpreferred-stack-boundary=@var{num} -mcx16 -msahf @gol
550 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm @gol
551 -mthreads -mno-align-stringops -minline-all-stringops @gol
552 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
553 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
554 -mpc32 -mpc64 -mpc80 mstackrealign @gol
555 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
556 -mcmodel=@var{code-model} @gol
557 -m32 -m64 -mlarge-data-threshold=@var{num}}
560 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
561 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
562 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
563 -minline-float-divide-max-throughput @gol
564 -minline-int-divide-min-latency @gol
565 -minline-int-divide-max-throughput @gol
566 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
567 -mno-dwarf2-asm -mearly-stop-bits @gol
568 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
569 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
570 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
571 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
572 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
573 -mno-sched-prefer-non-data-spec-insns @gol
574 -mno-sched-prefer-non-control-spec-insns @gol
575 -mno-sched-count-spec-in-critical-path}
577 @emph{M32R/D Options}
578 @gccoptlist{-m32r2 -m32rx -m32r @gol
580 -malign-loops -mno-align-loops @gol
581 -missue-rate=@var{number} @gol
582 -mbranch-cost=@var{number} @gol
583 -mmodel=@var{code-size-model-type} @gol
584 -msdata=@var{sdata-type} @gol
585 -mno-flush-func -mflush-func=@var{name} @gol
586 -mno-flush-trap -mflush-trap=@var{number} @gol
590 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
592 @emph{M680x0 Options}
593 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
594 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
595 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
596 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
597 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
598 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
599 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
600 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
602 @emph{M68hc1x Options}
603 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
604 -mauto-incdec -minmax -mlong-calls -mshort @gol
605 -msoft-reg-count=@var{count}}
608 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
609 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
610 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
611 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
612 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
615 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
616 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
617 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
618 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
619 -mfp32 -mfp64 -mhard-float -msoft-float @gol
620 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
621 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
622 -mips3d -mno-mips3d -mmt -mno-mt @gol
623 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
624 -G@var{num} -membedded-data -mno-embedded-data @gol
625 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
626 -msplit-addresses -mno-split-addresses @gol
627 -mexplicit-relocs -mno-explicit-relocs @gol
628 -mcheck-zero-division -mno-check-zero-division @gol
629 -mdivide-traps -mdivide-breaks @gol
630 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
631 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
632 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
633 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
634 -mfix-sb1 -mno-fix-sb1 @gol
635 -mflush-func=@var{func} -mno-flush-func @gol
636 -mbranch-likely -mno-branch-likely @gol
637 -mfp-exceptions -mno-fp-exceptions @gol
638 -mvr4130-align -mno-vr4130-align}
641 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
642 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
643 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
644 -mno-base-addresses -msingle-exit -mno-single-exit}
646 @emph{MN10300 Options}
647 @gccoptlist{-mmult-bug -mno-mult-bug @gol
648 -mam33 -mno-am33 @gol
649 -mam33-2 -mno-am33-2 @gol
650 -mreturn-pointer-on-d0 @gol
654 @gccoptlist{-mno-crt0 -mbacc -msim @gol
655 -march=@var{cpu-type} }
657 @emph{PDP-11 Options}
658 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
659 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
660 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
661 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
662 -mbranch-expensive -mbranch-cheap @gol
663 -msplit -mno-split -munix-asm -mdec-asm}
665 @emph{PowerPC Options}
666 See RS/6000 and PowerPC Options.
668 @emph{RS/6000 and PowerPC Options}
669 @gccoptlist{-mcpu=@var{cpu-type} @gol
670 -mtune=@var{cpu-type} @gol
671 -mpower -mno-power -mpower2 -mno-power2 @gol
672 -mpowerpc -mpowerpc64 -mno-powerpc @gol
673 -maltivec -mno-altivec @gol
674 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
675 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
676 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
677 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mdfp -mno-dfp @gol
678 -mnew-mnemonics -mold-mnemonics @gol
679 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
680 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
681 -malign-power -malign-natural @gol
682 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
683 -mstring -mno-string -mupdate -mno-update @gol
684 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
685 -mstrict-align -mno-strict-align -mrelocatable @gol
686 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
687 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
688 -mdynamic-no-pic -maltivec -mswdiv @gol
689 -mprioritize-restricted-insns=@var{priority} @gol
690 -msched-costly-dep=@var{dependence_type} @gol
691 -minsert-sched-nops=@var{scheme} @gol
692 -mcall-sysv -mcall-netbsd @gol
693 -maix-struct-return -msvr4-struct-return @gol
694 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
695 -misel -mno-isel @gol
696 -misel=yes -misel=no @gol
698 -mspe=yes -mspe=no @gol
699 -mvrsave -mno-vrsave @gol
700 -mmulhw -mno-mulhw @gol
701 -mdlmzb -mno-dlmzb @gol
702 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
703 -mprototype -mno-prototype @gol
704 -msim -mmvme -mads -myellowknife -memb -msdata @gol
705 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
707 @emph{S/390 and zSeries Options}
708 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
709 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
710 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
711 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
712 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
713 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
714 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
717 @gccoptlist{-meb -mel @gol
721 -mscore5 -mscore5u -mscore7 -mscore7d}
724 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
725 -m4-nofpu -m4-single-only -m4-single -m4 @gol
726 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
727 -m5-64media -m5-64media-nofpu @gol
728 -m5-32media -m5-32media-nofpu @gol
729 -m5-compact -m5-compact-nofpu @gol
730 -mb -ml -mdalign -mrelax @gol
731 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
732 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
733 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
734 -mdivsi3_libfunc=@var{name} @gol
735 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
739 @gccoptlist{-mcpu=@var{cpu-type} @gol
740 -mtune=@var{cpu-type} @gol
741 -mcmodel=@var{code-model} @gol
742 -m32 -m64 -mapp-regs -mno-app-regs @gol
743 -mfaster-structs -mno-faster-structs @gol
744 -mfpu -mno-fpu -mhard-float -msoft-float @gol
745 -mhard-quad-float -msoft-quad-float @gol
746 -mimpure-text -mno-impure-text -mlittle-endian @gol
747 -mstack-bias -mno-stack-bias @gol
748 -munaligned-doubles -mno-unaligned-doubles @gol
749 -mv8plus -mno-v8plus -mvis -mno-vis
750 -threads -pthreads -pthread}
753 @gccoptlist{-mwarn-reloc -merror-reloc @gol
754 -msafe-dma -munsafe-dma @gol
756 -msmall-mem -mlarge-mem -mstdmain @gol
757 -mfixed-range=@var{register-range}}
759 @emph{System V Options}
760 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
762 @emph{TMS320C3x/C4x Options}
763 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
764 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
765 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
766 -mparallel-insns -mparallel-mpy -mpreserve-float}
769 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
770 -mprolog-function -mno-prolog-function -mspace @gol
771 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
772 -mapp-regs -mno-app-regs @gol
773 -mdisable-callt -mno-disable-callt @gol
779 @gccoptlist{-mg -mgnu -munix}
781 @emph{VxWorks Options}
782 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
783 -Xbind-lazy -Xbind-now}
785 @emph{x86-64 Options}
786 See i386 and x86-64 Options.
788 @emph{Xstormy16 Options}
791 @emph{Xtensa Options}
792 @gccoptlist{-mconst16 -mno-const16 @gol
793 -mfused-madd -mno-fused-madd @gol
794 -mtext-section-literals -mno-text-section-literals @gol
795 -mtarget-align -mno-target-align @gol
796 -mlongcalls -mno-longcalls}
798 @emph{zSeries Options}
799 See S/390 and zSeries Options.
801 @item Code Generation Options
802 @xref{Code Gen Options,,Options for Code Generation Conventions}.
803 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
804 -ffixed-@var{reg} -fexceptions @gol
805 -fnon-call-exceptions -funwind-tables @gol
806 -fasynchronous-unwind-tables @gol
807 -finhibit-size-directive -finstrument-functions @gol
808 -fno-common -fno-ident @gol
809 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
810 -fno-jump-tables @gol
811 -frecord-gcc-switches @gol
812 -freg-struct-return -fshort-enums @gol
813 -fshort-double -fshort-wchar @gol
814 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
815 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
816 -fargument-alias -fargument-noalias @gol
817 -fargument-noalias-global -fargument-noalias-anything
818 -fleading-underscore -ftls-model=@var{model} @gol
819 -ftrapv -fwrapv -fbounds-check @gol
824 * Overall Options:: Controlling the kind of output:
825 an executable, object files, assembler files,
826 or preprocessed source.
827 * C Dialect Options:: Controlling the variant of C language compiled.
828 * C++ Dialect Options:: Variations on C++.
829 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
831 * Language Independent Options:: Controlling how diagnostics should be
833 * Warning Options:: How picky should the compiler be?
834 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
835 * Optimize Options:: How much optimization?
836 * Preprocessor Options:: Controlling header files and macro definitions.
837 Also, getting dependency information for Make.
838 * Assembler Options:: Passing options to the assembler.
839 * Link Options:: Specifying libraries and so on.
840 * Directory Options:: Where to find header files and libraries.
841 Where to find the compiler executable files.
842 * Spec Files:: How to pass switches to sub-processes.
843 * Target Options:: Running a cross-compiler, or an old version of GCC.
846 @node Overall Options
847 @section Options Controlling the Kind of Output
849 Compilation can involve up to four stages: preprocessing, compilation
850 proper, assembly and linking, always in that order. GCC is capable of
851 preprocessing and compiling several files either into several
852 assembler input files, or into one assembler input file; then each
853 assembler input file produces an object file, and linking combines all
854 the object files (those newly compiled, and those specified as input)
855 into an executable file.
857 @cindex file name suffix
858 For any given input file, the file name suffix determines what kind of
863 C source code which must be preprocessed.
866 C source code which should not be preprocessed.
869 C++ source code which should not be preprocessed.
872 Objective-C source code. Note that you must link with the @file{libobjc}
873 library to make an Objective-C program work.
876 Objective-C source code which should not be preprocessed.
880 Objective-C++ source code. Note that you must link with the @file{libobjc}
881 library to make an Objective-C++ program work. Note that @samp{.M} refers
882 to a literal capital M@.
885 Objective-C++ source code which should not be preprocessed.
888 C, C++, Objective-C or Objective-C++ header file to be turned into a
893 @itemx @var{file}.cxx
894 @itemx @var{file}.cpp
895 @itemx @var{file}.CPP
896 @itemx @var{file}.c++
898 C++ source code which must be preprocessed. Note that in @samp{.cxx},
899 the last two letters must both be literally @samp{x}. Likewise,
900 @samp{.C} refers to a literal capital C@.
904 Objective-C++ source code which must be preprocessed.
907 Objective-C++ source code which should not be preprocessed.
911 C++ header file to be turned into a precompiled header.
914 @itemx @var{file}.for
915 @itemx @var{file}.FOR
916 Fixed form Fortran source code which should not be preprocessed.
919 @itemx @var{file}.fpp
920 @itemx @var{file}.FPP
921 Fixed form Fortran source code which must be preprocessed (with the traditional
925 @itemx @var{file}.f95
926 Free form Fortran source code which should not be preprocessed.
929 @itemx @var{file}.F95
930 Free form Fortran source code which must be preprocessed (with the
931 traditional preprocessor).
933 @c FIXME: Descriptions of Java file types.
940 Ada source code file which contains a library unit declaration (a
941 declaration of a package, subprogram, or generic, or a generic
942 instantiation), or a library unit renaming declaration (a package,
943 generic, or subprogram renaming declaration). Such files are also
946 @itemx @var{file}.adb
947 Ada source code file containing a library unit body (a subprogram or
948 package body). Such files are also called @dfn{bodies}.
950 @c GCC also knows about some suffixes for languages not yet included:
961 Assembler code which must be preprocessed.
964 An object file to be fed straight into linking.
965 Any file name with no recognized suffix is treated this way.
969 You can specify the input language explicitly with the @option{-x} option:
972 @item -x @var{language}
973 Specify explicitly the @var{language} for the following input files
974 (rather than letting the compiler choose a default based on the file
975 name suffix). This option applies to all following input files until
976 the next @option{-x} option. Possible values for @var{language} are:
978 c c-header c-cpp-output
979 c++ c++-header c++-cpp-output
980 objective-c objective-c-header objective-c-cpp-output
981 objective-c++ objective-c++-header objective-c++-cpp-output
982 assembler assembler-with-cpp
990 Turn off any specification of a language, so that subsequent files are
991 handled according to their file name suffixes (as they are if @option{-x}
992 has not been used at all).
994 @item -pass-exit-codes
995 @opindex pass-exit-codes
996 Normally the @command{gcc} program will exit with the code of 1 if any
997 phase of the compiler returns a non-success return code. If you specify
998 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
999 numerically highest error produced by any phase that returned an error
1000 indication. The C, C++, and Fortran frontends return 4, if an internal
1001 compiler error is encountered.
1004 If you only want some of the stages of compilation, you can use
1005 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1006 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1007 @command{gcc} is to stop. Note that some combinations (for example,
1008 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1013 Compile or assemble the source files, but do not link. The linking
1014 stage simply is not done. The ultimate output is in the form of an
1015 object file for each source file.
1017 By default, the object file name for a source file is made by replacing
1018 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1020 Unrecognized input files, not requiring compilation or assembly, are
1025 Stop after the stage of compilation proper; do not assemble. The output
1026 is in the form of an assembler code file for each non-assembler input
1029 By default, the assembler file name for a source file is made by
1030 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1032 Input files that don't require compilation are ignored.
1036 Stop after the preprocessing stage; do not run the compiler proper. The
1037 output is in the form of preprocessed source code, which is sent to the
1040 Input files which don't require preprocessing are ignored.
1042 @cindex output file option
1045 Place output in file @var{file}. This applies regardless to whatever
1046 sort of output is being produced, whether it be an executable file,
1047 an object file, an assembler file or preprocessed C code.
1049 If @option{-o} is not specified, the default is to put an executable
1050 file in @file{a.out}, the object file for
1051 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1052 assembler file in @file{@var{source}.s}, a precompiled header file in
1053 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1058 Print (on standard error output) the commands executed to run the stages
1059 of compilation. Also print the version number of the compiler driver
1060 program and of the preprocessor and the compiler proper.
1064 Like @option{-v} except the commands are not executed and all command
1065 arguments are quoted. This is useful for shell scripts to capture the
1066 driver-generated command lines.
1070 Use pipes rather than temporary files for communication between the
1071 various stages of compilation. This fails to work on some systems where
1072 the assembler is unable to read from a pipe; but the GNU assembler has
1077 If you are compiling multiple source files, this option tells the driver
1078 to pass all the source files to the compiler at once (for those
1079 languages for which the compiler can handle this). This will allow
1080 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1081 language for which this is supported is C@. If you pass source files for
1082 multiple languages to the driver, using this option, the driver will invoke
1083 the compiler(s) that support IMA once each, passing each compiler all the
1084 source files appropriate for it. For those languages that do not support
1085 IMA this option will be ignored, and the compiler will be invoked once for
1086 each source file in that language. If you use this option in conjunction
1087 with @option{-save-temps}, the compiler will generate multiple
1089 (one for each source file), but only one (combined) @file{.o} or
1094 Print (on the standard output) a description of the command line options
1095 understood by @command{gcc}. If the @option{-v} option is also specified
1096 then @option{--help} will also be passed on to the various processes
1097 invoked by @command{gcc}, so that they can display the command line options
1098 they accept. If the @option{-Wextra} option has also been specified
1099 (prior to the @option{--help} option), then command line options which
1100 have no documentation associated with them will also be displayed.
1103 @opindex target-help
1104 Print (on the standard output) a description of target-specific command
1105 line options for each tool.
1107 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1108 Print (on the standard output) a description of the command line
1109 options understood by the compiler that fit into a specific class.
1110 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1111 @samp{params}, or @var{language}:
1114 @item @samp{optimizers}
1115 This will display all of the optimization options supported by the
1118 @item @samp{warnings}
1119 This will display all of the options controlling warning messages
1120 produced by the compiler.
1123 This will display target-specific options. Unlike the
1124 @option{--target-help} option however, target-specific options of the
1125 linker and assembler will not be displayed. This is because those
1126 tools do not currently support the extended @option{--help=} syntax.
1129 This will display the values recognized by the @option{--param}
1132 @item @var{language}
1133 This will display the options supported for @var{language}, where
1134 @var{language} is the name of one of the languages supported in this
1138 This will display the options that are common to all languages.
1141 It is possible to further refine the output of the @option{--help=}
1142 option by adding a comma separated list of qualifiers after the
1143 class. These can be any from the following list:
1146 @item @samp{undocumented}
1147 Display only those options which are undocumented.
1150 Display options which take an argument that appears after an equal
1151 sign in the same continuous piece of text, such as:
1152 @samp{--help=target}.
1154 @item @samp{separate}
1155 Display options which take an argument that appears as a separate word
1156 following the original option, such as: @samp{-o output-file}.
1159 Thus for example to display all the undocumented target-specific
1160 switches supported by the compiler the following can be used:
1163 --help=target,undocumented
1166 The sense of a qualifier can be inverted by prefixing it with the
1167 @var{^} character, so for example to display all binary warning
1168 options (i.e. ones that are either on or off and that do not take an
1169 argument), which have a description the following can be used:
1172 --help=warnings,^joined,^undocumented
1175 A class can also be used as a qualifier, although this usually
1176 restricts the output by so much that there is nothing to display. One
1177 case where it does work however is when one of the classes is
1178 @var{target}. So for example to display all the target-specific
1179 optimization options the following can be used:
1182 --help=target,optimizers
1185 The @option{--help=} option can be repeated on the command line. Each
1186 successive use will display its requested class of options, skipping
1187 those that have already been displayed.
1189 If the @option{-Q} option appears on the command line before the
1190 @option{--help=} option, then the descriptive text displayed by
1191 @option{--help=} is changed. Instead of describing the displayed
1192 options, an indication is given as to whether the option is enabled,
1193 disabled or set to a specific value (assuming that the compiler
1194 knows this at the point where the @option{--help=} option is used).
1196 Here is a truncated example from the ARM port of @command{gcc}:
1199 % gcc -Q -mabi=2 --help=target -c
1200 The following options are target specific:
1202 -mabort-on-noreturn [disabled]
1206 The output is sensitive to the effects of previous command line
1207 options, so for example it is possible to find out which optimizations
1208 are enabled at @option{-O2} by using:
1211 -O2 --help=optimizers
1214 Alternatively you can discover which binary optimizations are enabled
1215 by @option{-O3} by using:
1218 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1219 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1220 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1225 Display the version number and copyrights of the invoked GCC@.
1227 @include @value{srcdir}/../libiberty/at-file.texi
1231 @section Compiling C++ Programs
1233 @cindex suffixes for C++ source
1234 @cindex C++ source file suffixes
1235 C++ source files conventionally use one of the suffixes @samp{.C},
1236 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1237 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1238 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1239 files with these names and compiles them as C++ programs even if you
1240 call the compiler the same way as for compiling C programs (usually
1241 with the name @command{gcc}).
1245 However, the use of @command{gcc} does not add the C++ library.
1246 @command{g++} is a program that calls GCC and treats @samp{.c},
1247 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1248 files unless @option{-x} is used, and automatically specifies linking
1249 against the C++ library. This program is also useful when
1250 precompiling a C header file with a @samp{.h} extension for use in C++
1251 compilations. On many systems, @command{g++} is also installed with
1252 the name @command{c++}.
1254 @cindex invoking @command{g++}
1255 When you compile C++ programs, you may specify many of the same
1256 command-line options that you use for compiling programs in any
1257 language; or command-line options meaningful for C and related
1258 languages; or options that are meaningful only for C++ programs.
1259 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1260 explanations of options for languages related to C@.
1261 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1262 explanations of options that are meaningful only for C++ programs.
1264 @node C Dialect Options
1265 @section Options Controlling C Dialect
1266 @cindex dialect options
1267 @cindex language dialect options
1268 @cindex options, dialect
1270 The following options control the dialect of C (or languages derived
1271 from C, such as C++, Objective-C and Objective-C++) that the compiler
1275 @cindex ANSI support
1279 In C mode, support all ISO C90 programs. In C++ mode,
1280 remove GNU extensions that conflict with ISO C++.
1282 This turns off certain features of GCC that are incompatible with ISO
1283 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1284 such as the @code{asm} and @code{typeof} keywords, and
1285 predefined macros such as @code{unix} and @code{vax} that identify the
1286 type of system you are using. It also enables the undesirable and
1287 rarely used ISO trigraph feature. For the C compiler,
1288 it disables recognition of C++ style @samp{//} comments as well as
1289 the @code{inline} keyword.
1291 The alternate keywords @code{__asm__}, @code{__extension__},
1292 @code{__inline__} and @code{__typeof__} continue to work despite
1293 @option{-ansi}. You would not want to use them in an ISO C program, of
1294 course, but it is useful to put them in header files that might be included
1295 in compilations done with @option{-ansi}. Alternate predefined macros
1296 such as @code{__unix__} and @code{__vax__} are also available, with or
1297 without @option{-ansi}.
1299 The @option{-ansi} option does not cause non-ISO programs to be
1300 rejected gratuitously. For that, @option{-pedantic} is required in
1301 addition to @option{-ansi}. @xref{Warning Options}.
1303 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1304 option is used. Some header files may notice this macro and refrain
1305 from declaring certain functions or defining certain macros that the
1306 ISO standard doesn't call for; this is to avoid interfering with any
1307 programs that might use these names for other things.
1309 Functions which would normally be built in but do not have semantics
1310 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1311 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1312 built-in functions provided by GCC}, for details of the functions
1317 Determine the language standard. This option is currently only
1318 supported when compiling C or C++. A value for this option must be
1319 provided; possible values are
1324 ISO C90 (same as @option{-ansi}).
1326 @item iso9899:199409
1327 ISO C90 as modified in amendment 1.
1333 ISO C99. Note that this standard is not yet fully supported; see
1334 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1335 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1338 Default, ISO C90 plus GNU extensions (including some C99 features).
1342 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1343 this will become the default. The name @samp{gnu9x} is deprecated.
1346 The 1998 ISO C++ standard plus amendments.
1349 The same as @option{-std=c++98} plus GNU extensions. This is the
1350 default for C++ code.
1353 The working draft of the upcoming ISO C++0x standard. This option
1354 enables experimental features that are likely to be included in
1355 C++0x. The working draft is constantly changing, and any feature that is
1356 enabled by this flag may be removed from future versions of GCC if it is
1357 not part of the C++0x standard.
1360 The same as @option{-std=c++0x} plus GNU extensions. As with
1361 @option{-std=c++0x}, this option enables experimental features that may
1362 be removed in future versions of GCC.
1365 Even when this option is not specified, you can still use some of the
1366 features of newer standards in so far as they do not conflict with
1367 previous C standards. For example, you may use @code{__restrict__} even
1368 when @option{-std=c99} is not specified.
1370 The @option{-std} options specifying some version of ISO C have the same
1371 effects as @option{-ansi}, except that features that were not in ISO C90
1372 but are in the specified version (for example, @samp{//} comments and
1373 the @code{inline} keyword in ISO C99) are not disabled.
1375 @xref{Standards,,Language Standards Supported by GCC}, for details of
1376 these standard versions.
1378 @item -fgnu89-inline
1379 @opindex fgnu89-inline
1380 The option @option{-fgnu89-inline} tells GCC to use the traditional
1381 GNU semantics for @code{inline} functions when in C99 mode.
1382 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1383 is accepted and ignored by GCC versions 4.1.3 up to but not including
1384 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1385 C99 mode. Using this option is roughly equivalent to adding the
1386 @code{gnu_inline} function attribute to all inline functions
1387 (@pxref{Function Attributes}).
1389 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1390 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1391 specifies the default behavior). This option was first supported in
1392 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1394 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1395 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1396 in effect for @code{inline} functions. @xref{Common Predefined
1397 Macros,,,cpp,The C Preprocessor}.
1399 @item -aux-info @var{filename}
1401 Output to the given filename prototyped declarations for all functions
1402 declared and/or defined in a translation unit, including those in header
1403 files. This option is silently ignored in any language other than C@.
1405 Besides declarations, the file indicates, in comments, the origin of
1406 each declaration (source file and line), whether the declaration was
1407 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1408 @samp{O} for old, respectively, in the first character after the line
1409 number and the colon), and whether it came from a declaration or a
1410 definition (@samp{C} or @samp{F}, respectively, in the following
1411 character). In the case of function definitions, a K&R-style list of
1412 arguments followed by their declarations is also provided, inside
1413 comments, after the declaration.
1417 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1418 keyword, so that code can use these words as identifiers. You can use
1419 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1420 instead. @option{-ansi} implies @option{-fno-asm}.
1422 In C++, this switch only affects the @code{typeof} keyword, since
1423 @code{asm} and @code{inline} are standard keywords. You may want to
1424 use the @option{-fno-gnu-keywords} flag instead, which has the same
1425 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1426 switch only affects the @code{asm} and @code{typeof} keywords, since
1427 @code{inline} is a standard keyword in ISO C99.
1430 @itemx -fno-builtin-@var{function}
1431 @opindex fno-builtin
1432 @cindex built-in functions
1433 Don't recognize built-in functions that do not begin with
1434 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1435 functions provided by GCC}, for details of the functions affected,
1436 including those which are not built-in functions when @option{-ansi} or
1437 @option{-std} options for strict ISO C conformance are used because they
1438 do not have an ISO standard meaning.
1440 GCC normally generates special code to handle certain built-in functions
1441 more efficiently; for instance, calls to @code{alloca} may become single
1442 instructions that adjust the stack directly, and calls to @code{memcpy}
1443 may become inline copy loops. The resulting code is often both smaller
1444 and faster, but since the function calls no longer appear as such, you
1445 cannot set a breakpoint on those calls, nor can you change the behavior
1446 of the functions by linking with a different library. In addition,
1447 when a function is recognized as a built-in function, GCC may use
1448 information about that function to warn about problems with calls to
1449 that function, or to generate more efficient code, even if the
1450 resulting code still contains calls to that function. For example,
1451 warnings are given with @option{-Wformat} for bad calls to
1452 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1453 known not to modify global memory.
1455 With the @option{-fno-builtin-@var{function}} option
1456 only the built-in function @var{function} is
1457 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1458 function is named this is not built-in in this version of GCC, this
1459 option is ignored. There is no corresponding
1460 @option{-fbuiltin-@var{function}} option; if you wish to enable
1461 built-in functions selectively when using @option{-fno-builtin} or
1462 @option{-ffreestanding}, you may define macros such as:
1465 #define abs(n) __builtin_abs ((n))
1466 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1471 @cindex hosted environment
1473 Assert that compilation takes place in a hosted environment. This implies
1474 @option{-fbuiltin}. A hosted environment is one in which the
1475 entire standard library is available, and in which @code{main} has a return
1476 type of @code{int}. Examples are nearly everything except a kernel.
1477 This is equivalent to @option{-fno-freestanding}.
1479 @item -ffreestanding
1480 @opindex ffreestanding
1481 @cindex hosted environment
1483 Assert that compilation takes place in a freestanding environment. This
1484 implies @option{-fno-builtin}. A freestanding environment
1485 is one in which the standard library may not exist, and program startup may
1486 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1487 This is equivalent to @option{-fno-hosted}.
1489 @xref{Standards,,Language Standards Supported by GCC}, for details of
1490 freestanding and hosted environments.
1494 @cindex openmp parallel
1495 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1496 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1497 compiler generates parallel code according to the OpenMP Application
1498 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1500 @item -fms-extensions
1501 @opindex fms-extensions
1502 Accept some non-standard constructs used in Microsoft header files.
1504 Some cases of unnamed fields in structures and unions are only
1505 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1506 fields within structs/unions}, for details.
1510 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1511 options for strict ISO C conformance) implies @option{-trigraphs}.
1513 @item -no-integrated-cpp
1514 @opindex no-integrated-cpp
1515 Performs a compilation in two passes: preprocessing and compiling. This
1516 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1517 @option{-B} option. The user supplied compilation step can then add in
1518 an additional preprocessing step after normal preprocessing but before
1519 compiling. The default is to use the integrated cpp (internal cpp)
1521 The semantics of this option will change if "cc1", "cc1plus", and
1522 "cc1obj" are merged.
1524 @cindex traditional C language
1525 @cindex C language, traditional
1527 @itemx -traditional-cpp
1528 @opindex traditional-cpp
1529 @opindex traditional
1530 Formerly, these options caused GCC to attempt to emulate a pre-standard
1531 C compiler. They are now only supported with the @option{-E} switch.
1532 The preprocessor continues to support a pre-standard mode. See the GNU
1533 CPP manual for details.
1535 @item -fcond-mismatch
1536 @opindex fcond-mismatch
1537 Allow conditional expressions with mismatched types in the second and
1538 third arguments. The value of such an expression is void. This option
1539 is not supported for C++.
1541 @item -flax-vector-conversions
1542 @opindex flax-vector-conversions
1543 Allow implicit conversions between vectors with differing numbers of
1544 elements and/or incompatible element types. This option should not be
1547 @item -funsigned-char
1548 @opindex funsigned-char
1549 Let the type @code{char} be unsigned, like @code{unsigned char}.
1551 Each kind of machine has a default for what @code{char} should
1552 be. It is either like @code{unsigned char} by default or like
1553 @code{signed char} by default.
1555 Ideally, a portable program should always use @code{signed char} or
1556 @code{unsigned char} when it depends on the signedness of an object.
1557 But many programs have been written to use plain @code{char} and
1558 expect it to be signed, or expect it to be unsigned, depending on the
1559 machines they were written for. This option, and its inverse, let you
1560 make such a program work with the opposite default.
1562 The type @code{char} is always a distinct type from each of
1563 @code{signed char} or @code{unsigned char}, even though its behavior
1564 is always just like one of those two.
1567 @opindex fsigned-char
1568 Let the type @code{char} be signed, like @code{signed char}.
1570 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1571 the negative form of @option{-funsigned-char}. Likewise, the option
1572 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1574 @item -fsigned-bitfields
1575 @itemx -funsigned-bitfields
1576 @itemx -fno-signed-bitfields
1577 @itemx -fno-unsigned-bitfields
1578 @opindex fsigned-bitfields
1579 @opindex funsigned-bitfields
1580 @opindex fno-signed-bitfields
1581 @opindex fno-unsigned-bitfields
1582 These options control whether a bit-field is signed or unsigned, when the
1583 declaration does not use either @code{signed} or @code{unsigned}. By
1584 default, such a bit-field is signed, because this is consistent: the
1585 basic integer types such as @code{int} are signed types.
1588 @node C++ Dialect Options
1589 @section Options Controlling C++ Dialect
1591 @cindex compiler options, C++
1592 @cindex C++ options, command line
1593 @cindex options, C++
1594 This section describes the command-line options that are only meaningful
1595 for C++ programs; but you can also use most of the GNU compiler options
1596 regardless of what language your program is in. For example, you
1597 might compile a file @code{firstClass.C} like this:
1600 g++ -g -frepo -O -c firstClass.C
1604 In this example, only @option{-frepo} is an option meant
1605 only for C++ programs; you can use the other options with any
1606 language supported by GCC@.
1608 Here is a list of options that are @emph{only} for compiling C++ programs:
1612 @item -fabi-version=@var{n}
1613 @opindex fabi-version
1614 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1615 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1616 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1617 the version that conforms most closely to the C++ ABI specification.
1618 Therefore, the ABI obtained using version 0 will change as ABI bugs
1621 The default is version 2.
1623 @item -fno-access-control
1624 @opindex fno-access-control
1625 Turn off all access checking. This switch is mainly useful for working
1626 around bugs in the access control code.
1630 Check that the pointer returned by @code{operator new} is non-null
1631 before attempting to modify the storage allocated. This check is
1632 normally unnecessary because the C++ standard specifies that
1633 @code{operator new} will only return @code{0} if it is declared
1634 @samp{throw()}, in which case the compiler will always check the
1635 return value even without this option. In all other cases, when
1636 @code{operator new} has a non-empty exception specification, memory
1637 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1638 @samp{new (nothrow)}.
1640 @item -fconserve-space
1641 @opindex fconserve-space
1642 Put uninitialized or runtime-initialized global variables into the
1643 common segment, as C does. This saves space in the executable at the
1644 cost of not diagnosing duplicate definitions. If you compile with this
1645 flag and your program mysteriously crashes after @code{main()} has
1646 completed, you may have an object that is being destroyed twice because
1647 two definitions were merged.
1649 This option is no longer useful on most targets, now that support has
1650 been added for putting variables into BSS without making them common.
1652 @item -ffriend-injection
1653 @opindex ffriend-injection
1654 Inject friend functions into the enclosing namespace, so that they are
1655 visible outside the scope of the class in which they are declared.
1656 Friend functions were documented to work this way in the old Annotated
1657 C++ Reference Manual, and versions of G++ before 4.1 always worked
1658 that way. However, in ISO C++ a friend function which is not declared
1659 in an enclosing scope can only be found using argument dependent
1660 lookup. This option causes friends to be injected as they were in
1663 This option is for compatibility, and may be removed in a future
1666 @item -fno-elide-constructors
1667 @opindex fno-elide-constructors
1668 The C++ standard allows an implementation to omit creating a temporary
1669 which is only used to initialize another object of the same type.
1670 Specifying this option disables that optimization, and forces G++ to
1671 call the copy constructor in all cases.
1673 @item -fno-enforce-eh-specs
1674 @opindex fno-enforce-eh-specs
1675 Don't generate code to check for violation of exception specifications
1676 at runtime. This option violates the C++ standard, but may be useful
1677 for reducing code size in production builds, much like defining
1678 @samp{NDEBUG}. This does not give user code permission to throw
1679 exceptions in violation of the exception specifications; the compiler
1680 will still optimize based on the specifications, so throwing an
1681 unexpected exception will result in undefined behavior.
1684 @itemx -fno-for-scope
1686 @opindex fno-for-scope
1687 If @option{-ffor-scope} is specified, the scope of variables declared in
1688 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1689 as specified by the C++ standard.
1690 If @option{-fno-for-scope} is specified, the scope of variables declared in
1691 a @i{for-init-statement} extends to the end of the enclosing scope,
1692 as was the case in old versions of G++, and other (traditional)
1693 implementations of C++.
1695 The default if neither flag is given to follow the standard,
1696 but to allow and give a warning for old-style code that would
1697 otherwise be invalid, or have different behavior.
1699 @item -fno-gnu-keywords
1700 @opindex fno-gnu-keywords
1701 Do not recognize @code{typeof} as a keyword, so that code can use this
1702 word as an identifier. You can use the keyword @code{__typeof__} instead.
1703 @option{-ansi} implies @option{-fno-gnu-keywords}.
1705 @item -fno-implicit-templates
1706 @opindex fno-implicit-templates
1707 Never emit code for non-inline templates which are instantiated
1708 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1709 @xref{Template Instantiation}, for more information.
1711 @item -fno-implicit-inline-templates
1712 @opindex fno-implicit-inline-templates
1713 Don't emit code for implicit instantiations of inline templates, either.
1714 The default is to handle inlines differently so that compiles with and
1715 without optimization will need the same set of explicit instantiations.
1717 @item -fno-implement-inlines
1718 @opindex fno-implement-inlines
1719 To save space, do not emit out-of-line copies of inline functions
1720 controlled by @samp{#pragma implementation}. This will cause linker
1721 errors if these functions are not inlined everywhere they are called.
1723 @item -fms-extensions
1724 @opindex fms-extensions
1725 Disable pedantic warnings about constructs used in MFC, such as implicit
1726 int and getting a pointer to member function via non-standard syntax.
1728 @item -fno-nonansi-builtins
1729 @opindex fno-nonansi-builtins
1730 Disable built-in declarations of functions that are not mandated by
1731 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1732 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1734 @item -fno-operator-names
1735 @opindex fno-operator-names
1736 Do not treat the operator name keywords @code{and}, @code{bitand},
1737 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1738 synonyms as keywords.
1740 @item -fno-optional-diags
1741 @opindex fno-optional-diags
1742 Disable diagnostics that the standard says a compiler does not need to
1743 issue. Currently, the only such diagnostic issued by G++ is the one for
1744 a name having multiple meanings within a class.
1747 @opindex fpermissive
1748 Downgrade some diagnostics about nonconformant code from errors to
1749 warnings. Thus, using @option{-fpermissive} will allow some
1750 nonconforming code to compile.
1754 Enable automatic template instantiation at link time. This option also
1755 implies @option{-fno-implicit-templates}. @xref{Template
1756 Instantiation}, for more information.
1760 Disable generation of information about every class with virtual
1761 functions for use by the C++ runtime type identification features
1762 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1763 of the language, you can save some space by using this flag. Note that
1764 exception handling uses the same information, but it will generate it as
1765 needed. The @samp{dynamic_cast} operator can still be used for casts that
1766 do not require runtime type information, i.e. casts to @code{void *} or to
1767 unambiguous base classes.
1771 Emit statistics about front-end processing at the end of the compilation.
1772 This information is generally only useful to the G++ development team.
1774 @item -ftemplate-depth-@var{n}
1775 @opindex ftemplate-depth
1776 Set the maximum instantiation depth for template classes to @var{n}.
1777 A limit on the template instantiation depth is needed to detect
1778 endless recursions during template class instantiation. ANSI/ISO C++
1779 conforming programs must not rely on a maximum depth greater than 17.
1781 @item -fno-threadsafe-statics
1782 @opindex fno-threadsafe-statics
1783 Do not emit the extra code to use the routines specified in the C++
1784 ABI for thread-safe initialization of local statics. You can use this
1785 option to reduce code size slightly in code that doesn't need to be
1788 @item -fuse-cxa-atexit
1789 @opindex fuse-cxa-atexit
1790 Register destructors for objects with static storage duration with the
1791 @code{__cxa_atexit} function rather than the @code{atexit} function.
1792 This option is required for fully standards-compliant handling of static
1793 destructors, but will only work if your C library supports
1794 @code{__cxa_atexit}.
1796 @item -fno-use-cxa-get-exception-ptr
1797 @opindex fno-use-cxa-get-exception-ptr
1798 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1799 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1800 if the runtime routine is not available.
1802 @item -fvisibility-inlines-hidden
1803 @opindex fvisibility-inlines-hidden
1804 This switch declares that the user does not attempt to compare
1805 pointers to inline methods where the addresses of the two functions
1806 were taken in different shared objects.
1808 The effect of this is that GCC may, effectively, mark inline methods with
1809 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1810 appear in the export table of a DSO and do not require a PLT indirection
1811 when used within the DSO@. Enabling this option can have a dramatic effect
1812 on load and link times of a DSO as it massively reduces the size of the
1813 dynamic export table when the library makes heavy use of templates.
1815 The behavior of this switch is not quite the same as marking the
1816 methods as hidden directly, because it does not affect static variables
1817 local to the function or cause the compiler to deduce that
1818 the function is defined in only one shared object.
1820 You may mark a method as having a visibility explicitly to negate the
1821 effect of the switch for that method. For example, if you do want to
1822 compare pointers to a particular inline method, you might mark it as
1823 having default visibility. Marking the enclosing class with explicit
1824 visibility will have no effect.
1826 Explicitly instantiated inline methods are unaffected by this option
1827 as their linkage might otherwise cross a shared library boundary.
1828 @xref{Template Instantiation}.
1832 Do not use weak symbol support, even if it is provided by the linker.
1833 By default, G++ will use weak symbols if they are available. This
1834 option exists only for testing, and should not be used by end-users;
1835 it will result in inferior code and has no benefits. This option may
1836 be removed in a future release of G++.
1840 Do not search for header files in the standard directories specific to
1841 C++, but do still search the other standard directories. (This option
1842 is used when building the C++ library.)
1845 In addition, these optimization, warning, and code generation options
1846 have meanings only for C++ programs:
1849 @item -fno-default-inline
1850 @opindex fno-default-inline
1851 Do not assume @samp{inline} for functions defined inside a class scope.
1852 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1853 functions will have linkage like inline functions; they just won't be
1856 @item -Wabi @r{(C++ only)}
1858 Warn when G++ generates code that is probably not compatible with the
1859 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1860 all such cases, there are probably some cases that are not warned about,
1861 even though G++ is generating incompatible code. There may also be
1862 cases where warnings are emitted even though the code that is generated
1865 You should rewrite your code to avoid these warnings if you are
1866 concerned about the fact that code generated by G++ may not be binary
1867 compatible with code generated by other compilers.
1869 The known incompatibilities at this point include:
1874 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1875 pack data into the same byte as a base class. For example:
1878 struct A @{ virtual void f(); int f1 : 1; @};
1879 struct B : public A @{ int f2 : 1; @};
1883 In this case, G++ will place @code{B::f2} into the same byte
1884 as@code{A::f1}; other compilers will not. You can avoid this problem
1885 by explicitly padding @code{A} so that its size is a multiple of the
1886 byte size on your platform; that will cause G++ and other compilers to
1887 layout @code{B} identically.
1890 Incorrect handling of tail-padding for virtual bases. G++ does not use
1891 tail padding when laying out virtual bases. For example:
1894 struct A @{ virtual void f(); char c1; @};
1895 struct B @{ B(); char c2; @};
1896 struct C : public A, public virtual B @{@};
1900 In this case, G++ will not place @code{B} into the tail-padding for
1901 @code{A}; other compilers will. You can avoid this problem by
1902 explicitly padding @code{A} so that its size is a multiple of its
1903 alignment (ignoring virtual base classes); that will cause G++ and other
1904 compilers to layout @code{C} identically.
1907 Incorrect handling of bit-fields with declared widths greater than that
1908 of their underlying types, when the bit-fields appear in a union. For
1912 union U @{ int i : 4096; @};
1916 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1917 union too small by the number of bits in an @code{int}.
1920 Empty classes can be placed at incorrect offsets. For example:
1930 struct C : public B, public A @{@};
1934 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1935 it should be placed at offset zero. G++ mistakenly believes that the
1936 @code{A} data member of @code{B} is already at offset zero.
1939 Names of template functions whose types involve @code{typename} or
1940 template template parameters can be mangled incorrectly.
1943 template <typename Q>
1944 void f(typename Q::X) @{@}
1946 template <template <typename> class Q>
1947 void f(typename Q<int>::X) @{@}
1951 Instantiations of these templates may be mangled incorrectly.
1955 @item -Wctor-dtor-privacy @r{(C++ only)}
1956 @opindex Wctor-dtor-privacy
1957 Warn when a class seems unusable because all the constructors or
1958 destructors in that class are private, and it has neither friends nor
1959 public static member functions.
1961 @item -Wnon-virtual-dtor @r{(C++ only)}
1962 @opindex Wnon-virtual-dtor
1963 Warn when a class appears to be polymorphic, thereby requiring a virtual
1964 destructor, yet it declares a non-virtual one. This warning is also
1965 enabled if -Weffc++ is specified.
1967 @item -Wreorder @r{(C++ only)}
1969 @cindex reordering, warning
1970 @cindex warning for reordering of member initializers
1971 Warn when the order of member initializers given in the code does not
1972 match the order in which they must be executed. For instance:
1978 A(): j (0), i (1) @{ @}
1982 The compiler will rearrange the member initializers for @samp{i}
1983 and @samp{j} to match the declaration order of the members, emitting
1984 a warning to that effect. This warning is enabled by @option{-Wall}.
1987 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1990 @item -Weffc++ @r{(C++ only)}
1992 Warn about violations of the following style guidelines from Scott Meyers'
1993 @cite{Effective C++} book:
1997 Item 11: Define a copy constructor and an assignment operator for classes
1998 with dynamically allocated memory.
2001 Item 12: Prefer initialization to assignment in constructors.
2004 Item 14: Make destructors virtual in base classes.
2007 Item 15: Have @code{operator=} return a reference to @code{*this}.
2010 Item 23: Don't try to return a reference when you must return an object.
2014 Also warn about violations of the following style guidelines from
2015 Scott Meyers' @cite{More Effective C++} book:
2019 Item 6: Distinguish between prefix and postfix forms of increment and
2020 decrement operators.
2023 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2027 When selecting this option, be aware that the standard library
2028 headers do not obey all of these guidelines; use @samp{grep -v}
2029 to filter out those warnings.
2031 @item -Wno-deprecated @r{(C++ only)}
2032 @opindex Wno-deprecated
2033 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
2035 @item -Wstrict-null-sentinel @r{(C++ only)}
2036 @opindex Wstrict-null-sentinel
2037 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2038 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2039 to @code{__null}. Although it is a null pointer constant not a null pointer,
2040 it is guaranteed to of the same size as a pointer. But this use is
2041 not portable across different compilers.
2043 @item -Wno-non-template-friend @r{(C++ only)}
2044 @opindex Wno-non-template-friend
2045 Disable warnings when non-templatized friend functions are declared
2046 within a template. Since the advent of explicit template specification
2047 support in G++, if the name of the friend is an unqualified-id (i.e.,
2048 @samp{friend foo(int)}), the C++ language specification demands that the
2049 friend declare or define an ordinary, nontemplate function. (Section
2050 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2051 could be interpreted as a particular specialization of a templatized
2052 function. Because this non-conforming behavior is no longer the default
2053 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2054 check existing code for potential trouble spots and is on by default.
2055 This new compiler behavior can be turned off with
2056 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2057 but disables the helpful warning.
2059 @item -Wold-style-cast @r{(C++ only)}
2060 @opindex Wold-style-cast
2061 Warn if an old-style (C-style) cast to a non-void type is used within
2062 a C++ program. The new-style casts (@samp{dynamic_cast},
2063 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2064 less vulnerable to unintended effects and much easier to search for.
2066 @item -Woverloaded-virtual @r{(C++ only)}
2067 @opindex Woverloaded-virtual
2068 @cindex overloaded virtual fn, warning
2069 @cindex warning for overloaded virtual fn
2070 Warn when a function declaration hides virtual functions from a
2071 base class. For example, in:
2078 struct B: public A @{
2083 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2091 will fail to compile.
2093 @item -Wno-pmf-conversions @r{(C++ only)}
2094 @opindex Wno-pmf-conversions
2095 Disable the diagnostic for converting a bound pointer to member function
2098 @item -Wsign-promo @r{(C++ only)}
2099 @opindex Wsign-promo
2100 Warn when overload resolution chooses a promotion from unsigned or
2101 enumerated type to a signed type, over a conversion to an unsigned type of
2102 the same size. Previous versions of G++ would try to preserve
2103 unsignedness, but the standard mandates the current behavior.
2108 A& operator = (int);
2118 In this example, G++ will synthesize a default @samp{A& operator =
2119 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2122 @node Objective-C and Objective-C++ Dialect Options
2123 @section Options Controlling Objective-C and Objective-C++ Dialects
2125 @cindex compiler options, Objective-C and Objective-C++
2126 @cindex Objective-C and Objective-C++ options, command line
2127 @cindex options, Objective-C and Objective-C++
2128 (NOTE: This manual does not describe the Objective-C and Objective-C++
2129 languages themselves. See @xref{Standards,,Language Standards
2130 Supported by GCC}, for references.)
2132 This section describes the command-line options that are only meaningful
2133 for Objective-C and Objective-C++ programs, but you can also use most of
2134 the language-independent GNU compiler options.
2135 For example, you might compile a file @code{some_class.m} like this:
2138 gcc -g -fgnu-runtime -O -c some_class.m
2142 In this example, @option{-fgnu-runtime} is an option meant only for
2143 Objective-C and Objective-C++ programs; you can use the other options with
2144 any language supported by GCC@.
2146 Note that since Objective-C is an extension of the C language, Objective-C
2147 compilations may also use options specific to the C front-end (e.g.,
2148 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2149 C++-specific options (e.g., @option{-Wabi}).
2151 Here is a list of options that are @emph{only} for compiling Objective-C
2152 and Objective-C++ programs:
2155 @item -fconstant-string-class=@var{class-name}
2156 @opindex fconstant-string-class
2157 Use @var{class-name} as the name of the class to instantiate for each
2158 literal string specified with the syntax @code{@@"@dots{}"}. The default
2159 class name is @code{NXConstantString} if the GNU runtime is being used, and
2160 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2161 @option{-fconstant-cfstrings} option, if also present, will override the
2162 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2163 to be laid out as constant CoreFoundation strings.
2166 @opindex fgnu-runtime
2167 Generate object code compatible with the standard GNU Objective-C
2168 runtime. This is the default for most types of systems.
2170 @item -fnext-runtime
2171 @opindex fnext-runtime
2172 Generate output compatible with the NeXT runtime. This is the default
2173 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2174 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2177 @item -fno-nil-receivers
2178 @opindex fno-nil-receivers
2179 Assume that all Objective-C message dispatches (e.g.,
2180 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2181 is not @code{nil}. This allows for more efficient entry points in the runtime
2182 to be used. Currently, this option is only available in conjunction with
2183 the NeXT runtime on Mac OS X 10.3 and later.
2185 @item -fobjc-call-cxx-cdtors
2186 @opindex fobjc-call-cxx-cdtors
2187 For each Objective-C class, check if any of its instance variables is a
2188 C++ object with a non-trivial default constructor. If so, synthesize a
2189 special @code{- (id) .cxx_construct} instance method that will run
2190 non-trivial default constructors on any such instance variables, in order,
2191 and then return @code{self}. Similarly, check if any instance variable
2192 is a C++ object with a non-trivial destructor, and if so, synthesize a
2193 special @code{- (void) .cxx_destruct} method that will run
2194 all such default destructors, in reverse order.
2196 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2197 thusly generated will only operate on instance variables declared in the
2198 current Objective-C class, and not those inherited from superclasses. It
2199 is the responsibility of the Objective-C runtime to invoke all such methods
2200 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2201 will be invoked by the runtime immediately after a new object
2202 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2203 be invoked immediately before the runtime deallocates an object instance.
2205 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2206 support for invoking the @code{- (id) .cxx_construct} and
2207 @code{- (void) .cxx_destruct} methods.
2209 @item -fobjc-direct-dispatch
2210 @opindex fobjc-direct-dispatch
2211 Allow fast jumps to the message dispatcher. On Darwin this is
2212 accomplished via the comm page.
2214 @item -fobjc-exceptions
2215 @opindex fobjc-exceptions
2216 Enable syntactic support for structured exception handling in Objective-C,
2217 similar to what is offered by C++ and Java. This option is
2218 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2227 @@catch (AnObjCClass *exc) @{
2234 @@catch (AnotherClass *exc) @{
2237 @@catch (id allOthers) @{
2247 The @code{@@throw} statement may appear anywhere in an Objective-C or
2248 Objective-C++ program; when used inside of a @code{@@catch} block, the
2249 @code{@@throw} may appear without an argument (as shown above), in which case
2250 the object caught by the @code{@@catch} will be rethrown.
2252 Note that only (pointers to) Objective-C objects may be thrown and
2253 caught using this scheme. When an object is thrown, it will be caught
2254 by the nearest @code{@@catch} clause capable of handling objects of that type,
2255 analogously to how @code{catch} blocks work in C++ and Java. A
2256 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2257 any and all Objective-C exceptions not caught by previous @code{@@catch}
2260 The @code{@@finally} clause, if present, will be executed upon exit from the
2261 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2262 regardless of whether any exceptions are thrown, caught or rethrown
2263 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2264 of the @code{finally} clause in Java.
2266 There are several caveats to using the new exception mechanism:
2270 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2271 idioms provided by the @code{NSException} class, the new
2272 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2273 systems, due to additional functionality needed in the (NeXT) Objective-C
2277 As mentioned above, the new exceptions do not support handling
2278 types other than Objective-C objects. Furthermore, when used from
2279 Objective-C++, the Objective-C exception model does not interoperate with C++
2280 exceptions at this time. This means you cannot @code{@@throw} an exception
2281 from Objective-C and @code{catch} it in C++, or vice versa
2282 (i.e., @code{throw @dots{} @@catch}).
2285 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2286 blocks for thread-safe execution:
2289 @@synchronized (ObjCClass *guard) @{
2294 Upon entering the @code{@@synchronized} block, a thread of execution shall
2295 first check whether a lock has been placed on the corresponding @code{guard}
2296 object by another thread. If it has, the current thread shall wait until
2297 the other thread relinquishes its lock. Once @code{guard} becomes available,
2298 the current thread will place its own lock on it, execute the code contained in
2299 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2300 making @code{guard} available to other threads).
2302 Unlike Java, Objective-C does not allow for entire methods to be marked
2303 @code{@@synchronized}. Note that throwing exceptions out of
2304 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2305 to be unlocked properly.
2309 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2311 @item -freplace-objc-classes
2312 @opindex freplace-objc-classes
2313 Emit a special marker instructing @command{ld(1)} not to statically link in
2314 the resulting object file, and allow @command{dyld(1)} to load it in at
2315 run time instead. This is used in conjunction with the Fix-and-Continue
2316 debugging mode, where the object file in question may be recompiled and
2317 dynamically reloaded in the course of program execution, without the need
2318 to restart the program itself. Currently, Fix-and-Continue functionality
2319 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2324 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2325 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2326 compile time) with static class references that get initialized at load time,
2327 which improves run-time performance. Specifying the @option{-fzero-link} flag
2328 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2329 to be retained. This is useful in Zero-Link debugging mode, since it allows
2330 for individual class implementations to be modified during program execution.
2334 Dump interface declarations for all classes seen in the source file to a
2335 file named @file{@var{sourcename}.decl}.
2337 @item -Wassign-intercept
2338 @opindex Wassign-intercept
2339 Warn whenever an Objective-C assignment is being intercepted by the
2343 @opindex Wno-protocol
2344 If a class is declared to implement a protocol, a warning is issued for
2345 every method in the protocol that is not implemented by the class. The
2346 default behavior is to issue a warning for every method not explicitly
2347 implemented in the class, even if a method implementation is inherited
2348 from the superclass. If you use the @option{-Wno-protocol} option, then
2349 methods inherited from the superclass are considered to be implemented,
2350 and no warning is issued for them.
2354 Warn if multiple methods of different types for the same selector are
2355 found during compilation. The check is performed on the list of methods
2356 in the final stage of compilation. Additionally, a check is performed
2357 for each selector appearing in a @code{@@selector(@dots{})}
2358 expression, and a corresponding method for that selector has been found
2359 during compilation. Because these checks scan the method table only at
2360 the end of compilation, these warnings are not produced if the final
2361 stage of compilation is not reached, for example because an error is
2362 found during compilation, or because the @option{-fsyntax-only} option is
2365 @item -Wstrict-selector-match
2366 @opindex Wstrict-selector-match
2367 Warn if multiple methods with differing argument and/or return types are
2368 found for a given selector when attempting to send a message using this
2369 selector to a receiver of type @code{id} or @code{Class}. When this flag
2370 is off (which is the default behavior), the compiler will omit such warnings
2371 if any differences found are confined to types which share the same size
2374 @item -Wundeclared-selector
2375 @opindex Wundeclared-selector
2376 Warn if a @code{@@selector(@dots{})} expression referring to an
2377 undeclared selector is found. A selector is considered undeclared if no
2378 method with that name has been declared before the
2379 @code{@@selector(@dots{})} expression, either explicitly in an
2380 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2381 an @code{@@implementation} section. This option always performs its
2382 checks as soon as a @code{@@selector(@dots{})} expression is found,
2383 while @option{-Wselector} only performs its checks in the final stage of
2384 compilation. This also enforces the coding style convention
2385 that methods and selectors must be declared before being used.
2387 @item -print-objc-runtime-info
2388 @opindex print-objc-runtime-info
2389 Generate C header describing the largest structure that is passed by
2394 @node Language Independent Options
2395 @section Options to Control Diagnostic Messages Formatting
2396 @cindex options to control diagnostics formatting
2397 @cindex diagnostic messages
2398 @cindex message formatting
2400 Traditionally, diagnostic messages have been formatted irrespective of
2401 the output device's aspect (e.g.@: its width, @dots{}). The options described
2402 below can be used to control the diagnostic messages formatting
2403 algorithm, e.g.@: how many characters per line, how often source location
2404 information should be reported. Right now, only the C++ front end can
2405 honor these options. However it is expected, in the near future, that
2406 the remaining front ends would be able to digest them correctly.
2409 @item -fmessage-length=@var{n}
2410 @opindex fmessage-length
2411 Try to format error messages so that they fit on lines of about @var{n}
2412 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2413 the front ends supported by GCC@. If @var{n} is zero, then no
2414 line-wrapping will be done; each error message will appear on a single
2417 @opindex fdiagnostics-show-location
2418 @item -fdiagnostics-show-location=once
2419 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2420 reporter to emit @emph{once} source location information; that is, in
2421 case the message is too long to fit on a single physical line and has to
2422 be wrapped, the source location won't be emitted (as prefix) again,
2423 over and over, in subsequent continuation lines. This is the default
2426 @item -fdiagnostics-show-location=every-line
2427 Only meaningful in line-wrapping mode. Instructs the diagnostic
2428 messages reporter to emit the same source location information (as
2429 prefix) for physical lines that result from the process of breaking
2430 a message which is too long to fit on a single line.
2432 @item -fdiagnostics-show-option
2433 @opindex fdiagnostics-show-option
2434 This option instructs the diagnostic machinery to add text to each
2435 diagnostic emitted, which indicates which command line option directly
2436 controls that diagnostic, when such an option is known to the
2437 diagnostic machinery.
2439 @item -Wcoverage-mismatch
2440 @opindex Wcoverage-mismatch
2441 Warn if feedback profiles do not match when using the
2442 @option{-fprofile-use} option.
2443 If a source file was changed between @option{-fprofile-gen} and
2444 @option{-fprofile-use}, the files with the profile feedback can fail
2445 to match the source file and GCC can not use the profile feedback
2446 information. By default, GCC emits an error message in this case.
2447 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2448 error. GCC does not use appropriate feedback profiles, so using this
2449 option can result in poorly optimized code. This option is useful
2450 only in the case of very minor changes such as bug fixes to an
2455 @node Warning Options
2456 @section Options to Request or Suppress Warnings
2457 @cindex options to control warnings
2458 @cindex warning messages
2459 @cindex messages, warning
2460 @cindex suppressing warnings
2462 Warnings are diagnostic messages that report constructions which
2463 are not inherently erroneous but which are risky or suggest there
2464 may have been an error.
2466 You can request many specific warnings with options beginning @samp{-W},
2467 for example @option{-Wimplicit} to request warnings on implicit
2468 declarations. Each of these specific warning options also has a
2469 negative form beginning @samp{-Wno-} to turn off warnings;
2470 for example, @option{-Wno-implicit}. This manual lists only one of the
2471 two forms, whichever is not the default.
2473 The following options control the amount and kinds of warnings produced
2474 by GCC; for further, language-specific options also refer to
2475 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2479 @cindex syntax checking
2481 @opindex fsyntax-only
2482 Check the code for syntax errors, but don't do anything beyond that.
2486 Issue all the warnings demanded by strict ISO C and ISO C++;
2487 reject all programs that use forbidden extensions, and some other
2488 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2489 version of the ISO C standard specified by any @option{-std} option used.
2491 Valid ISO C and ISO C++ programs should compile properly with or without
2492 this option (though a rare few will require @option{-ansi} or a
2493 @option{-std} option specifying the required version of ISO C)@. However,
2494 without this option, certain GNU extensions and traditional C and C++
2495 features are supported as well. With this option, they are rejected.
2497 @option{-pedantic} does not cause warning messages for use of the
2498 alternate keywords whose names begin and end with @samp{__}. Pedantic
2499 warnings are also disabled in the expression that follows
2500 @code{__extension__}. However, only system header files should use
2501 these escape routes; application programs should avoid them.
2502 @xref{Alternate Keywords}.
2504 Some users try to use @option{-pedantic} to check programs for strict ISO
2505 C conformance. They soon find that it does not do quite what they want:
2506 it finds some non-ISO practices, but not all---only those for which
2507 ISO C @emph{requires} a diagnostic, and some others for which
2508 diagnostics have been added.
2510 A feature to report any failure to conform to ISO C might be useful in
2511 some instances, but would require considerable additional work and would
2512 be quite different from @option{-pedantic}. We don't have plans to
2513 support such a feature in the near future.
2515 Where the standard specified with @option{-std} represents a GNU
2516 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2517 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2518 extended dialect is based. Warnings from @option{-pedantic} are given
2519 where they are required by the base standard. (It would not make sense
2520 for such warnings to be given only for features not in the specified GNU
2521 C dialect, since by definition the GNU dialects of C include all
2522 features the compiler supports with the given option, and there would be
2523 nothing to warn about.)
2525 @item -pedantic-errors
2526 @opindex pedantic-errors
2527 Like @option{-pedantic}, except that errors are produced rather than
2532 Inhibit all warning messages.
2536 Inhibit warning messages about the use of @samp{#import}.
2538 @item -Wchar-subscripts
2539 @opindex Wchar-subscripts
2540 Warn if an array subscript has type @code{char}. This is a common cause
2541 of error, as programmers often forget that this type is signed on some
2543 This warning is enabled by @option{-Wall}.
2547 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2548 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2549 This warning is enabled by @option{-Wall}.
2551 @item -Wfatal-errors
2552 @opindex Wfatal-errors
2553 This option causes the compiler to abort compilation on the first error
2554 occurred rather than trying to keep going and printing further error
2559 @opindex ffreestanding
2560 @opindex fno-builtin
2561 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2562 the arguments supplied have types appropriate to the format string
2563 specified, and that the conversions specified in the format string make
2564 sense. This includes standard functions, and others specified by format
2565 attributes (@pxref{Function Attributes}), in the @code{printf},
2566 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2567 not in the C standard) families (or other target-specific families).
2568 Which functions are checked without format attributes having been
2569 specified depends on the standard version selected, and such checks of
2570 functions without the attribute specified are disabled by
2571 @option{-ffreestanding} or @option{-fno-builtin}.
2573 The formats are checked against the format features supported by GNU
2574 libc version 2.2. These include all ISO C90 and C99 features, as well
2575 as features from the Single Unix Specification and some BSD and GNU
2576 extensions. Other library implementations may not support all these
2577 features; GCC does not support warning about features that go beyond a
2578 particular library's limitations. However, if @option{-pedantic} is used
2579 with @option{-Wformat}, warnings will be given about format features not
2580 in the selected standard version (but not for @code{strfmon} formats,
2581 since those are not in any version of the C standard). @xref{C Dialect
2582 Options,,Options Controlling C Dialect}.
2584 Since @option{-Wformat} also checks for null format arguments for
2585 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2587 @option{-Wformat} is included in @option{-Wall}. For more control over some
2588 aspects of format checking, the options @option{-Wformat-y2k},
2589 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2590 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2591 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2594 @opindex Wformat-y2k
2595 If @option{-Wformat} is specified, also warn about @code{strftime}
2596 formats which may yield only a two-digit year.
2598 @item -Wno-format-extra-args
2599 @opindex Wno-format-extra-args
2600 If @option{-Wformat} is specified, do not warn about excess arguments to a
2601 @code{printf} or @code{scanf} format function. The C standard specifies
2602 that such arguments are ignored.
2604 Where the unused arguments lie between used arguments that are
2605 specified with @samp{$} operand number specifications, normally
2606 warnings are still given, since the implementation could not know what
2607 type to pass to @code{va_arg} to skip the unused arguments. However,
2608 in the case of @code{scanf} formats, this option will suppress the
2609 warning if the unused arguments are all pointers, since the Single
2610 Unix Specification says that such unused arguments are allowed.
2612 @item -Wno-format-zero-length
2613 @opindex Wno-format-zero-length
2614 If @option{-Wformat} is specified, do not warn about zero-length formats.
2615 The C standard specifies that zero-length formats are allowed.
2617 @item -Wformat-nonliteral
2618 @opindex Wformat-nonliteral
2619 If @option{-Wformat} is specified, also warn if the format string is not a
2620 string literal and so cannot be checked, unless the format function
2621 takes its format arguments as a @code{va_list}.
2623 @item -Wformat-security
2624 @opindex Wformat-security
2625 If @option{-Wformat} is specified, also warn about uses of format
2626 functions that represent possible security problems. At present, this
2627 warns about calls to @code{printf} and @code{scanf} functions where the
2628 format string is not a string literal and there are no format arguments,
2629 as in @code{printf (foo);}. This may be a security hole if the format
2630 string came from untrusted input and contains @samp{%n}. (This is
2631 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2632 in future warnings may be added to @option{-Wformat-security} that are not
2633 included in @option{-Wformat-nonliteral}.)
2637 Enable @option{-Wformat} plus format checks not included in
2638 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2639 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2643 Warn about passing a null pointer for arguments marked as
2644 requiring a non-null value by the @code{nonnull} function attribute.
2646 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2647 can be disabled with the @option{-Wno-nonnull} option.
2649 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2651 Warn about uninitialized variables which are initialized with themselves.
2652 Note this option can only be used with the @option{-Wuninitialized} option,
2653 which in turn only works with @option{-O1} and above.
2655 For example, GCC will warn about @code{i} being uninitialized in the
2656 following snippet only when @option{-Winit-self} has been specified:
2667 @item -Wimplicit-int
2668 @opindex Wimplicit-int
2669 Warn when a declaration does not specify a type.
2670 This warning is enabled by @option{-Wall}.
2672 @item -Wimplicit-function-declaration
2673 @opindex Wimplicit-function-declaration
2674 @opindex Wno-implicit-function-declaration
2675 Give a warning whenever a function is used before being declared. In
2676 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2677 enabled by default and it is made into an error by
2678 @option{-pedantic-errors}. This warning is also enabled by
2683 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2684 This warning is enabled by @option{-Wall}.
2688 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2689 function with external linkage, returning int, taking either zero
2690 arguments, two, or three arguments of appropriate types.
2691 This warning is enabled by @option{-Wall}.
2693 @item -Wmissing-braces
2694 @opindex Wmissing-braces
2695 Warn if an aggregate or union initializer is not fully bracketed. In
2696 the following example, the initializer for @samp{a} is not fully
2697 bracketed, but that for @samp{b} is fully bracketed.
2700 int a[2][2] = @{ 0, 1, 2, 3 @};
2701 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2704 This warning is enabled by @option{-Wall}.
2706 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2707 @opindex Wmissing-include-dirs
2708 Warn if a user-supplied include directory does not exist.
2711 @opindex Wparentheses
2712 Warn if parentheses are omitted in certain contexts, such
2713 as when there is an assignment in a context where a truth value
2714 is expected, or when operators are nested whose precedence people
2715 often get confused about.
2717 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2718 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2719 interpretation from that of ordinary mathematical notation.
2721 Also warn about constructions where there may be confusion to which
2722 @code{if} statement an @code{else} branch belongs. Here is an example of
2737 In C/C++, every @code{else} branch belongs to the innermost possible
2738 @code{if} statement, which in this example is @code{if (b)}. This is
2739 often not what the programmer expected, as illustrated in the above
2740 example by indentation the programmer chose. When there is the
2741 potential for this confusion, GCC will issue a warning when this flag
2742 is specified. To eliminate the warning, add explicit braces around
2743 the innermost @code{if} statement so there is no way the @code{else}
2744 could belong to the enclosing @code{if}. The resulting code would
2761 This warning is enabled by @option{-Wall}.
2763 @item -Wsequence-point
2764 @opindex Wsequence-point
2765 Warn about code that may have undefined semantics because of violations
2766 of sequence point rules in the C and C++ standards.
2768 The C and C++ standards defines the order in which expressions in a C/C++
2769 program are evaluated in terms of @dfn{sequence points}, which represent
2770 a partial ordering between the execution of parts of the program: those
2771 executed before the sequence point, and those executed after it. These
2772 occur after the evaluation of a full expression (one which is not part
2773 of a larger expression), after the evaluation of the first operand of a
2774 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2775 function is called (but after the evaluation of its arguments and the
2776 expression denoting the called function), and in certain other places.
2777 Other than as expressed by the sequence point rules, the order of
2778 evaluation of subexpressions of an expression is not specified. All
2779 these rules describe only a partial order rather than a total order,
2780 since, for example, if two functions are called within one expression
2781 with no sequence point between them, the order in which the functions
2782 are called is not specified. However, the standards committee have
2783 ruled that function calls do not overlap.
2785 It is not specified when between sequence points modifications to the
2786 values of objects take effect. Programs whose behavior depends on this
2787 have undefined behavior; the C and C++ standards specify that ``Between
2788 the previous and next sequence point an object shall have its stored
2789 value modified at most once by the evaluation of an expression.
2790 Furthermore, the prior value shall be read only to determine the value
2791 to be stored.''. If a program breaks these rules, the results on any
2792 particular implementation are entirely unpredictable.
2794 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2795 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2796 diagnosed by this option, and it may give an occasional false positive
2797 result, but in general it has been found fairly effective at detecting
2798 this sort of problem in programs.
2800 The standard is worded confusingly, therefore there is some debate
2801 over the precise meaning of the sequence point rules in subtle cases.
2802 Links to discussions of the problem, including proposed formal
2803 definitions, may be found on the GCC readings page, at
2804 @w{@uref{http://gcc.gnu.org/readings.html}}.
2806 This warning is enabled by @option{-Wall} for C and C++.
2809 @opindex Wreturn-type
2810 Warn whenever a function is defined with a return-type that defaults to
2811 @code{int}. Also warn about any @code{return} statement with no
2812 return-value in a function whose return-type is not @code{void}.
2814 Also warn if the return type of a function has a type qualifier
2815 such as @code{const}. For ISO C such a type qualifier has no effect,
2816 since the value returned by a function is not an lvalue.
2817 For C++, the warning is only emitted for scalar types or @code{void}.
2818 ISO C prohibits qualified @code{void} return types on function
2819 definitions, so such return types always receive a warning
2820 even without this option.
2822 For C++, a function without return type always produces a diagnostic
2823 message, even when @option{-Wno-return-type} is specified. The only
2824 exceptions are @samp{main} and functions defined in system headers.
2826 This warning is enabled by @option{-Wall}.
2830 Warn whenever a @code{switch} statement has an index of enumerated type
2831 and lacks a @code{case} for one or more of the named codes of that
2832 enumeration. (The presence of a @code{default} label prevents this
2833 warning.) @code{case} labels outside the enumeration range also
2834 provoke warnings when this option is used.
2835 This warning is enabled by @option{-Wall}.
2837 @item -Wswitch-default
2838 @opindex Wswitch-switch
2839 Warn whenever a @code{switch} statement does not have a @code{default}
2843 @opindex Wswitch-enum
2844 Warn whenever a @code{switch} statement has an index of enumerated type
2845 and lacks a @code{case} for one or more of the named codes of that
2846 enumeration. @code{case} labels outside the enumeration range also
2847 provoke warnings when this option is used.
2851 Warn if any trigraphs are encountered that might change the meaning of
2852 the program (trigraphs within comments are not warned about).
2853 This warning is enabled by @option{-Wall}.
2855 @item -Wunused-function
2856 @opindex Wunused-function
2857 Warn whenever a static function is declared but not defined or a
2858 non-inline static function is unused.
2859 This warning is enabled by @option{-Wall}.
2861 @item -Wunused-label
2862 @opindex Wunused-label
2863 Warn whenever a label is declared but not used.
2864 This warning is enabled by @option{-Wall}.
2866 To suppress this warning use the @samp{unused} attribute
2867 (@pxref{Variable Attributes}).
2869 @item -Wunused-parameter
2870 @opindex Wunused-parameter
2871 Warn whenever a function parameter is unused aside from its declaration.
2873 To suppress this warning use the @samp{unused} attribute
2874 (@pxref{Variable Attributes}).
2876 @item -Wunused-variable
2877 @opindex Wunused-variable
2878 Warn whenever a local variable or non-constant static variable is unused
2879 aside from its declaration.
2880 This warning is enabled by @option{-Wall}.
2882 To suppress this warning use the @samp{unused} attribute
2883 (@pxref{Variable Attributes}).
2885 @item -Wunused-value
2886 @opindex Wunused-value
2887 Warn whenever a statement computes a result that is explicitly not
2888 used. To suppress this warning cast the unused expression to
2889 @samp{void}. This includes an expression-statement or the left-hand
2890 side of a comma expression that contains no side effects. For example,
2891 an expression such as @samp{x[i,j]} will cause a warning, while
2892 @samp{x[(void)i,j]} will not.
2894 This warning is enabled by @option{-Wall}.
2898 All the above @option{-Wunused} options combined.
2900 In order to get a warning about an unused function parameter, you must
2901 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2902 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2904 @item -Wuninitialized
2905 @opindex Wuninitialized
2906 Warn if an automatic variable is used without first being initialized or
2907 if a variable may be clobbered by a @code{setjmp} call.
2909 These warnings are possible only in optimizing compilation,
2910 because they require data flow information that is computed only
2911 when optimizing. If you do not specify @option{-O}, you will not get
2912 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2913 requiring @option{-O}.
2915 If you want to warn about code which uses the uninitialized value of the
2916 variable in its own initializer, use the @option{-Winit-self} option.
2918 These warnings occur for individual uninitialized or clobbered
2919 elements of structure, union or array variables as well as for
2920 variables which are uninitialized or clobbered as a whole. They do
2921 not occur for variables or elements declared @code{volatile}. Because
2922 these warnings depend on optimization, the exact variables or elements
2923 for which there are warnings will depend on the precise optimization
2924 options and version of GCC used.
2926 Note that there may be no warning about a variable that is used only
2927 to compute a value that itself is never used, because such
2928 computations may be deleted by data flow analysis before the warnings
2931 These warnings are made optional because GCC is not smart
2932 enough to see all the reasons why the code might be correct
2933 despite appearing to have an error. Here is one example of how
2954 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2955 always initialized, but GCC doesn't know this. Here is
2956 another common case:
2961 if (change_y) save_y = y, y = new_y;
2963 if (change_y) y = save_y;
2968 This has no bug because @code{save_y} is used only if it is set.
2970 @cindex @code{longjmp} warnings
2971 This option also warns when a non-volatile automatic variable might be
2972 changed by a call to @code{longjmp}. These warnings as well are possible
2973 only in optimizing compilation.
2975 The compiler sees only the calls to @code{setjmp}. It cannot know
2976 where @code{longjmp} will be called; in fact, a signal handler could
2977 call it at any point in the code. As a result, you may get a warning
2978 even when there is in fact no problem because @code{longjmp} cannot
2979 in fact be called at the place which would cause a problem.
2981 Some spurious warnings can be avoided if you declare all the functions
2982 you use that never return as @code{noreturn}. @xref{Function
2985 This warning is enabled by @option{-Wall}.
2987 @item -Wunknown-pragmas
2988 @opindex Wunknown-pragmas
2989 @cindex warning for unknown pragmas
2990 @cindex unknown pragmas, warning
2991 @cindex pragmas, warning of unknown
2992 Warn when a #pragma directive is encountered which is not understood by
2993 GCC@. If this command line option is used, warnings will even be issued
2994 for unknown pragmas in system header files. This is not the case if
2995 the warnings were only enabled by the @option{-Wall} command line option.
2998 @opindex Wno-pragmas
3000 Do not warn about misuses of pragmas, such as incorrect parameters,
3001 invalid syntax, or conflicts between pragmas. See also
3002 @samp{-Wunknown-pragmas}.
3004 @item -Wstrict-aliasing
3005 @opindex Wstrict-aliasing
3006 This option is only active when @option{-fstrict-aliasing} is active.
3007 It warns about code which might break the strict aliasing rules that the
3008 compiler is using for optimization. The warning does not catch all
3009 cases, but does attempt to catch the more common pitfalls. It is
3010 included in @option{-Wall}.
3012 @item -Wstrict-aliasing=2
3013 @opindex Wstrict-aliasing=2
3014 This option is only active when @option{-fstrict-aliasing} is active.
3015 It warns about code which might break the strict aliasing rules that the
3016 compiler is using for optimization. This warning catches more cases than
3017 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
3018 cases that are safe.
3020 @item -Wstrict-overflow
3021 @item -Wstrict-overflow=@var{n}
3022 @opindex Wstrict-overflow
3023 This option is only active when @option{-fstrict-overflow} is active.
3024 It warns about cases where the compiler optimizes based on the
3025 assumption that signed overflow does not occur. Note that it does not
3026 warn about all cases where the code might overflow: it only warns
3027 about cases where the compiler implements some optimization. Thus
3028 this warning depends on the optimization level.
3030 An optimization which assumes that signed overflow does not occur is
3031 perfectly safe if the values of the variables involved are such that
3032 overflow never does, in fact, occur. Therefore this warning can
3033 easily give a false positive: a warning about code which is not
3034 actually a problem. To help focus on important issues, several
3035 warning levels are defined. No warnings are issued for the use of
3036 undefined signed overflow when estimating how many iterations a loop
3037 will require, in particular when determining whether a loop will be
3041 @item -Wstrict-overflow=1
3042 Warn about cases which are both questionable and easy to avoid. For
3043 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3044 compiler will simplify this to @code{1}. This level of
3045 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3046 are not, and must be explicitly requested.
3048 @item -Wstrict-overflow=2
3049 Also warn about other cases where a comparison is simplified to a
3050 constant. For example: @code{abs (x) >= 0}. This can only be
3051 simplified when @option{-fstrict-overflow} is in effect, because
3052 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3053 zero. @option{-Wstrict-overflow} (with no level) is the same as
3054 @option{-Wstrict-overflow=2}.
3056 @item -Wstrict-overflow=3
3057 Also warn about other cases where a comparison is simplified. For
3058 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3060 @item -Wstrict-overflow=4
3061 Also warn about other simplifications not covered by the above cases.
3062 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3064 @item -Wstrict-overflow=5
3065 Also warn about cases where the compiler reduces the magnitude of a
3066 constant involved in a comparison. For example: @code{x + 2 > y} will
3067 be simplified to @code{x + 1 >= y}. This is reported only at the
3068 highest warning level because this simplification applies to many
3069 comparisons, so this warning level will give a very large number of
3073 @item -Warray-bounds
3074 @opindex Wno-array-bounds
3075 @opindex Warray-bounds
3076 This option is only active when @option{-ftree-vrp} is active
3077 (default for -O2 and above). It warns about subscripts to arrays
3078 that are always out of bounds. This warning is enabled by @option{-Wall}.
3082 All of the above @samp{-W} options combined. This enables all the
3083 warnings about constructions that some users consider questionable, and
3084 that are easy to avoid (or modify to prevent the warning), even in
3085 conjunction with macros. This also enables some language-specific
3086 warnings described in @ref{C++ Dialect Options} and
3087 @ref{Objective-C and Objective-C++ Dialect Options}.
3090 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
3091 Some of them warn about constructions that users generally do not
3092 consider questionable, but which occasionally you might wish to check
3093 for; others warn about constructions that are necessary or hard to avoid
3094 in some cases, and there is no simple way to modify the code to suppress
3101 (This option used to be called @option{-W}. The older name is still
3102 supported, but the newer name is more descriptive.) Print extra warning
3103 messages for these events:
3107 A function can return either with or without a value. (Falling
3108 off the end of the function body is considered returning without
3109 a value.) For example, this function would evoke such a
3123 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
3126 Storage-class specifiers like @code{static} are not the first things
3127 in a declaration. According to the C Standard, this usage is
3128 obsolescent. This warning can be independently controlled by
3129 @option{-Wold-style-declaration}.
3132 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
3136 A comparison between signed and unsigned values could produce an
3137 incorrect result when the signed value is converted to unsigned.
3138 (But don't warn if @option{-Wno-sign-compare} is also specified.)
3141 An aggregate has an initializer which does not initialize all members.
3142 This warning can be independently controlled by
3143 @option{-Wmissing-field-initializers}.
3146 An initialized field without side effects is overridden when using
3147 designated initializers (@pxref{Designated Inits, , Designated
3148 Initializers}). This warning can be independently controlled by
3149 @option{-Woverride-init}.
3152 A function parameter is declared without a type specifier in K&R-style
3153 functions. This warning can be independently controlled by
3154 @option{-Wmissing-parameter-type}.
3157 An empty body occurs in an @samp{if}, @samp{else} or
3158 @samp{do while} statement. This warning can be independently
3159 controlled by @option{-Wempty-body}.
3161 @item @r{(C++ only)}
3162 An empty body occurs in a @samp{while} or @samp{for} statement with no
3163 whitespacing before the semicolon. This warning can be independently
3164 controlled by @option{-Wempty-body}.
3167 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3168 @samp{>}, or @samp{>=}.
3171 A variable might be changed by @samp{longjmp} or @samp{vfork}.
3172 This warning can be independently controlled by @option{-Wclobbered}.
3174 @item @r{(C++ only)}
3175 An enumerator and a non-enumerator both appear in a conditional expression.
3177 @item @r{(C++ only)}
3178 A non-static reference or non-static @samp{const} member appears in a
3179 class without constructors.
3181 @item @r{(C++ only)}
3182 Ambiguous virtual bases.
3184 @item @r{(C++ only)}
3185 Subscripting an array which has been declared @samp{register}.
3187 @item @r{(C++ only)}
3188 Taking the address of a variable which has been declared @samp{register}.
3190 @item @r{(C++ only)}
3191 A base class is not initialized in a derived class' copy constructor.
3194 @item -Wno-div-by-zero
3195 @opindex Wno-div-by-zero
3196 @opindex Wdiv-by-zero
3197 Do not warn about compile-time integer division by zero. Floating point
3198 division by zero is not warned about, as it can be a legitimate way of
3199 obtaining infinities and NaNs.
3201 @item -Wsystem-headers
3202 @opindex Wsystem-headers
3203 @cindex warnings from system headers
3204 @cindex system headers, warnings from
3205 Print warning messages for constructs found in system header files.
3206 Warnings from system headers are normally suppressed, on the assumption
3207 that they usually do not indicate real problems and would only make the
3208 compiler output harder to read. Using this command line option tells
3209 GCC to emit warnings from system headers as if they occurred in user
3210 code. However, note that using @option{-Wall} in conjunction with this
3211 option will @emph{not} warn about unknown pragmas in system
3212 headers---for that, @option{-Wunknown-pragmas} must also be used.
3215 @opindex Wfloat-equal
3216 Warn if floating point values are used in equality comparisons.
3218 The idea behind this is that sometimes it is convenient (for the
3219 programmer) to consider floating-point values as approximations to
3220 infinitely precise real numbers. If you are doing this, then you need
3221 to compute (by analyzing the code, or in some other way) the maximum or
3222 likely maximum error that the computation introduces, and allow for it
3223 when performing comparisons (and when producing output, but that's a
3224 different problem). In particular, instead of testing for equality, you
3225 would check to see whether the two values have ranges that overlap; and
3226 this is done with the relational operators, so equality comparisons are
3229 @item -Wtraditional @r{(C only)}
3230 @opindex Wtraditional
3231 Warn about certain constructs that behave differently in traditional and
3232 ISO C@. Also warn about ISO C constructs that have no traditional C
3233 equivalent, and/or problematic constructs which should be avoided.
3237 Macro parameters that appear within string literals in the macro body.
3238 In traditional C macro replacement takes place within string literals,
3239 but does not in ISO C@.
3242 In traditional C, some preprocessor directives did not exist.
3243 Traditional preprocessors would only consider a line to be a directive
3244 if the @samp{#} appeared in column 1 on the line. Therefore
3245 @option{-Wtraditional} warns about directives that traditional C
3246 understands but would ignore because the @samp{#} does not appear as the
3247 first character on the line. It also suggests you hide directives like
3248 @samp{#pragma} not understood by traditional C by indenting them. Some
3249 traditional implementations would not recognize @samp{#elif}, so it
3250 suggests avoiding it altogether.
3253 A function-like macro that appears without arguments.
3256 The unary plus operator.
3259 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3260 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3261 constants.) Note, these suffixes appear in macros defined in the system
3262 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3263 Use of these macros in user code might normally lead to spurious
3264 warnings, however GCC's integrated preprocessor has enough context to
3265 avoid warning in these cases.
3268 A function declared external in one block and then used after the end of
3272 A @code{switch} statement has an operand of type @code{long}.
3275 A non-@code{static} function declaration follows a @code{static} one.
3276 This construct is not accepted by some traditional C compilers.
3279 The ISO type of an integer constant has a different width or
3280 signedness from its traditional type. This warning is only issued if
3281 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3282 typically represent bit patterns, are not warned about.
3285 Usage of ISO string concatenation is detected.
3288 Initialization of automatic aggregates.
3291 Identifier conflicts with labels. Traditional C lacks a separate
3292 namespace for labels.
3295 Initialization of unions. If the initializer is zero, the warning is
3296 omitted. This is done under the assumption that the zero initializer in
3297 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3298 initializer warnings and relies on default initialization to zero in the
3302 Conversions by prototypes between fixed/floating point values and vice
3303 versa. The absence of these prototypes when compiling with traditional
3304 C would cause serious problems. This is a subset of the possible
3305 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3308 Use of ISO C style function definitions. This warning intentionally is
3309 @emph{not} issued for prototype declarations or variadic functions
3310 because these ISO C features will appear in your code when using
3311 libiberty's traditional C compatibility macros, @code{PARAMS} and
3312 @code{VPARAMS}. This warning is also bypassed for nested functions
3313 because that feature is already a GCC extension and thus not relevant to
3314 traditional C compatibility.
3317 @item -Wtraditional-conversion @r{(C only)}
3318 @opindex Wtraditional-conversion
3319 Warn if a prototype causes a type conversion that is different from what
3320 would happen to the same argument in the absence of a prototype. This
3321 includes conversions of fixed point to floating and vice versa, and
3322 conversions changing the width or signedness of a fixed point argument
3323 except when the same as the default promotion.
3325 @item -Wdeclaration-after-statement @r{(C only)}
3326 @opindex Wdeclaration-after-statement
3327 Warn when a declaration is found after a statement in a block. This
3328 construct, known from C++, was introduced with ISO C99 and is by default
3329 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3330 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3334 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3336 @item -Wno-endif-labels
3337 @opindex Wno-endif-labels
3338 @opindex Wendif-labels
3339 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3343 Warn whenever a local variable shadows another local variable, parameter or
3344 global variable or whenever a built-in function is shadowed.
3346 @item -Wlarger-than-@var{len}
3347 @opindex Wlarger-than
3348 Warn whenever an object of larger than @var{len} bytes is defined.
3350 @item -Wunsafe-loop-optimizations
3351 @opindex Wunsafe-loop-optimizations
3352 Warn if the loop cannot be optimized because the compiler could not
3353 assume anything on the bounds of the loop indices. With
3354 @option{-funsafe-loop-optimizations} warn if the compiler made
3357 @item -Wpointer-arith
3358 @opindex Wpointer-arith
3359 Warn about anything that depends on the ``size of'' a function type or
3360 of @code{void}. GNU C assigns these types a size of 1, for
3361 convenience in calculations with @code{void *} pointers and pointers
3362 to functions. In C++, warn also when an arithmetic operation involves
3363 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3365 @item -Wbad-function-cast @r{(C only)}
3366 @opindex Wbad-function-cast
3367 Warn whenever a function call is cast to a non-matching type.
3368 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3371 Warn about ISO C constructs that are outside of the common subset of
3372 ISO C and ISO C++, e.g.@: request for implicit conversion from
3373 @code{void *} to a pointer to non-@code{void} type.
3375 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3376 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3377 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3378 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3382 Warn whenever a pointer is cast so as to remove a type qualifier from
3383 the target type. For example, warn if a @code{const char *} is cast
3384 to an ordinary @code{char *}.
3387 @opindex Wcast-align
3388 Warn whenever a pointer is cast such that the required alignment of the
3389 target is increased. For example, warn if a @code{char *} is cast to
3390 an @code{int *} on machines where integers can only be accessed at
3391 two- or four-byte boundaries.
3393 @item -Wwrite-strings
3394 @opindex Wwrite-strings
3395 When compiling C, give string constants the type @code{const
3396 char[@var{length}]} so that
3397 copying the address of one into a non-@code{const} @code{char *}
3398 pointer will get a warning; when compiling C++, warn about the
3399 deprecated conversion from string literals to @code{char *}. This
3400 warning, by default, is enabled for C++ programs.
3401 These warnings will help you find at
3402 compile time code that can try to write into a string constant, but
3403 only if you have been very careful about using @code{const} in
3404 declarations and prototypes. Otherwise, it will just be a nuisance;
3405 this is why we did not make @option{-Wall} request these warnings.
3409 Warn for variables that might be changed by @samp{longjmp} or
3410 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3413 @opindex Wconversion
3414 Warn for implicit conversions that may alter a value. This includes
3415 conversions between real and integer, like @code{abs (x)} when
3416 @code{x} is @code{double}; conversions between signed and unsigned,
3417 like @code{unsigned ui = -1}; and conversions to smaller types, like
3418 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3419 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3420 changed by the conversion like in @code{abs (2.0)}.
3422 For C++, also warn for conversions between @code{NULL} and non-pointer
3423 types; confusing overload resolution for user-defined conversions; and
3424 conversions that will never use a type conversion operator:
3425 conversions to @code{void}, the same type, a base class or a reference
3429 @opindex Wempty-body
3430 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3431 while} statement. Additionally, in C++, warn when an empty body occurs
3432 in a @samp{while} or @samp{for} statement with no whitespacing before
3433 the semicolon. This warning is also enabled by @option{-Wextra}.
3435 @item -Wsign-compare
3436 @opindex Wsign-compare
3437 @cindex warning for comparison of signed and unsigned values
3438 @cindex comparison of signed and unsigned values, warning
3439 @cindex signed and unsigned values, comparison warning
3440 Warn when a comparison between signed and unsigned values could produce
3441 an incorrect result when the signed value is converted to unsigned.
3442 This warning is also enabled by @option{-Wextra}; to get the other warnings
3443 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3447 @opindex Wno-address
3448 Warn about suspicious uses of memory addresses. These include using
3449 the address of a function in a conditional expression, such as
3450 @code{void func(void); if (func)}, and comparisons against the memory
3451 address of a string literal, such as @code{if (x == "abc")}. Such
3452 uses typically indicate a programmer error: the address of a function
3453 always evaluates to true, so their use in a conditional usually
3454 indicate that the programmer forgot the parentheses in a function
3455 call; and comparisons against string literals result in unspecified
3456 behavior and are not portable in C, so they usually indicate that the
3457 programmer intended to use @code{strcmp}. This warning is enabled by
3461 @opindex Wlogical-op
3462 @opindex Wno-logical-op
3463 Warn about suspicious uses of logical operators in expressions.
3464 This includes using logical operators in contexts where a
3465 bit-wise operator is likely to be expected.
3467 @item -Waggregate-return
3468 @opindex Waggregate-return
3469 Warn if any functions that return structures or unions are defined or
3470 called. (In languages where you can return an array, this also elicits
3473 @item -Wno-attributes
3474 @opindex Wno-attributes
3475 @opindex Wattributes
3476 Do not warn if an unexpected @code{__attribute__} is used, such as
3477 unrecognized attributes, function attributes applied to variables,
3478 etc. This will not stop errors for incorrect use of supported
3481 @item -Wstrict-prototypes @r{(C only)}
3482 @opindex Wstrict-prototypes
3483 Warn if a function is declared or defined without specifying the
3484 argument types. (An old-style function definition is permitted without
3485 a warning if preceded by a declaration which specifies the argument
3488 @item -Wold-style-declaration @r{(C only)}
3489 @opindex Wold-style-declaration
3490 Warn for obsolescent usages, according to the C Standard, in a
3491 declaration. For example, warn if storage-class specifiers like
3492 @code{static} are not the first things in a declaration. This warning
3493 is also enabled by @option{-Wextra}.
3495 @item -Wold-style-definition @r{(C only)}
3496 @opindex Wold-style-definition
3497 Warn if an old-style function definition is used. A warning is given
3498 even if there is a previous prototype.
3500 @item -Wmissing-parameter-type @r{(C only)}
3501 @opindex Wmissing-parameter-type
3502 A function parameter is declared without a type specifier in K&R-style
3509 This warning is also enabled by @option{-Wextra}.
3511 @item -Wmissing-prototypes @r{(C only)}
3512 @opindex Wmissing-prototypes
3513 Warn if a global function is defined without a previous prototype
3514 declaration. This warning is issued even if the definition itself
3515 provides a prototype. The aim is to detect global functions that fail
3516 to be declared in header files.
3518 @item -Wmissing-declarations @r{(C and C++ only)}
3519 @opindex Wmissing-declarations
3520 Warn if a global function is defined without a previous declaration.
3521 Do so even if the definition itself provides a prototype.
3522 Use this option to detect global functions that are not declared in
3523 header files. In C++, no warnings are issued for function templates,
3524 or for inline functions, or for functions in anonymous namespaces.
3526 @item -Wmissing-field-initializers
3527 @opindex Wmissing-field-initializers
3530 Warn if a structure's initializer has some fields missing. For
3531 example, the following code would cause such a warning, because
3532 @code{x.h} is implicitly zero:
3535 struct s @{ int f, g, h; @};
3536 struct s x = @{ 3, 4 @};
3539 This option does not warn about designated initializers, so the following
3540 modification would not trigger a warning:
3543 struct s @{ int f, g, h; @};
3544 struct s x = @{ .f = 3, .g = 4 @};
3547 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3548 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3550 @item -Wmissing-noreturn
3551 @opindex Wmissing-noreturn
3552 Warn about functions which might be candidates for attribute @code{noreturn}.
3553 Note these are only possible candidates, not absolute ones. Care should
3554 be taken to manually verify functions actually do not ever return before
3555 adding the @code{noreturn} attribute, otherwise subtle code generation
3556 bugs could be introduced. You will not get a warning for @code{main} in
3557 hosted C environments.
3559 @item -Wmissing-format-attribute
3560 @opindex Wmissing-format-attribute
3562 Warn about function pointers which might be candidates for @code{format}
3563 attributes. Note these are only possible candidates, not absolute ones.
3564 GCC will guess that function pointers with @code{format} attributes that
3565 are used in assignment, initialization, parameter passing or return
3566 statements should have a corresponding @code{format} attribute in the
3567 resulting type. I.e.@: the left-hand side of the assignment or
3568 initialization, the type of the parameter variable, or the return type
3569 of the containing function respectively should also have a @code{format}
3570 attribute to avoid the warning.
3572 GCC will also warn about function definitions which might be
3573 candidates for @code{format} attributes. Again, these are only
3574 possible candidates. GCC will guess that @code{format} attributes
3575 might be appropriate for any function that calls a function like
3576 @code{vprintf} or @code{vscanf}, but this might not always be the
3577 case, and some functions for which @code{format} attributes are
3578 appropriate may not be detected.
3580 @item -Wno-multichar
3581 @opindex Wno-multichar
3583 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3584 Usually they indicate a typo in the user's code, as they have
3585 implementation-defined values, and should not be used in portable code.
3587 @item -Wnormalized=<none|id|nfc|nfkc>
3588 @opindex Wnormalized
3591 @cindex character set, input normalization
3592 In ISO C and ISO C++, two identifiers are different if they are
3593 different sequences of characters. However, sometimes when characters
3594 outside the basic ASCII character set are used, you can have two
3595 different character sequences that look the same. To avoid confusion,
3596 the ISO 10646 standard sets out some @dfn{normalization rules} which
3597 when applied ensure that two sequences that look the same are turned into
3598 the same sequence. GCC can warn you if you are using identifiers which
3599 have not been normalized; this option controls that warning.
3601 There are four levels of warning that GCC supports. The default is
3602 @option{-Wnormalized=nfc}, which warns about any identifier which is
3603 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3604 recommended form for most uses.
3606 Unfortunately, there are some characters which ISO C and ISO C++ allow
3607 in identifiers that when turned into NFC aren't allowable as
3608 identifiers. That is, there's no way to use these symbols in portable
3609 ISO C or C++ and have all your identifiers in NFC.
3610 @option{-Wnormalized=id} suppresses the warning for these characters.
3611 It is hoped that future versions of the standards involved will correct
3612 this, which is why this option is not the default.
3614 You can switch the warning off for all characters by writing
3615 @option{-Wnormalized=none}. You would only want to do this if you
3616 were using some other normalization scheme (like ``D''), because
3617 otherwise you can easily create bugs that are literally impossible to see.
3619 Some characters in ISO 10646 have distinct meanings but look identical
3620 in some fonts or display methodologies, especially once formatting has
3621 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3622 LETTER N'', will display just like a regular @code{n} which has been
3623 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3624 normalization scheme to convert all these into a standard form as
3625 well, and GCC will warn if your code is not in NFKC if you use
3626 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3627 about every identifier that contains the letter O because it might be
3628 confused with the digit 0, and so is not the default, but may be
3629 useful as a local coding convention if the programming environment is
3630 unable to be fixed to display these characters distinctly.
3632 @item -Wno-deprecated-declarations
3633 @opindex Wno-deprecated-declarations
3634 Do not warn about uses of functions (@pxref{Function Attributes}),
3635 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3636 Attributes}) marked as deprecated by using the @code{deprecated}
3640 @opindex Wno-overflow
3641 Do not warn about compile-time overflow in constant expressions.
3643 @item -Woverride-init
3644 @opindex Woverride-init
3647 Warn if an initialized field without side effects is overridden when
3648 using designated initializers (@pxref{Designated Inits, , Designated
3651 This warning is included in @option{-Wextra}. To get other
3652 @option{-Wextra} warnings without this one, use @samp{-Wextra
3653 -Wno-override-init}.
3657 Warn if a structure is given the packed attribute, but the packed
3658 attribute has no effect on the layout or size of the structure.
3659 Such structures may be mis-aligned for little benefit. For
3660 instance, in this code, the variable @code{f.x} in @code{struct bar}
3661 will be misaligned even though @code{struct bar} does not itself
3662 have the packed attribute:
3669 @} __attribute__((packed));
3679 Warn if padding is included in a structure, either to align an element
3680 of the structure or to align the whole structure. Sometimes when this
3681 happens it is possible to rearrange the fields of the structure to
3682 reduce the padding and so make the structure smaller.
3684 @item -Wredundant-decls
3685 @opindex Wredundant-decls
3686 Warn if anything is declared more than once in the same scope, even in
3687 cases where multiple declaration is valid and changes nothing.
3689 @item -Wnested-externs @r{(C only)}
3690 @opindex Wnested-externs
3691 Warn if an @code{extern} declaration is encountered within a function.
3693 @item -Wunreachable-code
3694 @opindex Wunreachable-code
3695 Warn if the compiler detects that code will never be executed.
3697 This option is intended to warn when the compiler detects that at
3698 least a whole line of source code will never be executed, because
3699 some condition is never satisfied or because it is after a
3700 procedure that never returns.
3702 It is possible for this option to produce a warning even though there
3703 are circumstances under which part of the affected line can be executed,
3704 so care should be taken when removing apparently-unreachable code.
3706 For instance, when a function is inlined, a warning may mean that the
3707 line is unreachable in only one inlined copy of the function.
3709 This option is not made part of @option{-Wall} because in a debugging
3710 version of a program there is often substantial code which checks
3711 correct functioning of the program and is, hopefully, unreachable
3712 because the program does work. Another common use of unreachable
3713 code is to provide behavior which is selectable at compile-time.
3717 Warn if a function can not be inlined and it was declared as inline.
3718 Even with this option, the compiler will not warn about failures to
3719 inline functions declared in system headers.
3721 The compiler uses a variety of heuristics to determine whether or not
3722 to inline a function. For example, the compiler takes into account
3723 the size of the function being inlined and the amount of inlining
3724 that has already been done in the current function. Therefore,
3725 seemingly insignificant changes in the source program can cause the
3726 warnings produced by @option{-Winline} to appear or disappear.
3728 @item -Wno-invalid-offsetof @r{(C++ only)}
3729 @opindex Wno-invalid-offsetof
3730 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3731 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3732 to a non-POD type is undefined. In existing C++ implementations,
3733 however, @samp{offsetof} typically gives meaningful results even when
3734 applied to certain kinds of non-POD types. (Such as a simple
3735 @samp{struct} that fails to be a POD type only by virtue of having a
3736 constructor.) This flag is for users who are aware that they are
3737 writing nonportable code and who have deliberately chosen to ignore the
3740 The restrictions on @samp{offsetof} may be relaxed in a future version
3741 of the C++ standard.
3743 @item -Wno-int-to-pointer-cast @r{(C only)}
3744 @opindex Wno-int-to-pointer-cast
3745 Suppress warnings from casts to pointer type of an integer of a
3748 @item -Wno-pointer-to-int-cast @r{(C only)}
3749 @opindex Wno-pointer-to-int-cast
3750 Suppress warnings from casts from a pointer to an integer type of a
3754 @opindex Winvalid-pch
3755 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3756 the search path but can't be used.
3760 @opindex Wno-long-long
3761 Warn if @samp{long long} type is used. This is default. To inhibit
3762 the warning messages, use @option{-Wno-long-long}. Flags
3763 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3764 only when @option{-pedantic} flag is used.
3766 @item -Wvariadic-macros
3767 @opindex Wvariadic-macros
3768 @opindex Wno-variadic-macros
3769 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3770 alternate syntax when in pedantic ISO C99 mode. This is default.
3771 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3776 Warn if variable length array is used in the code.
3777 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3778 the variable length array.
3780 @item -Wvolatile-register-var
3781 @opindex Wvolatile-register-var
3782 @opindex Wno-volatile-register-var
3783 Warn if a register variable is declared volatile. The volatile
3784 modifier does not inhibit all optimizations that may eliminate reads
3785 and/or writes to register variables.
3787 @item -Wdisabled-optimization
3788 @opindex Wdisabled-optimization
3789 Warn if a requested optimization pass is disabled. This warning does
3790 not generally indicate that there is anything wrong with your code; it
3791 merely indicates that GCC's optimizers were unable to handle the code
3792 effectively. Often, the problem is that your code is too big or too
3793 complex; GCC will refuse to optimize programs when the optimization
3794 itself is likely to take inordinate amounts of time.
3796 @item -Wpointer-sign
3797 @opindex Wpointer-sign
3798 @opindex Wno-pointer-sign
3799 Warn for pointer argument passing or assignment with different signedness.
3800 This option is only supported for C and Objective-C@. It is implied by
3801 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3802 @option{-Wno-pointer-sign}.
3806 Make all warnings into errors.
3810 Make the specified warning into an errors. The specifier for a
3811 warning is appended, for example @option{-Werror=switch} turns the
3812 warnings controlled by @option{-Wswitch} into errors. This switch
3813 takes a negative form, to be used to negate @option{-Werror} for
3814 specific warnings, for example @option{-Wno-error=switch} makes
3815 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3816 is in effect. You can use the @option{-fdiagnostics-show-option}
3817 option to have each controllable warning amended with the option which
3818 controls it, to determine what to use with this option.
3820 Note that specifying @option{-Werror=}@var{foo} automatically implies
3821 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3824 @item -Wstack-protector
3825 @opindex Wstack-protector
3826 This option is only active when @option{-fstack-protector} is active. It
3827 warns about functions that will not be protected against stack smashing.
3829 @item -Woverlength-strings
3830 @opindex Woverlength-strings
3831 Warn about string constants which are longer than the ``minimum
3832 maximum'' length specified in the C standard. Modern compilers
3833 generally allow string constants which are much longer than the
3834 standard's minimum limit, but very portable programs should avoid
3835 using longer strings.
3837 The limit applies @emph{after} string constant concatenation, and does
3838 not count the trailing NUL@. In C89, the limit was 509 characters; in
3839 C99, it was raised to 4095. C++98 does not specify a normative
3840 minimum maximum, so we do not diagnose overlength strings in C++@.
3842 This option is implied by @option{-pedantic}, and can be disabled with
3843 @option{-Wno-overlength-strings}.
3846 @node Debugging Options
3847 @section Options for Debugging Your Program or GCC
3848 @cindex options, debugging
3849 @cindex debugging information options
3851 GCC has various special options that are used for debugging
3852 either your program or GCC:
3857 Produce debugging information in the operating system's native format
3858 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3861 On most systems that use stabs format, @option{-g} enables use of extra
3862 debugging information that only GDB can use; this extra information
3863 makes debugging work better in GDB but will probably make other debuggers
3865 refuse to read the program. If you want to control for certain whether
3866 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3867 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3869 GCC allows you to use @option{-g} with
3870 @option{-O}. The shortcuts taken by optimized code may occasionally
3871 produce surprising results: some variables you declared may not exist
3872 at all; flow of control may briefly move where you did not expect it;
3873 some statements may not be executed because they compute constant
3874 results or their values were already at hand; some statements may
3875 execute in different places because they were moved out of loops.
3877 Nevertheless it proves possible to debug optimized output. This makes
3878 it reasonable to use the optimizer for programs that might have bugs.
3880 The following options are useful when GCC is generated with the
3881 capability for more than one debugging format.
3885 Produce debugging information for use by GDB@. This means to use the
3886 most expressive format available (DWARF 2, stabs, or the native format
3887 if neither of those are supported), including GDB extensions if at all
3892 Produce debugging information in stabs format (if that is supported),
3893 without GDB extensions. This is the format used by DBX on most BSD
3894 systems. On MIPS, Alpha and System V Release 4 systems this option
3895 produces stabs debugging output which is not understood by DBX or SDB@.
3896 On System V Release 4 systems this option requires the GNU assembler.
3898 @item -feliminate-unused-debug-symbols
3899 @opindex feliminate-unused-debug-symbols
3900 Produce debugging information in stabs format (if that is supported),
3901 for only symbols that are actually used.
3903 @item -femit-class-debug-always
3904 Instead of emitting debugging information for a C++ class in only one
3905 object file, emit it in all object files using the class. This option
3906 should be used only with debuggers that are unable to handle the way GCC
3907 normally emits debugging information for classes because using this
3908 option will increase the size of debugging information by as much as a
3913 Produce debugging information in stabs format (if that is supported),
3914 using GNU extensions understood only by the GNU debugger (GDB)@. The
3915 use of these extensions is likely to make other debuggers crash or
3916 refuse to read the program.
3920 Produce debugging information in COFF format (if that is supported).
3921 This is the format used by SDB on most System V systems prior to
3926 Produce debugging information in XCOFF format (if that is supported).
3927 This is the format used by the DBX debugger on IBM RS/6000 systems.
3931 Produce debugging information in XCOFF format (if that is supported),
3932 using GNU extensions understood only by the GNU debugger (GDB)@. The
3933 use of these extensions is likely to make other debuggers crash or
3934 refuse to read the program, and may cause assemblers other than the GNU
3935 assembler (GAS) to fail with an error.
3939 Produce debugging information in DWARF version 2 format (if that is
3940 supported). This is the format used by DBX on IRIX 6. With this
3941 option, GCC uses features of DWARF version 3 when they are useful;
3942 version 3 is upward compatible with version 2, but may still cause
3943 problems for older debuggers.
3947 Produce debugging information in VMS debug format (if that is
3948 supported). This is the format used by DEBUG on VMS systems.
3951 @itemx -ggdb@var{level}
3952 @itemx -gstabs@var{level}
3953 @itemx -gcoff@var{level}
3954 @itemx -gxcoff@var{level}
3955 @itemx -gvms@var{level}
3956 Request debugging information and also use @var{level} to specify how
3957 much information. The default level is 2.
3959 Level 0 produces no debug information at all. Thus, @option{-g0} negates
3962 Level 1 produces minimal information, enough for making backtraces in
3963 parts of the program that you don't plan to debug. This includes
3964 descriptions of functions and external variables, but no information
3965 about local variables and no line numbers.
3967 Level 3 includes extra information, such as all the macro definitions
3968 present in the program. Some debuggers support macro expansion when
3969 you use @option{-g3}.
3971 @option{-gdwarf-2} does not accept a concatenated debug level, because
3972 GCC used to support an option @option{-gdwarf} that meant to generate
3973 debug information in version 1 of the DWARF format (which is very
3974 different from version 2), and it would have been too confusing. That
3975 debug format is long obsolete, but the option cannot be changed now.
3976 Instead use an additional @option{-g@var{level}} option to change the
3977 debug level for DWARF2.
3979 @item -feliminate-dwarf2-dups
3980 @opindex feliminate-dwarf2-dups
3981 Compress DWARF2 debugging information by eliminating duplicated
3982 information about each symbol. This option only makes sense when
3983 generating DWARF2 debugging information with @option{-gdwarf-2}.
3985 @item -femit-struct-debug-baseonly
3986 Emit debug information for struct-like types
3987 only when the base name of the compilation source file
3988 matches the base name of file in which the struct was defined.
3990 This option substantially reduces the size of debugging information,
3991 but at significant potential loss in type information to the debugger.
3992 See @option{-femit-struct-debug-reduced} for a less aggressive option.
3993 See @option{-femit-struct-debug-detailed} for more detailed control.
3995 This option works only with DWARF 2.
3997 @item -femit-struct-debug-reduced
3998 Emit debug information for struct-like types
3999 only when the base name of the compilation source file
4000 matches the base name of file in which the type was defined,
4001 unless the struct is a template or defined in a system header.
4003 This option significantly reduces the size of debugging information,
4004 with some potential loss in type information to the debugger.
4005 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4006 See @option{-femit-struct-debug-detailed} for more detailed control.
4008 This option works only with DWARF 2.
4010 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4011 Specify the struct-like types
4012 for which the compiler will generate debug information.
4013 The intent is to reduce duplicate struct debug information
4014 between different object files within the same program.
4016 This option is a detailed version of
4017 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4018 which will serve for most needs.
4020 A specification has the syntax
4021 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4023 The optional first word limits the specification to
4024 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4025 A struct type is used directly when it is the type of a variable, member.
4026 Indirect uses arise through pointers to structs.
4027 That is, when use of an incomplete struct would be legal, the use is indirect.
4029 @samp{struct one direct; struct two * indirect;}.
4031 The optional second word limits the specification to
4032 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4033 Generic structs are a bit complicated to explain.
4034 For C++, these are non-explicit specializations of template classes,
4035 or non-template classes within the above.
4036 Other programming languages have generics,
4037 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4039 The third word specifies the source files for those
4040 structs for which the compiler will emit debug information.
4041 The values @samp{none} and @samp{any} have the normal meaning.
4042 The value @samp{base} means that
4043 the base of name of the file in which the type declaration appears
4044 must match the base of the name of the main compilation file.
4045 In practice, this means that
4046 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4047 but types declared in other header will not.
4048 The value @samp{sys} means those types satisfying @samp{base}
4049 or declared in system or compiler headers.
4051 You may need to experiment to determine the best settings for your application.
4053 The default is @samp{-femit-struct-debug-detailed=all}.
4055 This option works only with DWARF 2.
4057 @cindex @command{prof}
4060 Generate extra code to write profile information suitable for the
4061 analysis program @command{prof}. You must use this option when compiling
4062 the source files you want data about, and you must also use it when
4065 @cindex @command{gprof}
4068 Generate extra code to write profile information suitable for the
4069 analysis program @command{gprof}. You must use this option when compiling
4070 the source files you want data about, and you must also use it when
4075 Makes the compiler print out each function name as it is compiled, and
4076 print some statistics about each pass when it finishes.
4079 @opindex ftime-report
4080 Makes the compiler print some statistics about the time consumed by each
4081 pass when it finishes.
4084 @opindex fmem-report
4085 Makes the compiler print some statistics about permanent memory
4086 allocation when it finishes.
4088 @item -fpre-ipa-mem-report
4089 @opindex fpre-ipa-mem-report
4090 @item -fpost-ipa-mem-report
4091 @opindex fpost-ipa-mem-report
4092 Makes the compiler print some statistics about permanent memory
4093 allocation before or after interprocedural optimization.
4095 @item -fprofile-arcs
4096 @opindex fprofile-arcs
4097 Add code so that program flow @dfn{arcs} are instrumented. During
4098 execution the program records how many times each branch and call is
4099 executed and how many times it is taken or returns. When the compiled
4100 program exits it saves this data to a file called
4101 @file{@var{auxname}.gcda} for each source file. The data may be used for
4102 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4103 test coverage analysis (@option{-ftest-coverage}). Each object file's
4104 @var{auxname} is generated from the name of the output file, if
4105 explicitly specified and it is not the final executable, otherwise it is
4106 the basename of the source file. In both cases any suffix is removed
4107 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4108 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4109 @xref{Cross-profiling}.
4111 @cindex @command{gcov}
4115 This option is used to compile and link code instrumented for coverage
4116 analysis. The option is a synonym for @option{-fprofile-arcs}
4117 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4118 linking). See the documentation for those options for more details.
4123 Compile the source files with @option{-fprofile-arcs} plus optimization
4124 and code generation options. For test coverage analysis, use the
4125 additional @option{-ftest-coverage} option. You do not need to profile
4126 every source file in a program.
4129 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4130 (the latter implies the former).
4133 Run the program on a representative workload to generate the arc profile
4134 information. This may be repeated any number of times. You can run
4135 concurrent instances of your program, and provided that the file system
4136 supports locking, the data files will be correctly updated. Also
4137 @code{fork} calls are detected and correctly handled (double counting
4141 For profile-directed optimizations, compile the source files again with
4142 the same optimization and code generation options plus
4143 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4144 Control Optimization}).
4147 For test coverage analysis, use @command{gcov} to produce human readable
4148 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4149 @command{gcov} documentation for further information.
4153 With @option{-fprofile-arcs}, for each function of your program GCC
4154 creates a program flow graph, then finds a spanning tree for the graph.
4155 Only arcs that are not on the spanning tree have to be instrumented: the
4156 compiler adds code to count the number of times that these arcs are
4157 executed. When an arc is the only exit or only entrance to a block, the
4158 instrumentation code can be added to the block; otherwise, a new basic
4159 block must be created to hold the instrumentation code.
4162 @item -ftest-coverage
4163 @opindex ftest-coverage
4164 Produce a notes file that the @command{gcov} code-coverage utility
4165 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4166 show program coverage. Each source file's note file is called
4167 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4168 above for a description of @var{auxname} and instructions on how to
4169 generate test coverage data. Coverage data will match the source files
4170 more closely, if you do not optimize.
4172 @item -d@var{letters}
4173 @item -fdump-rtl-@var{pass}
4175 Says to make debugging dumps during compilation at times specified by
4176 @var{letters}. This is used for debugging the RTL-based passes of the
4177 compiler. The file names for most of the dumps are made by appending a
4178 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4179 from the name of the output file, if explicitly specified and it is not
4180 an executable, otherwise it is the basename of the source file.
4182 Most debug dumps can be enabled either passing a letter to the @option{-d}
4183 option, or with a long @option{-fdump-rtl} switch; here are the possible
4184 letters for use in @var{letters} and @var{pass}, and their meanings:
4189 Annotate the assembler output with miscellaneous debugging information.
4192 @itemx -fdump-rtl-bbro
4194 @opindex fdump-rtl-bbro
4195 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4198 @itemx -fdump-rtl-combine
4200 @opindex fdump-rtl-combine
4201 Dump after the RTL instruction combination pass, to the file
4202 @file{@var{file}.129r.combine}.
4205 @itemx -fdump-rtl-ce1
4206 @itemx -fdump-rtl-ce2
4208 @opindex fdump-rtl-ce1
4209 @opindex fdump-rtl-ce2
4210 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
4211 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
4212 and @option{-fdump-rtl-ce2} enable dumping after the second if
4213 conversion, to the file @file{@var{file}.130r.ce2}.
4216 @itemx -fdump-rtl-btl
4217 @itemx -fdump-rtl-dbr
4219 @opindex fdump-rtl-btl
4220 @opindex fdump-rtl-dbr
4221 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
4222 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
4223 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
4224 scheduling, to @file{@var{file}.36.dbr}.
4228 Dump all macro definitions, at the end of preprocessing, in addition to
4232 @itemx -fdump-rtl-ce3
4234 @opindex fdump-rtl-ce3
4235 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4238 @itemx -fdump-rtl-cfg
4239 @itemx -fdump-rtl-life
4241 @opindex fdump-rtl-cfg
4242 @opindex fdump-rtl-life
4243 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
4244 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
4245 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4246 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4249 @itemx -fdump-rtl-greg
4251 @opindex fdump-rtl-greg
4252 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4255 @itemx -fdump-rtl-gcse
4256 @itemx -fdump-rtl-bypass
4258 @opindex fdump-rtl-gcse
4259 @opindex fdump-rtl-bypass
4260 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4261 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
4262 enable dumping after jump bypassing and control flow optimizations, to
4263 @file{@var{file}.115r.bypass}.
4266 @itemx -fdump-rtl-eh
4268 @opindex fdump-rtl-eh
4269 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4272 @itemx -fdump-rtl-sibling
4274 @opindex fdump-rtl-sibling
4275 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4278 @itemx -fdump-rtl-jump
4280 @opindex fdump-rtl-jump
4281 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4284 @itemx -fdump-rtl-stack
4286 @opindex fdump-rtl-stack
4287 Dump after conversion from GCC's "flat register file" registers to the
4288 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4291 @itemx -fdump-rtl-lreg
4293 @opindex fdump-rtl-lreg
4294 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4297 @itemx -fdump-rtl-loop2
4299 @opindex fdump-rtl-loop2
4300 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4301 loop optimization pass, to @file{@var{file}.119r.loop2},
4302 @file{@var{file}.120r.loop2_init},
4303 @file{@var{file}.121r.loop2_invariant}, and
4304 @file{@var{file}.125r.loop2_done}.
4307 @itemx -fdump-rtl-sms
4309 @opindex fdump-rtl-sms
4310 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4313 @itemx -fdump-rtl-mach
4315 @opindex fdump-rtl-mach
4316 Dump after performing the machine dependent reorganization pass, to
4317 @file{@var{file}.155r.mach}.
4320 @itemx -fdump-rtl-rnreg
4322 @opindex fdump-rtl-rnreg
4323 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4326 @itemx -fdump-rtl-regmove
4328 @opindex fdump-rtl-regmove
4329 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4332 @itemx -fdump-rtl-postreload
4334 @opindex fdump-rtl-postreload
4335 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4338 @itemx -fdump-rtl-expand
4340 @opindex fdump-rtl-expand
4341 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4344 @itemx -fdump-rtl-sched2
4346 @opindex fdump-rtl-sched2
4347 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4350 @itemx -fdump-rtl-cse
4352 @opindex fdump-rtl-cse
4353 Dump after CSE (including the jump optimization that sometimes follows
4354 CSE), to @file{@var{file}.113r.cse}.
4357 @itemx -fdump-rtl-sched1
4359 @opindex fdump-rtl-sched1
4360 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4363 @itemx -fdump-rtl-cse2
4365 @opindex fdump-rtl-cse2
4366 Dump after the second CSE pass (including the jump optimization that
4367 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4370 @itemx -fdump-rtl-tracer
4372 @opindex fdump-rtl-tracer
4373 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4376 @itemx -fdump-rtl-vpt
4377 @itemx -fdump-rtl-vartrack
4379 @opindex fdump-rtl-vpt
4380 @opindex fdump-rtl-vartrack
4381 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4382 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4383 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4384 to @file{@var{file}.154r.vartrack}.
4387 @itemx -fdump-rtl-flow2
4389 @opindex fdump-rtl-flow2
4390 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4393 @itemx -fdump-rtl-peephole2
4395 @opindex fdump-rtl-peephole2
4396 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4399 @itemx -fdump-rtl-web
4401 @opindex fdump-rtl-web
4402 Dump after live range splitting, to @file{@var{file}.126r.web}.
4405 @itemx -fdump-rtl-all
4407 @opindex fdump-rtl-all
4408 Produce all the dumps listed above.
4412 Produce a core dump whenever an error occurs.
4416 Print statistics on memory usage, at the end of the run, to
4421 Annotate the assembler output with a comment indicating which
4422 pattern and alternative was used. The length of each instruction is
4427 Dump the RTL in the assembler output as a comment before each instruction.
4428 Also turns on @option{-dp} annotation.
4432 For each of the other indicated dump files (either with @option{-d} or
4433 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4434 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4438 Just generate RTL for a function instead of compiling it. Usually used
4439 with @samp{r} (@option{-fdump-rtl-expand}).
4443 Dump debugging information during parsing, to standard error.
4447 @opindex fdump-noaddr
4448 When doing debugging dumps (see @option{-d} option above), suppress
4449 address output. This makes it more feasible to use diff on debugging
4450 dumps for compiler invocations with different compiler binaries and/or
4451 different text / bss / data / heap / stack / dso start locations.
4453 @item -fdump-unnumbered
4454 @opindex fdump-unnumbered
4455 When doing debugging dumps (see @option{-d} option above), suppress instruction
4456 numbers, line number note and address output. This makes it more feasible to
4457 use diff on debugging dumps for compiler invocations with different
4458 options, in particular with and without @option{-g}.
4460 @item -fdump-translation-unit @r{(C++ only)}
4461 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4462 @opindex fdump-translation-unit
4463 Dump a representation of the tree structure for the entire translation
4464 unit to a file. The file name is made by appending @file{.tu} to the
4465 source file name. If the @samp{-@var{options}} form is used, @var{options}
4466 controls the details of the dump as described for the
4467 @option{-fdump-tree} options.
4469 @item -fdump-class-hierarchy @r{(C++ only)}
4470 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4471 @opindex fdump-class-hierarchy
4472 Dump a representation of each class's hierarchy and virtual function
4473 table layout to a file. The file name is made by appending @file{.class}
4474 to the source file name. If the @samp{-@var{options}} form is used,
4475 @var{options} controls the details of the dump as described for the
4476 @option{-fdump-tree} options.
4478 @item -fdump-ipa-@var{switch}
4480 Control the dumping at various stages of inter-procedural analysis
4481 language tree to a file. The file name is generated by appending a switch
4482 specific suffix to the source file name. The following dumps are possible:
4486 Enables all inter-procedural analysis dumps; currently the only produced
4487 dump is the @samp{cgraph} dump.
4490 Dumps information about call-graph optimization, unused function removal,
4491 and inlining decisions.
4494 @item -fdump-tree-@var{switch}
4495 @itemx -fdump-tree-@var{switch}-@var{options}
4497 Control the dumping at various stages of processing the intermediate
4498 language tree to a file. The file name is generated by appending a switch
4499 specific suffix to the source file name. If the @samp{-@var{options}}
4500 form is used, @var{options} is a list of @samp{-} separated options that
4501 control the details of the dump. Not all options are applicable to all
4502 dumps, those which are not meaningful will be ignored. The following
4503 options are available
4507 Print the address of each node. Usually this is not meaningful as it
4508 changes according to the environment and source file. Its primary use
4509 is for tying up a dump file with a debug environment.
4511 Inhibit dumping of members of a scope or body of a function merely
4512 because that scope has been reached. Only dump such items when they
4513 are directly reachable by some other path. When dumping pretty-printed
4514 trees, this option inhibits dumping the bodies of control structures.
4516 Print a raw representation of the tree. By default, trees are
4517 pretty-printed into a C-like representation.
4519 Enable more detailed dumps (not honored by every dump option).
4521 Enable dumping various statistics about the pass (not honored by every dump
4524 Enable showing basic block boundaries (disabled in raw dumps).
4526 Enable showing virtual operands for every statement.
4528 Enable showing line numbers for statements.
4530 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4532 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4535 The following tree dumps are possible:
4539 Dump before any tree based optimization, to @file{@var{file}.original}.
4542 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4545 Dump after function inlining, to @file{@var{file}.inlined}.
4548 @opindex fdump-tree-gimple
4549 Dump each function before and after the gimplification pass to a file. The
4550 file name is made by appending @file{.gimple} to the source file name.
4553 @opindex fdump-tree-cfg
4554 Dump the control flow graph of each function to a file. The file name is
4555 made by appending @file{.cfg} to the source file name.
4558 @opindex fdump-tree-vcg
4559 Dump the control flow graph of each function to a file in VCG format. The
4560 file name is made by appending @file{.vcg} to the source file name. Note
4561 that if the file contains more than one function, the generated file cannot
4562 be used directly by VCG@. You will need to cut and paste each function's
4563 graph into its own separate file first.
4566 @opindex fdump-tree-ch
4567 Dump each function after copying loop headers. The file name is made by
4568 appending @file{.ch} to the source file name.
4571 @opindex fdump-tree-ssa
4572 Dump SSA related information to a file. The file name is made by appending
4573 @file{.ssa} to the source file name.
4576 @opindex fdump-tree-salias
4577 Dump structure aliasing variable information to a file. This file name
4578 is made by appending @file{.salias} to the source file name.
4581 @opindex fdump-tree-alias
4582 Dump aliasing information for each function. The file name is made by
4583 appending @file{.alias} to the source file name.
4586 @opindex fdump-tree-ccp
4587 Dump each function after CCP@. The file name is made by appending
4588 @file{.ccp} to the source file name.
4591 @opindex fdump-tree-storeccp
4592 Dump each function after STORE-CCP. The file name is made by appending
4593 @file{.storeccp} to the source file name.
4596 @opindex fdump-tree-pre
4597 Dump trees after partial redundancy elimination. The file name is made
4598 by appending @file{.pre} to the source file name.
4601 @opindex fdump-tree-fre
4602 Dump trees after full redundancy elimination. The file name is made
4603 by appending @file{.fre} to the source file name.
4606 @opindex fdump-tree-copyprop
4607 Dump trees after copy propagation. The file name is made
4608 by appending @file{.copyprop} to the source file name.
4610 @item store_copyprop
4611 @opindex fdump-tree-store_copyprop
4612 Dump trees after store copy-propagation. The file name is made
4613 by appending @file{.store_copyprop} to the source file name.
4616 @opindex fdump-tree-dce
4617 Dump each function after dead code elimination. The file name is made by
4618 appending @file{.dce} to the source file name.
4621 @opindex fdump-tree-mudflap
4622 Dump each function after adding mudflap instrumentation. The file name is
4623 made by appending @file{.mudflap} to the source file name.
4626 @opindex fdump-tree-sra
4627 Dump each function after performing scalar replacement of aggregates. The
4628 file name is made by appending @file{.sra} to the source file name.
4631 @opindex fdump-tree-sink
4632 Dump each function after performing code sinking. The file name is made
4633 by appending @file{.sink} to the source file name.
4636 @opindex fdump-tree-dom
4637 Dump each function after applying dominator tree optimizations. The file
4638 name is made by appending @file{.dom} to the source file name.
4641 @opindex fdump-tree-dse
4642 Dump each function after applying dead store elimination. The file
4643 name is made by appending @file{.dse} to the source file name.
4646 @opindex fdump-tree-phiopt
4647 Dump each function after optimizing PHI nodes into straightline code. The file
4648 name is made by appending @file{.phiopt} to the source file name.
4651 @opindex fdump-tree-forwprop
4652 Dump each function after forward propagating single use variables. The file
4653 name is made by appending @file{.forwprop} to the source file name.
4656 @opindex fdump-tree-copyrename
4657 Dump each function after applying the copy rename optimization. The file
4658 name is made by appending @file{.copyrename} to the source file name.
4661 @opindex fdump-tree-nrv
4662 Dump each function after applying the named return value optimization on
4663 generic trees. The file name is made by appending @file{.nrv} to the source
4667 @opindex fdump-tree-vect
4668 Dump each function after applying vectorization of loops. The file name is
4669 made by appending @file{.vect} to the source file name.
4672 @opindex fdump-tree-vrp
4673 Dump each function after Value Range Propagation (VRP). The file name
4674 is made by appending @file{.vrp} to the source file name.
4677 @opindex fdump-tree-all
4678 Enable all the available tree dumps with the flags provided in this option.
4681 @item -ftree-vectorizer-verbose=@var{n}
4682 @opindex ftree-vectorizer-verbose
4683 This option controls the amount of debugging output the vectorizer prints.
4684 This information is written to standard error, unless
4685 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4686 in which case it is output to the usual dump listing file, @file{.vect}.
4687 For @var{n}=0 no diagnostic information is reported.
4688 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4689 and the total number of loops that got vectorized.
4690 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4691 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4692 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4693 level that @option{-fdump-tree-vect-stats} uses.
4694 Higher verbosity levels mean either more information dumped for each
4695 reported loop, or same amount of information reported for more loops:
4696 If @var{n}=3, alignment related information is added to the reports.
4697 If @var{n}=4, data-references related information (e.g. memory dependences,
4698 memory access-patterns) is added to the reports.
4699 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4700 that did not pass the first analysis phase (i.e. may not be countable, or
4701 may have complicated control-flow).
4702 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4703 For @var{n}=7, all the information the vectorizer generates during its
4704 analysis and transformation is reported. This is the same verbosity level
4705 that @option{-fdump-tree-vect-details} uses.
4707 @item -frandom-seed=@var{string}
4708 @opindex frandom-string
4709 This option provides a seed that GCC uses when it would otherwise use
4710 random numbers. It is used to generate certain symbol names
4711 that have to be different in every compiled file. It is also used to
4712 place unique stamps in coverage data files and the object files that
4713 produce them. You can use the @option{-frandom-seed} option to produce
4714 reproducibly identical object files.
4716 The @var{string} should be different for every file you compile.
4718 @item -fsched-verbose=@var{n}
4719 @opindex fsched-verbose
4720 On targets that use instruction scheduling, this option controls the
4721 amount of debugging output the scheduler prints. This information is
4722 written to standard error, unless @option{-dS} or @option{-dR} is
4723 specified, in which case it is output to the usual dump
4724 listing file, @file{.sched} or @file{.sched2} respectively. However
4725 for @var{n} greater than nine, the output is always printed to standard
4728 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4729 same information as @option{-dRS}. For @var{n} greater than one, it
4730 also output basic block probabilities, detailed ready list information
4731 and unit/insn info. For @var{n} greater than two, it includes RTL
4732 at abort point, control-flow and regions info. And for @var{n} over
4733 four, @option{-fsched-verbose} also includes dependence info.
4737 Store the usual ``temporary'' intermediate files permanently; place them
4738 in the current directory and name them based on the source file. Thus,
4739 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4740 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4741 preprocessed @file{foo.i} output file even though the compiler now
4742 normally uses an integrated preprocessor.
4744 When used in combination with the @option{-x} command line option,
4745 @option{-save-temps} is sensible enough to avoid over writing an
4746 input source file with the same extension as an intermediate file.
4747 The corresponding intermediate file may be obtained by renaming the
4748 source file before using @option{-save-temps}.
4752 Report the CPU time taken by each subprocess in the compilation
4753 sequence. For C source files, this is the compiler proper and assembler
4754 (plus the linker if linking is done). The output looks like this:
4761 The first number on each line is the ``user time'', that is time spent
4762 executing the program itself. The second number is ``system time'',
4763 time spent executing operating system routines on behalf of the program.
4764 Both numbers are in seconds.
4766 @item -fvar-tracking
4767 @opindex fvar-tracking
4768 Run variable tracking pass. It computes where variables are stored at each
4769 position in code. Better debugging information is then generated
4770 (if the debugging information format supports this information).
4772 It is enabled by default when compiling with optimization (@option{-Os},
4773 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4774 the debug info format supports it.
4776 @item -print-file-name=@var{library}
4777 @opindex print-file-name
4778 Print the full absolute name of the library file @var{library} that
4779 would be used when linking---and don't do anything else. With this
4780 option, GCC does not compile or link anything; it just prints the
4783 @item -print-multi-directory
4784 @opindex print-multi-directory
4785 Print the directory name corresponding to the multilib selected by any
4786 other switches present in the command line. This directory is supposed
4787 to exist in @env{GCC_EXEC_PREFIX}.
4789 @item -print-multi-lib
4790 @opindex print-multi-lib
4791 Print the mapping from multilib directory names to compiler switches
4792 that enable them. The directory name is separated from the switches by
4793 @samp{;}, and each switch starts with an @samp{@@} instead of the
4794 @samp{-}, without spaces between multiple switches. This is supposed to
4795 ease shell-processing.
4797 @item -print-prog-name=@var{program}
4798 @opindex print-prog-name
4799 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4801 @item -print-libgcc-file-name
4802 @opindex print-libgcc-file-name
4803 Same as @option{-print-file-name=libgcc.a}.
4805 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4806 but you do want to link with @file{libgcc.a}. You can do
4809 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4812 @item -print-search-dirs
4813 @opindex print-search-dirs
4814 Print the name of the configured installation directory and a list of
4815 program and library directories @command{gcc} will search---and don't do anything else.
4817 This is useful when @command{gcc} prints the error message
4818 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4819 To resolve this you either need to put @file{cpp0} and the other compiler
4820 components where @command{gcc} expects to find them, or you can set the environment
4821 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4822 Don't forget the trailing @samp{/}.
4823 @xref{Environment Variables}.
4825 @item -print-sysroot-headers-suffix
4826 @opindex print-sysroot-headers-suffix
4827 Print the suffix added to the target sysroot when searching for
4828 headers, or give an error if the compiler is not configured with such
4829 a suffix---and don't do anything else.
4832 @opindex dumpmachine
4833 Print the compiler's target machine (for example,
4834 @samp{i686-pc-linux-gnu})---and don't do anything else.
4837 @opindex dumpversion
4838 Print the compiler version (for example, @samp{3.0})---and don't do
4843 Print the compiler's built-in specs---and don't do anything else. (This
4844 is used when GCC itself is being built.) @xref{Spec Files}.
4846 @item -feliminate-unused-debug-types
4847 @opindex feliminate-unused-debug-types
4848 Normally, when producing DWARF2 output, GCC will emit debugging
4849 information for all types declared in a compilation
4850 unit, regardless of whether or not they are actually used
4851 in that compilation unit. Sometimes this is useful, such as
4852 if, in the debugger, you want to cast a value to a type that is
4853 not actually used in your program (but is declared). More often,
4854 however, this results in a significant amount of wasted space.
4855 With this option, GCC will avoid producing debug symbol output
4856 for types that are nowhere used in the source file being compiled.
4859 @node Optimize Options
4860 @section Options That Control Optimization
4861 @cindex optimize options
4862 @cindex options, optimization
4864 These options control various sorts of optimizations.
4866 Without any optimization option, the compiler's goal is to reduce the
4867 cost of compilation and to make debugging produce the expected
4868 results. Statements are independent: if you stop the program with a
4869 breakpoint between statements, you can then assign a new value to any
4870 variable or change the program counter to any other statement in the
4871 function and get exactly the results you would expect from the source
4874 Turning on optimization flags makes the compiler attempt to improve
4875 the performance and/or code size at the expense of compilation time
4876 and possibly the ability to debug the program.
4878 The compiler performs optimization based on the knowledge it has of
4879 the program. Optimization levels @option{-O} and above, in
4880 particular, enable @emph{unit-at-a-time} mode, which allows the
4881 compiler to consider information gained from later functions in
4882 the file when compiling a function. Compiling multiple files at
4883 once to a single output file in @emph{unit-at-a-time} mode allows
4884 the compiler to use information gained from all of the files when
4885 compiling each of them.
4887 Not all optimizations are controlled directly by a flag. Only
4888 optimizations that have a flag are listed.
4895 Optimize. Optimizing compilation takes somewhat more time, and a lot
4896 more memory for a large function.
4898 With @option{-O}, the compiler tries to reduce code size and execution
4899 time, without performing any optimizations that take a great deal of
4902 @option{-O} turns on the following optimization flags:
4903 @gccoptlist{-fdefer-pop @gol
4904 -fdelayed-branch @gol
4905 -fguess-branch-probability @gol
4906 -fcprop-registers @gol
4907 -fif-conversion @gol
4908 -fif-conversion2 @gol
4909 -fsplit-wide-types @gol
4912 -ftree-dominator-opts @gol
4916 -ftree-copyrename @gol
4919 -funit-at-a-time @gol
4922 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4923 where doing so does not interfere with debugging.
4927 Optimize even more. GCC performs nearly all supported optimizations
4928 that do not involve a space-speed tradeoff. The compiler does not
4929 perform loop unrolling or function inlining when you specify @option{-O2}.
4930 As compared to @option{-O}, this option increases both compilation time
4931 and the performance of the generated code.
4933 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4934 also turns on the following optimization flags:
4935 @gccoptlist{-fthread-jumps @gol
4937 -foptimize-sibling-calls @gol
4938 -fcse-follow-jumps -fcse-skip-blocks @gol
4939 -fgcse -fgcse-lm @gol
4940 -fexpensive-optimizations @gol
4941 -frerun-cse-after-loop @gol
4944 -fschedule-insns -fschedule-insns2 @gol
4945 -fsched-interblock -fsched-spec @gol
4947 -fstrict-aliasing -fstrict-overflow @gol
4948 -fdelete-null-pointer-checks @gol
4949 -freorder-blocks -freorder-functions @gol
4950 -falign-functions -falign-jumps @gol
4951 -falign-loops -falign-labels @gol
4955 Please note the warning under @option{-fgcse} about
4956 invoking @option{-O2} on programs that use computed gotos.
4960 Optimize yet more. @option{-O3} turns on all optimizations specified by
4961 @option{-O2} and also turns on the @option{-finline-functions},
4962 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4966 Reduce compilation time and make debugging produce the expected
4967 results. This is the default.
4971 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4972 do not typically increase code size. It also performs further
4973 optimizations designed to reduce code size.
4975 @option{-Os} disables the following optimization flags:
4976 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4977 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4978 -fprefetch-loop-arrays -ftree-vect-loop-version}
4980 If you use multiple @option{-O} options, with or without level numbers,
4981 the last such option is the one that is effective.
4984 Options of the form @option{-f@var{flag}} specify machine-independent
4985 flags. Most flags have both positive and negative forms; the negative
4986 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4987 below, only one of the forms is listed---the one you typically will
4988 use. You can figure out the other form by either removing @samp{no-}
4991 The following options control specific optimizations. They are either
4992 activated by @option{-O} options or are related to ones that are. You
4993 can use the following flags in the rare cases when ``fine-tuning'' of
4994 optimizations to be performed is desired.
4997 @item -fno-default-inline
4998 @opindex fno-default-inline
4999 Do not make member functions inline by default merely because they are
5000 defined inside the class scope (C++ only). Otherwise, when you specify
5001 @w{@option{-O}}, member functions defined inside class scope are compiled
5002 inline by default; i.e., you don't need to add @samp{inline} in front of
5003 the member function name.
5005 @item -fno-defer-pop
5006 @opindex fno-defer-pop
5007 Always pop the arguments to each function call as soon as that function
5008 returns. For machines which must pop arguments after a function call,
5009 the compiler normally lets arguments accumulate on the stack for several
5010 function calls and pops them all at once.
5012 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5015 @opindex fforce-addr
5016 Force memory address constants to be copied into registers before
5017 doing arithmetic on them.
5019 @item -fforward-propagate
5020 @opindex fforward-propagate
5021 Perform a forward propagation pass on RTL. The pass tries to combine two
5022 instructions and checks if the result can be simplified. If loop unrolling
5023 is active, two passes are performed and the second is scheduled after
5026 This option is enabled by default at optimization levels @option{-O2},
5027 @option{-O3}, @option{-Os}.
5029 @item -fomit-frame-pointer
5030 @opindex fomit-frame-pointer
5031 Don't keep the frame pointer in a register for functions that
5032 don't need one. This avoids the instructions to save, set up and
5033 restore frame pointers; it also makes an extra register available
5034 in many functions. @strong{It also makes debugging impossible on
5037 On some machines, such as the VAX, this flag has no effect, because
5038 the standard calling sequence automatically handles the frame pointer
5039 and nothing is saved by pretending it doesn't exist. The
5040 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5041 whether a target machine supports this flag. @xref{Registers,,Register
5042 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5044 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5046 @item -foptimize-sibling-calls
5047 @opindex foptimize-sibling-calls
5048 Optimize sibling and tail recursive calls.
5050 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5054 Don't pay attention to the @code{inline} keyword. Normally this option
5055 is used to keep the compiler from expanding any functions inline.
5056 Note that if you are not optimizing, no functions can be expanded inline.
5058 @item -finline-functions
5059 @opindex finline-functions
5060 Integrate all simple functions into their callers. The compiler
5061 heuristically decides which functions are simple enough to be worth
5062 integrating in this way.
5064 If all calls to a given function are integrated, and the function is
5065 declared @code{static}, then the function is normally not output as
5066 assembler code in its own right.
5068 Enabled at level @option{-O3}.
5070 @item -finline-functions-called-once
5071 @opindex finline-functions-called-once
5072 Consider all @code{static} functions called once for inlining into their
5073 caller even if they are not marked @code{inline}. If a call to a given
5074 function is integrated, then the function is not output as assembler code
5077 Enabled if @option{-funit-at-a-time} is enabled.
5079 @item -fearly-inlining
5080 @opindex fearly-inlining
5081 Inline functions marked by @code{always_inline} and functions whose body seems
5082 smaller than the function call overhead early before doing
5083 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5084 makes profiling significantly cheaper and usually inlining faster on programs
5085 having large chains of nested wrapper functions.
5089 @item -finline-limit=@var{n}
5090 @opindex finline-limit
5091 By default, GCC limits the size of functions that can be inlined. This flag
5092 allows the control of this limit for functions that are explicitly marked as
5093 inline (i.e., marked with the inline keyword or defined within the class
5094 definition in c++). @var{n} is the size of functions that can be inlined in
5095 number of pseudo instructions (not counting parameter handling). The default
5096 value of @var{n} is 600.
5097 Increasing this value can result in more inlined code at
5098 the cost of compilation time and memory consumption. Decreasing usually makes
5099 the compilation faster and less code will be inlined (which presumably
5100 means slower programs). This option is particularly useful for programs that
5101 use inlining heavily such as those based on recursive templates with C++.
5103 Inlining is actually controlled by a number of parameters, which may be
5104 specified individually by using @option{--param @var{name}=@var{value}}.
5105 The @option{-finline-limit=@var{n}} option sets some of these parameters
5109 @item max-inline-insns-single
5110 is set to @var{n}/2.
5111 @item max-inline-insns-auto
5112 is set to @var{n}/2.
5113 @item min-inline-insns
5114 is set to 130 or @var{n}/4, whichever is smaller.
5115 @item max-inline-insns-rtl
5119 See below for a documentation of the individual
5120 parameters controlling inlining.
5122 @emph{Note:} pseudo instruction represents, in this particular context, an
5123 abstract measurement of function's size. In no way does it represent a count
5124 of assembly instructions and as such its exact meaning might change from one
5125 release to an another.
5127 @item -fkeep-inline-functions
5128 @opindex fkeep-inline-functions
5129 In C, emit @code{static} functions that are declared @code{inline}
5130 into the object file, even if the function has been inlined into all
5131 of its callers. This switch does not affect functions using the
5132 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5133 inline functions into the object file.
5135 @item -fkeep-static-consts
5136 @opindex fkeep-static-consts
5137 Emit variables declared @code{static const} when optimization isn't turned
5138 on, even if the variables aren't referenced.
5140 GCC enables this option by default. If you want to force the compiler to
5141 check if the variable was referenced, regardless of whether or not
5142 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5144 @item -fmerge-constants
5145 Attempt to merge identical constants (string constants and floating point
5146 constants) across compilation units.
5148 This option is the default for optimized compilation if the assembler and
5149 linker support it. Use @option{-fno-merge-constants} to inhibit this
5152 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5154 @item -fmerge-all-constants
5155 Attempt to merge identical constants and identical variables.
5157 This option implies @option{-fmerge-constants}. In addition to
5158 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5159 arrays or initialized constant variables with integral or floating point
5160 types. Languages like C or C++ require each non-automatic variable to
5161 have distinct location, so using this option will result in non-conforming
5164 @item -fmodulo-sched
5165 @opindex fmodulo-sched
5166 Perform swing modulo scheduling immediately before the first scheduling
5167 pass. This pass looks at innermost loops and reorders their
5168 instructions by overlapping different iterations.
5170 @item -fno-branch-count-reg
5171 @opindex fno-branch-count-reg
5172 Do not use ``decrement and branch'' instructions on a count register,
5173 but instead generate a sequence of instructions that decrement a
5174 register, compare it against zero, then branch based upon the result.
5175 This option is only meaningful on architectures that support such
5176 instructions, which include x86, PowerPC, IA-64 and S/390.
5178 The default is @option{-fbranch-count-reg}.
5180 @item -fno-function-cse
5181 @opindex fno-function-cse
5182 Do not put function addresses in registers; make each instruction that
5183 calls a constant function contain the function's address explicitly.
5185 This option results in less efficient code, but some strange hacks
5186 that alter the assembler output may be confused by the optimizations
5187 performed when this option is not used.
5189 The default is @option{-ffunction-cse}
5191 @item -fno-zero-initialized-in-bss
5192 @opindex fno-zero-initialized-in-bss
5193 If the target supports a BSS section, GCC by default puts variables that
5194 are initialized to zero into BSS@. This can save space in the resulting
5197 This option turns off this behavior because some programs explicitly
5198 rely on variables going to the data section. E.g., so that the
5199 resulting executable can find the beginning of that section and/or make
5200 assumptions based on that.
5202 The default is @option{-fzero-initialized-in-bss}.
5204 @item -fbounds-check
5205 @opindex fbounds-check
5206 For front-ends that support it, generate additional code to check that
5207 indices used to access arrays are within the declared range. This is
5208 currently only supported by the Java and Fortran front-ends, where
5209 this option defaults to true and false respectively.
5211 @item -fmudflap -fmudflapth -fmudflapir
5215 @cindex bounds checking
5217 For front-ends that support it (C and C++), instrument all risky
5218 pointer/array dereferencing operations, some standard library
5219 string/heap functions, and some other associated constructs with
5220 range/validity tests. Modules so instrumented should be immune to
5221 buffer overflows, invalid heap use, and some other classes of C/C++
5222 programming errors. The instrumentation relies on a separate runtime
5223 library (@file{libmudflap}), which will be linked into a program if
5224 @option{-fmudflap} is given at link time. Run-time behavior of the
5225 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5226 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5229 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5230 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5231 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5232 instrumentation should ignore pointer reads. This produces less
5233 instrumentation (and therefore faster execution) and still provides
5234 some protection against outright memory corrupting writes, but allows
5235 erroneously read data to propagate within a program.
5237 @item -fthread-jumps
5238 @opindex fthread-jumps
5239 Perform optimizations where we check to see if a jump branches to a
5240 location where another comparison subsumed by the first is found. If
5241 so, the first branch is redirected to either the destination of the
5242 second branch or a point immediately following it, depending on whether
5243 the condition is known to be true or false.
5245 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5247 @item -fsplit-wide-types
5248 @opindex fsplit-wide-types
5249 When using a type that occupies multiple registers, such as @code{long
5250 long} on a 32-bit system, split the registers apart and allocate them
5251 independently. This normally generates better code for those types,
5252 but may make debugging more difficult.
5254 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5257 @item -fcse-follow-jumps
5258 @opindex fcse-follow-jumps
5259 In common subexpression elimination, scan through jump instructions
5260 when the target of the jump is not reached by any other path. For
5261 example, when CSE encounters an @code{if} statement with an
5262 @code{else} clause, CSE will follow the jump when the condition
5265 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5267 @item -fcse-skip-blocks
5268 @opindex fcse-skip-blocks
5269 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5270 follow jumps which conditionally skip over blocks. When CSE
5271 encounters a simple @code{if} statement with no else clause,
5272 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5273 body of the @code{if}.
5275 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5277 @item -frerun-cse-after-loop
5278 @opindex frerun-cse-after-loop
5279 Re-run common subexpression elimination after loop optimizations has been
5282 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5286 Perform a global common subexpression elimination pass.
5287 This pass also performs global constant and copy propagation.
5289 @emph{Note:} When compiling a program using computed gotos, a GCC
5290 extension, you may get better runtime performance if you disable
5291 the global common subexpression elimination pass by adding
5292 @option{-fno-gcse} to the command line.
5294 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5298 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5299 attempt to move loads which are only killed by stores into themselves. This
5300 allows a loop containing a load/store sequence to be changed to a load outside
5301 the loop, and a copy/store within the loop.
5303 Enabled by default when gcse is enabled.
5307 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5308 global common subexpression elimination. This pass will attempt to move
5309 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5310 loops containing a load/store sequence can be changed to a load before
5311 the loop and a store after the loop.
5313 Not enabled at any optimization level.
5317 When @option{-fgcse-las} is enabled, the global common subexpression
5318 elimination pass eliminates redundant loads that come after stores to the
5319 same memory location (both partial and full redundancies).
5321 Not enabled at any optimization level.
5323 @item -fgcse-after-reload
5324 @opindex fgcse-after-reload
5325 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5326 pass is performed after reload. The purpose of this pass is to cleanup
5329 @item -funsafe-loop-optimizations
5330 @opindex funsafe-loop-optimizations
5331 If given, the loop optimizer will assume that loop indices do not
5332 overflow, and that the loops with nontrivial exit condition are not
5333 infinite. This enables a wider range of loop optimizations even if
5334 the loop optimizer itself cannot prove that these assumptions are valid.
5335 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5336 if it finds this kind of loop.
5338 @item -fcrossjumping
5339 @opindex crossjumping
5340 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5341 resulting code may or may not perform better than without cross-jumping.
5343 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5345 @item -fif-conversion
5346 @opindex if-conversion
5347 Attempt to transform conditional jumps into branch-less equivalents. This
5348 include use of conditional moves, min, max, set flags and abs instructions, and
5349 some tricks doable by standard arithmetics. The use of conditional execution
5350 on chips where it is available is controlled by @code{if-conversion2}.
5352 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5354 @item -fif-conversion2
5355 @opindex if-conversion2
5356 Use conditional execution (where available) to transform conditional jumps into
5357 branch-less equivalents.
5359 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5361 @item -fdelete-null-pointer-checks
5362 @opindex fdelete-null-pointer-checks
5363 Use global dataflow analysis to identify and eliminate useless checks
5364 for null pointers. The compiler assumes that dereferencing a null
5365 pointer would have halted the program. If a pointer is checked after
5366 it has already been dereferenced, it cannot be null.
5368 In some environments, this assumption is not true, and programs can
5369 safely dereference null pointers. Use
5370 @option{-fno-delete-null-pointer-checks} to disable this optimization
5371 for programs which depend on that behavior.
5373 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5375 @item -fexpensive-optimizations
5376 @opindex fexpensive-optimizations
5377 Perform a number of minor optimizations that are relatively expensive.
5379 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5381 @item -foptimize-register-move
5383 @opindex foptimize-register-move
5385 Attempt to reassign register numbers in move instructions and as
5386 operands of other simple instructions in order to maximize the amount of
5387 register tying. This is especially helpful on machines with two-operand
5390 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5393 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5395 @item -fdelayed-branch
5396 @opindex fdelayed-branch
5397 If supported for the target machine, attempt to reorder instructions
5398 to exploit instruction slots available after delayed branch
5401 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5403 @item -fschedule-insns
5404 @opindex fschedule-insns
5405 If supported for the target machine, attempt to reorder instructions to
5406 eliminate execution stalls due to required data being unavailable. This
5407 helps machines that have slow floating point or memory load instructions
5408 by allowing other instructions to be issued until the result of the load
5409 or floating point instruction is required.
5411 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5413 @item -fschedule-insns2
5414 @opindex fschedule-insns2
5415 Similar to @option{-fschedule-insns}, but requests an additional pass of
5416 instruction scheduling after register allocation has been done. This is
5417 especially useful on machines with a relatively small number of
5418 registers and where memory load instructions take more than one cycle.
5420 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5422 @item -fno-sched-interblock
5423 @opindex fno-sched-interblock
5424 Don't schedule instructions across basic blocks. This is normally
5425 enabled by default when scheduling before register allocation, i.e.@:
5426 with @option{-fschedule-insns} or at @option{-O2} or higher.
5428 @item -fno-sched-spec
5429 @opindex fno-sched-spec
5430 Don't allow speculative motion of non-load instructions. This is normally
5431 enabled by default when scheduling before register allocation, i.e.@:
5432 with @option{-fschedule-insns} or at @option{-O2} or higher.
5434 @item -fsched-spec-load
5435 @opindex fsched-spec-load
5436 Allow speculative motion of some load instructions. This only makes
5437 sense when scheduling before register allocation, i.e.@: with
5438 @option{-fschedule-insns} or at @option{-O2} or higher.
5440 @item -fsched-spec-load-dangerous
5441 @opindex fsched-spec-load-dangerous
5442 Allow speculative motion of more load instructions. This only makes
5443 sense when scheduling before register allocation, i.e.@: with
5444 @option{-fschedule-insns} or at @option{-O2} or higher.
5446 @item -fsched-stalled-insns=@var{n}
5447 @opindex fsched-stalled-insns
5448 Define how many insns (if any) can be moved prematurely from the queue
5449 of stalled insns into the ready list, during the second scheduling pass.
5451 @item -fsched-stalled-insns-dep=@var{n}
5452 @opindex fsched-stalled-insns-dep
5453 Define how many insn groups (cycles) will be examined for a dependency
5454 on a stalled insn that is candidate for premature removal from the queue
5455 of stalled insns. Has an effect only during the second scheduling pass,
5456 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5458 @item -fsched2-use-superblocks
5459 @opindex fsched2-use-superblocks
5460 When scheduling after register allocation, do use superblock scheduling
5461 algorithm. Superblock scheduling allows motion across basic block boundaries
5462 resulting on faster schedules. This option is experimental, as not all machine
5463 descriptions used by GCC model the CPU closely enough to avoid unreliable
5464 results from the algorithm.
5466 This only makes sense when scheduling after register allocation, i.e.@: with
5467 @option{-fschedule-insns2} or at @option{-O2} or higher.
5469 @item -fsched2-use-traces
5470 @opindex fsched2-use-traces
5471 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5472 allocation and additionally perform code duplication in order to increase the
5473 size of superblocks using tracer pass. See @option{-ftracer} for details on
5476 This mode should produce faster but significantly longer programs. Also
5477 without @option{-fbranch-probabilities} the traces constructed may not
5478 match the reality and hurt the performance. This only makes
5479 sense when scheduling after register allocation, i.e.@: with
5480 @option{-fschedule-insns2} or at @option{-O2} or higher.
5484 Eliminates redundant extension instructions and move the non redundant
5485 ones to optimal placement using LCM.
5487 @item -freschedule-modulo-scheduled-loops
5488 @opindex fscheduling-in-modulo-scheduled-loops
5489 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5490 we may want to prevent the later scheduling passes from changing its schedule, we use this
5491 option to control that.
5493 @item -fcaller-saves
5494 @opindex fcaller-saves
5495 Enable values to be allocated in registers that will be clobbered by
5496 function calls, by emitting extra instructions to save and restore the
5497 registers around such calls. Such allocation is done only when it
5498 seems to result in better code than would otherwise be produced.
5500 This option is always enabled by default on certain machines, usually
5501 those which have no call-preserved registers to use instead.
5503 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5506 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5507 enabled by default at @option{-O2} and @option{-O3}.
5510 Perform Full Redundancy Elimination (FRE) on trees. The difference
5511 between FRE and PRE is that FRE only considers expressions
5512 that are computed on all paths leading to the redundant computation.
5513 This analysis is faster than PRE, though it exposes fewer redundancies.
5514 This flag is enabled by default at @option{-O} and higher.
5516 @item -ftree-copy-prop
5517 Perform copy propagation on trees. This pass eliminates unnecessary
5518 copy operations. This flag is enabled by default at @option{-O} and
5521 @item -ftree-store-copy-prop
5522 Perform copy propagation of memory loads and stores. This pass
5523 eliminates unnecessary copy operations in memory references
5524 (structures, global variables, arrays, etc). This flag is enabled by
5525 default at @option{-O2} and higher.
5528 Perform structural alias analysis on trees. This flag
5529 is enabled by default at @option{-O} and higher.
5532 Perform interprocedural pointer analysis.
5535 Perform forward store motion on trees. This flag is
5536 enabled by default at @option{-O} and higher.
5539 Perform sparse conditional constant propagation (CCP) on trees. This
5540 pass only operates on local scalar variables and is enabled by default
5541 at @option{-O} and higher.
5543 @item -ftree-store-ccp
5544 Perform sparse conditional constant propagation (CCP) on trees. This
5545 pass operates on both local scalar variables and memory stores and
5546 loads (global variables, structures, arrays, etc). This flag is
5547 enabled by default at @option{-O2} and higher.
5550 Perform dead code elimination (DCE) on trees. This flag is enabled by
5551 default at @option{-O} and higher.
5553 @item -ftree-dominator-opts
5554 Perform a variety of simple scalar cleanups (constant/copy
5555 propagation, redundancy elimination, range propagation and expression
5556 simplification) based on a dominator tree traversal. This also
5557 performs jump threading (to reduce jumps to jumps). This flag is
5558 enabled by default at @option{-O} and higher.
5561 Perform loop header copying on trees. This is beneficial since it increases
5562 effectiveness of code motion optimizations. It also saves one jump. This flag
5563 is enabled by default at @option{-O} and higher. It is not enabled
5564 for @option{-Os}, since it usually increases code size.
5566 @item -ftree-loop-optimize
5567 Perform loop optimizations on trees. This flag is enabled by default
5568 at @option{-O} and higher.
5570 @item -ftree-loop-linear
5571 Perform linear loop transformations on tree. This flag can improve cache
5572 performance and allow further loop optimizations to take place.
5574 @item -fcheck-data-deps
5575 Compare the results of several data dependence analyzers. This option
5576 is used for debugging the data dependence analyzers.
5578 @item -ftree-loop-im
5579 Perform loop invariant motion on trees. This pass moves only invariants that
5580 would be hard to handle at RTL level (function calls, operations that expand to
5581 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5582 operands of conditions that are invariant out of the loop, so that we can use
5583 just trivial invariantness analysis in loop unswitching. The pass also includes
5586 @item -ftree-loop-ivcanon
5587 Create a canonical counter for number of iterations in the loop for that
5588 determining number of iterations requires complicated analysis. Later
5589 optimizations then may determine the number easily. Useful especially
5590 in connection with unrolling.
5593 Perform induction variable optimizations (strength reduction, induction
5594 variable merging and induction variable elimination) on trees.
5597 Perform scalar replacement of aggregates. This pass replaces structure
5598 references with scalars to prevent committing structures to memory too
5599 early. This flag is enabled by default at @option{-O} and higher.
5601 @item -ftree-copyrename
5602 Perform copy renaming on trees. This pass attempts to rename compiler
5603 temporaries to other variables at copy locations, usually resulting in
5604 variable names which more closely resemble the original variables. This flag
5605 is enabled by default at @option{-O} and higher.
5608 Perform temporary expression replacement during the SSA->normal phase. Single
5609 use/single def temporaries are replaced at their use location with their
5610 defining expression. This results in non-GIMPLE code, but gives the expanders
5611 much more complex trees to work on resulting in better RTL generation. This is
5612 enabled by default at @option{-O} and higher.
5614 @item -ftree-vectorize
5615 Perform loop vectorization on trees.
5617 @item -ftree-vect-loop-version
5618 @opindex ftree-vect-loop-version
5619 Perform loop versioning when doing loop vectorization on trees. When a loop
5620 appears to be vectorizable except that data alignment or data dependence cannot
5621 be determined at compile time then vectorized and non-vectorized versions of
5622 the loop are generated along with runtime checks for alignment or dependence
5623 to control which version is executed. This option is enabled by default
5624 except at level @option{-Os} where it is disabled.
5627 Perform Value Range Propagation on trees. This is similar to the
5628 constant propagation pass, but instead of values, ranges of values are
5629 propagated. This allows the optimizers to remove unnecessary range
5630 checks like array bound checks and null pointer checks. This is
5631 enabled by default at @option{-O2} and higher. Null pointer check
5632 elimination is only done if @option{-fdelete-null-pointer-checks} is
5637 Perform tail duplication to enlarge superblock size. This transformation
5638 simplifies the control flow of the function allowing other optimizations to do
5641 @item -funroll-loops
5642 @opindex funroll-loops
5643 Unroll loops whose number of iterations can be determined at compile
5644 time or upon entry to the loop. @option{-funroll-loops} implies
5645 @option{-frerun-cse-after-loop}. This option makes code larger,
5646 and may or may not make it run faster.
5648 @item -funroll-all-loops
5649 @opindex funroll-all-loops
5650 Unroll all loops, even if their number of iterations is uncertain when
5651 the loop is entered. This usually makes programs run more slowly.
5652 @option{-funroll-all-loops} implies the same options as
5653 @option{-funroll-loops},
5655 @item -fsplit-ivs-in-unroller
5656 @opindex fsplit-ivs-in-unroller
5657 Enables expressing of values of induction variables in later iterations
5658 of the unrolled loop using the value in the first iteration. This breaks
5659 long dependency chains, thus improving efficiency of the scheduling passes.
5661 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5662 same effect. However in cases the loop body is more complicated than
5663 a single basic block, this is not reliable. It also does not work at all
5664 on some of the architectures due to restrictions in the CSE pass.
5666 This optimization is enabled by default.
5668 @item -fvariable-expansion-in-unroller
5669 @opindex fvariable-expansion-in-unroller
5670 With this option, the compiler will create multiple copies of some
5671 local variables when unrolling a loop which can result in superior code.
5673 @item -fprefetch-loop-arrays
5674 @opindex fprefetch-loop-arrays
5675 If supported by the target machine, generate instructions to prefetch
5676 memory to improve the performance of loops that access large arrays.
5678 This option may generate better or worse code; results are highly
5679 dependent on the structure of loops within the source code.
5681 Disabled at level @option{-Os}.
5684 @itemx -fno-peephole2
5685 @opindex fno-peephole
5686 @opindex fno-peephole2
5687 Disable any machine-specific peephole optimizations. The difference
5688 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5689 are implemented in the compiler; some targets use one, some use the
5690 other, a few use both.
5692 @option{-fpeephole} is enabled by default.
5693 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5695 @item -fno-guess-branch-probability
5696 @opindex fno-guess-branch-probability
5697 Do not guess branch probabilities using heuristics.
5699 GCC will use heuristics to guess branch probabilities if they are
5700 not provided by profiling feedback (@option{-fprofile-arcs}). These
5701 heuristics are based on the control flow graph. If some branch probabilities
5702 are specified by @samp{__builtin_expect}, then the heuristics will be
5703 used to guess branch probabilities for the rest of the control flow graph,
5704 taking the @samp{__builtin_expect} info into account. The interactions
5705 between the heuristics and @samp{__builtin_expect} can be complex, and in
5706 some cases, it may be useful to disable the heuristics so that the effects
5707 of @samp{__builtin_expect} are easier to understand.
5709 The default is @option{-fguess-branch-probability} at levels
5710 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5712 @item -freorder-blocks
5713 @opindex freorder-blocks
5714 Reorder basic blocks in the compiled function in order to reduce number of
5715 taken branches and improve code locality.
5717 Enabled at levels @option{-O2}, @option{-O3}.
5719 @item -freorder-blocks-and-partition
5720 @opindex freorder-blocks-and-partition
5721 In addition to reordering basic blocks in the compiled function, in order
5722 to reduce number of taken branches, partitions hot and cold basic blocks
5723 into separate sections of the assembly and .o files, to improve
5724 paging and cache locality performance.
5726 This optimization is automatically turned off in the presence of
5727 exception handling, for linkonce sections, for functions with a user-defined
5728 section attribute and on any architecture that does not support named
5731 @item -freorder-functions
5732 @opindex freorder-functions
5733 Reorder functions in the object file in order to
5734 improve code locality. This is implemented by using special
5735 subsections @code{.text.hot} for most frequently executed functions and
5736 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5737 the linker so object file format must support named sections and linker must
5738 place them in a reasonable way.
5740 Also profile feedback must be available in to make this option effective. See
5741 @option{-fprofile-arcs} for details.
5743 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5745 @item -fstrict-aliasing
5746 @opindex fstrict-aliasing
5747 Allows the compiler to assume the strictest aliasing rules applicable to
5748 the language being compiled. For C (and C++), this activates
5749 optimizations based on the type of expressions. In particular, an
5750 object of one type is assumed never to reside at the same address as an
5751 object of a different type, unless the types are almost the same. For
5752 example, an @code{unsigned int} can alias an @code{int}, but not a
5753 @code{void*} or a @code{double}. A character type may alias any other
5756 Pay special attention to code like this:
5769 The practice of reading from a different union member than the one most
5770 recently written to (called ``type-punning'') is common. Even with
5771 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5772 is accessed through the union type. So, the code above will work as
5773 expected. However, this code might not:
5784 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5786 @item -fstrict-overflow
5787 @opindex fstrict-overflow
5788 Allow the compiler to assume strict signed overflow rules, depending
5789 on the language being compiled. For C (and C++) this means that
5790 overflow when doing arithmetic with signed numbers is undefined, which
5791 means that the compiler may assume that it will not happen. This
5792 permits various optimizations. For example, the compiler will assume
5793 that an expression like @code{i + 10 > i} will always be true for
5794 signed @code{i}. This assumption is only valid if signed overflow is
5795 undefined, as the expression is false if @code{i + 10} overflows when
5796 using twos complement arithmetic. When this option is in effect any
5797 attempt to determine whether an operation on signed numbers will
5798 overflow must be written carefully to not actually involve overflow.
5800 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5801 that signed overflow is fully defined: it wraps. When
5802 @option{-fwrapv} is used, there is no difference between
5803 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5804 @option{-fwrapv} certain types of overflow are permitted. For
5805 example, if the compiler gets an overflow when doing arithmetic on
5806 constants, the overflowed value can still be used with
5807 @option{-fwrapv}, but not otherwise.
5809 The @option{-fstrict-overflow} option is enabled at levels
5810 @option{-O2}, @option{-O3}, @option{-Os}.
5812 @item -falign-functions
5813 @itemx -falign-functions=@var{n}
5814 @opindex falign-functions
5815 Align the start of functions to the next power-of-two greater than
5816 @var{n}, skipping up to @var{n} bytes. For instance,
5817 @option{-falign-functions=32} aligns functions to the next 32-byte
5818 boundary, but @option{-falign-functions=24} would align to the next
5819 32-byte boundary only if this can be done by skipping 23 bytes or less.
5821 @option{-fno-align-functions} and @option{-falign-functions=1} are
5822 equivalent and mean that functions will not be aligned.
5824 Some assemblers only support this flag when @var{n} is a power of two;
5825 in that case, it is rounded up.
5827 If @var{n} is not specified or is zero, use a machine-dependent default.
5829 Enabled at levels @option{-O2}, @option{-O3}.
5831 @item -falign-labels
5832 @itemx -falign-labels=@var{n}
5833 @opindex falign-labels
5834 Align all branch targets to a power-of-two boundary, skipping up to
5835 @var{n} bytes like @option{-falign-functions}. This option can easily
5836 make code slower, because it must insert dummy operations for when the
5837 branch target is reached in the usual flow of the code.
5839 @option{-fno-align-labels} and @option{-falign-labels=1} are
5840 equivalent and mean that labels will not be aligned.
5842 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5843 are greater than this value, then their values are used instead.
5845 If @var{n} is not specified or is zero, use a machine-dependent default
5846 which is very likely to be @samp{1}, meaning no alignment.
5848 Enabled at levels @option{-O2}, @option{-O3}.
5851 @itemx -falign-loops=@var{n}
5852 @opindex falign-loops
5853 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5854 like @option{-falign-functions}. The hope is that the loop will be
5855 executed many times, which will make up for any execution of the dummy
5858 @option{-fno-align-loops} and @option{-falign-loops=1} are
5859 equivalent and mean that loops will not be aligned.
5861 If @var{n} is not specified or is zero, use a machine-dependent default.
5863 Enabled at levels @option{-O2}, @option{-O3}.
5866 @itemx -falign-jumps=@var{n}
5867 @opindex falign-jumps
5868 Align branch targets to a power-of-two boundary, for branch targets
5869 where the targets can only be reached by jumping, skipping up to @var{n}
5870 bytes like @option{-falign-functions}. In this case, no dummy operations
5873 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5874 equivalent and mean that loops will not be aligned.
5876 If @var{n} is not specified or is zero, use a machine-dependent default.
5878 Enabled at levels @option{-O2}, @option{-O3}.
5880 @item -funit-at-a-time
5881 @opindex funit-at-a-time
5882 Parse the whole compilation unit before starting to produce code.
5883 This allows some extra optimizations to take place but consumes
5884 more memory (in general). There are some compatibility issues
5885 with @emph{unit-at-a-time} mode:
5888 enabling @emph{unit-at-a-time} mode may change the order
5889 in which functions, variables, and top-level @code{asm} statements
5890 are emitted, and will likely break code relying on some particular
5891 ordering. The majority of such top-level @code{asm} statements,
5892 though, can be replaced by @code{section} attributes. The
5893 @option{fno-toplevel-reorder} option may be used to keep the ordering
5894 used in the input file, at the cost of some optimizations.
5897 @emph{unit-at-a-time} mode removes unreferenced static variables
5898 and functions. This may result in undefined references
5899 when an @code{asm} statement refers directly to variables or functions
5900 that are otherwise unused. In that case either the variable/function
5901 shall be listed as an operand of the @code{asm} statement operand or,
5902 in the case of top-level @code{asm} statements the attribute @code{used}
5903 shall be used on the declaration.
5906 Static functions now can use non-standard passing conventions that
5907 may break @code{asm} statements calling functions directly. Again,
5908 attribute @code{used} will prevent this behavior.
5911 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5912 but this scheme may not be supported by future releases of GCC@.
5914 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5916 @item -fno-toplevel-reorder
5917 Do not reorder top-level functions, variables, and @code{asm}
5918 statements. Output them in the same order that they appear in the
5919 input file. When this option is used, unreferenced static variables
5920 will not be removed. This option is intended to support existing code
5921 which relies on a particular ordering. For new code, it is better to
5926 Constructs webs as commonly used for register allocation purposes and assign
5927 each web individual pseudo register. This allows the register allocation pass
5928 to operate on pseudos directly, but also strengthens several other optimization
5929 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5930 however, make debugging impossible, since variables will no longer stay in a
5933 Enabled by default with @option{-funroll-loops}.
5935 @item -fwhole-program
5936 @opindex fwhole-program
5937 Assume that the current compilation unit represents whole program being
5938 compiled. All public functions and variables with the exception of @code{main}
5939 and those merged by attribute @code{externally_visible} become static functions
5940 and in a affect gets more aggressively optimized by interprocedural optimizers.
5941 While this option is equivalent to proper use of @code{static} keyword for
5942 programs consisting of single file, in combination with option
5943 @option{--combine} this flag can be used to compile most of smaller scale C
5944 programs since the functions and variables become local for the whole combined
5945 compilation unit, not for the single source file itself.
5948 @item -fno-cprop-registers
5949 @opindex fno-cprop-registers
5950 After register allocation and post-register allocation instruction splitting,
5951 we perform a copy-propagation pass to try to reduce scheduling dependencies
5952 and occasionally eliminate the copy.
5954 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5956 @item -fprofile-generate
5957 @opindex fprofile-generate
5959 Enable options usually used for instrumenting application to produce
5960 profile useful for later recompilation with profile feedback based
5961 optimization. You must use @option{-fprofile-generate} both when
5962 compiling and when linking your program.
5964 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5967 @opindex fprofile-use
5968 Enable profile feedback directed optimizations, and optimizations
5969 generally profitable only with profile feedback available.
5971 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5972 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5974 By default, GCC emits an error message if the feedback profiles do not
5975 match the source code. This error can be turned into a warning by using
5976 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
5980 The following options control compiler behavior regarding floating
5981 point arithmetic. These options trade off between speed and
5982 correctness. All must be specifically enabled.
5986 @opindex ffloat-store
5987 Do not store floating point variables in registers, and inhibit other
5988 options that might change whether a floating point value is taken from a
5991 @cindex floating point precision
5992 This option prevents undesirable excess precision on machines such as
5993 the 68000 where the floating registers (of the 68881) keep more
5994 precision than a @code{double} is supposed to have. Similarly for the
5995 x86 architecture. For most programs, the excess precision does only
5996 good, but a few programs rely on the precise definition of IEEE floating
5997 point. Use @option{-ffloat-store} for such programs, after modifying
5998 them to store all pertinent intermediate computations into variables.
6002 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
6003 @option{-fno-trapping-math}, @option{-ffinite-math-only},
6004 @option{-fno-rounding-math}, @option{-fno-signaling-nans},
6005 @option{-fno-signed-zeros} and @option{fcx-limited-range}.
6007 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6009 This option is not turned on by any @option{-O} option since
6010 it can result in incorrect output for programs which depend on
6011 an exact implementation of IEEE or ISO rules/specifications for
6012 math functions. It may, however, yield faster code for programs
6013 that do not require the guarantees of these specifications.
6015 @item -fno-math-errno
6016 @opindex fno-math-errno
6017 Do not set ERRNO after calling math functions that are executed
6018 with a single instruction, e.g., sqrt. A program that relies on
6019 IEEE exceptions for math error handling may want to use this flag
6020 for speed while maintaining IEEE arithmetic compatibility.
6022 This option is not turned on by any @option{-O} option since
6023 it can result in incorrect output for programs which depend on
6024 an exact implementation of IEEE or ISO rules/specifications for
6025 math functions. It may, however, yield faster code for programs
6026 that do not require the guarantees of these specifications.
6028 The default is @option{-fmath-errno}.
6030 On Darwin systems, the math library never sets @code{errno}. There is
6031 therefore no reason for the compiler to consider the possibility that
6032 it might, and @option{-fno-math-errno} is the default.
6034 @item -funsafe-math-optimizations
6035 @opindex funsafe-math-optimizations
6036 Allow optimizations for floating-point arithmetic that (a) assume
6037 that arguments and results are valid and (b) may violate IEEE or
6038 ANSI standards. When used at link-time, it may include libraries
6039 or startup files that change the default FPU control word or other
6040 similar optimizations.
6042 This option is not turned on by any @option{-O} option since
6043 it can result in incorrect output for programs which depend on
6044 an exact implementation of IEEE or ISO rules/specifications for
6045 math functions. It may, however, yield faster code for programs
6046 that do not require the guarantees of these specifications.
6048 The default is @option{-fno-unsafe-math-optimizations}.
6050 @item -ffinite-math-only
6051 @opindex ffinite-math-only
6052 Allow optimizations for floating-point arithmetic that assume
6053 that arguments and results are not NaNs or +-Infs.
6055 This option is not turned on by any @option{-O} option since
6056 it can result in incorrect output for programs which depend on
6057 an exact implementation of IEEE or ISO rules/specifications for
6058 math functions. It may, however, yield faster code for programs
6059 that do not require the guarantees of these specifications.
6061 The default is @option{-fno-finite-math-only}.
6063 @item -fno-signed-zeros
6064 @opindex fno-signed-zeros
6065 Allow optimizations for floating point arithmetic that ignore the
6066 signedness of zero. IEEE arithmetic specifies the behavior of
6067 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6068 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6069 This option implies that the sign of a zero result isn't significant.
6071 The default is @option{-fsigned-zeros}.
6073 @item -fno-trapping-math
6074 @opindex fno-trapping-math
6075 Compile code assuming that floating-point operations cannot generate
6076 user-visible traps. These traps include division by zero, overflow,
6077 underflow, inexact result and invalid operation. This option implies
6078 @option{-fno-signaling-nans}. Setting this option may allow faster
6079 code if one relies on ``non-stop'' IEEE arithmetic, for example.
6081 This option should never be turned on by any @option{-O} option since
6082 it can result in incorrect output for programs which depend on
6083 an exact implementation of IEEE or ISO rules/specifications for
6086 The default is @option{-ftrapping-math}.
6088 @item -frounding-math
6089 @opindex frounding-math
6090 Disable transformations and optimizations that assume default floating
6091 point rounding behavior. This is round-to-zero for all floating point
6092 to integer conversions, and round-to-nearest for all other arithmetic
6093 truncations. This option should be specified for programs that change
6094 the FP rounding mode dynamically, or that may be executed with a
6095 non-default rounding mode. This option disables constant folding of
6096 floating point expressions at compile-time (which may be affected by
6097 rounding mode) and arithmetic transformations that are unsafe in the
6098 presence of sign-dependent rounding modes.
6100 The default is @option{-fno-rounding-math}.
6102 This option is experimental and does not currently guarantee to
6103 disable all GCC optimizations that are affected by rounding mode.
6104 Future versions of GCC may provide finer control of this setting
6105 using C99's @code{FENV_ACCESS} pragma. This command line option
6106 will be used to specify the default state for @code{FENV_ACCESS}.
6108 @item -frtl-abstract-sequences
6109 @opindex frtl-abstract-sequences
6110 It is a size optimization method. This option is to find identical
6111 sequences of code, which can be turned into pseudo-procedures and
6112 then replace all occurrences with calls to the newly created
6113 subroutine. It is kind of an opposite of @option{-finline-functions}.
6114 This optimization runs at RTL level.
6116 @item -fsignaling-nans
6117 @opindex fsignaling-nans
6118 Compile code assuming that IEEE signaling NaNs may generate user-visible
6119 traps during floating-point operations. Setting this option disables
6120 optimizations that may change the number of exceptions visible with
6121 signaling NaNs. This option implies @option{-ftrapping-math}.
6123 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6126 The default is @option{-fno-signaling-nans}.
6128 This option is experimental and does not currently guarantee to
6129 disable all GCC optimizations that affect signaling NaN behavior.
6131 @item -fsingle-precision-constant
6132 @opindex fsingle-precision-constant
6133 Treat floating point constant as single precision constant instead of
6134 implicitly converting it to double precision constant.
6136 @item -fcx-limited-range
6137 @itemx -fno-cx-limited-range
6138 @opindex fcx-limited-range
6139 @opindex fno-cx-limited-range
6140 When enabled, this option states that a range reduction step is not
6141 needed when performing complex division. The default is
6142 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
6144 This option controls the default setting of the ISO C99
6145 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6150 The following options control optimizations that may improve
6151 performance, but are not enabled by any @option{-O} options. This
6152 section includes experimental options that may produce broken code.
6155 @item -fbranch-probabilities
6156 @opindex fbranch-probabilities
6157 After running a program compiled with @option{-fprofile-arcs}
6158 (@pxref{Debugging Options,, Options for Debugging Your Program or
6159 @command{gcc}}), you can compile it a second time using
6160 @option{-fbranch-probabilities}, to improve optimizations based on
6161 the number of times each branch was taken. When the program
6162 compiled with @option{-fprofile-arcs} exits it saves arc execution
6163 counts to a file called @file{@var{sourcename}.gcda} for each source
6164 file. The information in this data file is very dependent on the
6165 structure of the generated code, so you must use the same source code
6166 and the same optimization options for both compilations.
6168 With @option{-fbranch-probabilities}, GCC puts a
6169 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6170 These can be used to improve optimization. Currently, they are only
6171 used in one place: in @file{reorg.c}, instead of guessing which path a
6172 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6173 exactly determine which path is taken more often.
6175 @item -fprofile-values
6176 @opindex fprofile-values
6177 If combined with @option{-fprofile-arcs}, it adds code so that some
6178 data about values of expressions in the program is gathered.
6180 With @option{-fbranch-probabilities}, it reads back the data gathered
6181 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6182 notes to instructions for their later usage in optimizations.
6184 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6188 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6189 a code to gather information about values of expressions.
6191 With @option{-fbranch-probabilities}, it reads back the data gathered
6192 and actually performs the optimizations based on them.
6193 Currently the optimizations include specialization of division operation
6194 using the knowledge about the value of the denominator.
6196 @item -frename-registers
6197 @opindex frename-registers
6198 Attempt to avoid false dependencies in scheduled code by making use
6199 of registers left over after register allocation. This optimization
6200 will most benefit processors with lots of registers. Depending on the
6201 debug information format adopted by the target, however, it can
6202 make debugging impossible, since variables will no longer stay in
6203 a ``home register''.
6205 Enabled by default with @option{-funroll-loops}.
6209 Perform tail duplication to enlarge superblock size. This transformation
6210 simplifies the control flow of the function allowing other optimizations to do
6213 Enabled with @option{-fprofile-use}.
6215 @item -funroll-loops
6216 @opindex funroll-loops
6217 Unroll loops whose number of iterations can be determined at compile time or
6218 upon entry to the loop. @option{-funroll-loops} implies
6219 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6220 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6221 small constant number of iterations). This option makes code larger, and may
6222 or may not make it run faster.
6224 Enabled with @option{-fprofile-use}.
6226 @item -funroll-all-loops
6227 @opindex funroll-all-loops
6228 Unroll all loops, even if their number of iterations is uncertain when
6229 the loop is entered. This usually makes programs run more slowly.
6230 @option{-funroll-all-loops} implies the same options as
6231 @option{-funroll-loops}.
6234 @opindex fpeel-loops
6235 Peels the loops for that there is enough information that they do not
6236 roll much (from profile feedback). It also turns on complete loop peeling
6237 (i.e.@: complete removal of loops with small constant number of iterations).
6239 Enabled with @option{-fprofile-use}.
6241 @item -fmove-loop-invariants
6242 @opindex fmove-loop-invariants
6243 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6244 at level @option{-O1}
6246 @item -funswitch-loops
6247 @opindex funswitch-loops
6248 Move branches with loop invariant conditions out of the loop, with duplicates
6249 of the loop on both branches (modified according to result of the condition).
6251 @item -ffunction-sections
6252 @itemx -fdata-sections
6253 @opindex ffunction-sections
6254 @opindex fdata-sections
6255 Place each function or data item into its own section in the output
6256 file if the target supports arbitrary sections. The name of the
6257 function or the name of the data item determines the section's name
6260 Use these options on systems where the linker can perform optimizations
6261 to improve locality of reference in the instruction space. Most systems
6262 using the ELF object format and SPARC processors running Solaris 2 have
6263 linkers with such optimizations. AIX may have these optimizations in
6266 Only use these options when there are significant benefits from doing
6267 so. When you specify these options, the assembler and linker will
6268 create larger object and executable files and will also be slower.
6269 You will not be able to use @code{gprof} on all systems if you
6270 specify this option and you may have problems with debugging if
6271 you specify both this option and @option{-g}.
6273 @item -fbranch-target-load-optimize
6274 @opindex fbranch-target-load-optimize
6275 Perform branch target register load optimization before prologue / epilogue
6277 The use of target registers can typically be exposed only during reload,
6278 thus hoisting loads out of loops and doing inter-block scheduling needs
6279 a separate optimization pass.
6281 @item -fbranch-target-load-optimize2
6282 @opindex fbranch-target-load-optimize2
6283 Perform branch target register load optimization after prologue / epilogue
6286 @item -fbtr-bb-exclusive
6287 @opindex fbtr-bb-exclusive
6288 When performing branch target register load optimization, don't reuse
6289 branch target registers in within any basic block.
6291 @item -fstack-protector
6292 Emit extra code to check for buffer overflows, such as stack smashing
6293 attacks. This is done by adding a guard variable to functions with
6294 vulnerable objects. This includes functions that call alloca, and
6295 functions with buffers larger than 8 bytes. The guards are initialized
6296 when a function is entered and then checked when the function exits.
6297 If a guard check fails, an error message is printed and the program exits.
6299 @item -fstack-protector-all
6300 Like @option{-fstack-protector} except that all functions are protected.
6302 @item -fsection-anchors
6303 @opindex fsection-anchors
6304 Try to reduce the number of symbolic address calculations by using
6305 shared ``anchor'' symbols to address nearby objects. This transformation
6306 can help to reduce the number of GOT entries and GOT accesses on some
6309 For example, the implementation of the following function @code{foo}:
6313 int foo (void) @{ return a + b + c; @}
6316 would usually calculate the addresses of all three variables, but if you
6317 compile it with @option{-fsection-anchors}, it will access the variables
6318 from a common anchor point instead. The effect is similar to the
6319 following pseudocode (which isn't valid C):
6324 register int *xr = &x;
6325 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6329 Not all targets support this option.
6331 @item --param @var{name}=@var{value}
6333 In some places, GCC uses various constants to control the amount of
6334 optimization that is done. For example, GCC will not inline functions
6335 that contain more that a certain number of instructions. You can
6336 control some of these constants on the command-line using the
6337 @option{--param} option.
6339 The names of specific parameters, and the meaning of the values, are
6340 tied to the internals of the compiler, and are subject to change
6341 without notice in future releases.
6343 In each case, the @var{value} is an integer. The allowable choices for
6344 @var{name} are given in the following table:
6347 @item salias-max-implicit-fields
6348 The maximum number of fields in a variable without direct
6349 structure accesses for which structure aliasing will consider trying
6350 to track each field. The default is 5
6352 @item salias-max-array-elements
6353 The maximum number of elements an array can have and its elements
6354 still be tracked individually by structure aliasing. The default is 4
6356 @item sra-max-structure-size
6357 The maximum structure size, in bytes, at which the scalar replacement
6358 of aggregates (SRA) optimization will perform block copies. The
6359 default value, 0, implies that GCC will select the most appropriate
6362 @item sra-field-structure-ratio
6363 The threshold ratio (as a percentage) between instantiated fields and
6364 the complete structure size. We say that if the ratio of the number
6365 of bytes in instantiated fields to the number of bytes in the complete
6366 structure exceeds this parameter, then block copies are not used. The
6369 @item max-crossjump-edges
6370 The maximum number of incoming edges to consider for crossjumping.
6371 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6372 the number of edges incoming to each block. Increasing values mean
6373 more aggressive optimization, making the compile time increase with
6374 probably small improvement in executable size.
6376 @item min-crossjump-insns
6377 The minimum number of instructions which must be matched at the end
6378 of two blocks before crossjumping will be performed on them. This
6379 value is ignored in the case where all instructions in the block being
6380 crossjumped from are matched. The default value is 5.
6382 @item max-grow-copy-bb-insns
6383 The maximum code size expansion factor when copying basic blocks
6384 instead of jumping. The expansion is relative to a jump instruction.
6385 The default value is 8.
6387 @item max-goto-duplication-insns
6388 The maximum number of instructions to duplicate to a block that jumps
6389 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6390 passes, GCC factors computed gotos early in the compilation process,
6391 and unfactors them as late as possible. Only computed jumps at the
6392 end of a basic blocks with no more than max-goto-duplication-insns are
6393 unfactored. The default value is 8.
6395 @item max-delay-slot-insn-search
6396 The maximum number of instructions to consider when looking for an
6397 instruction to fill a delay slot. If more than this arbitrary number of
6398 instructions is searched, the time savings from filling the delay slot
6399 will be minimal so stop searching. Increasing values mean more
6400 aggressive optimization, making the compile time increase with probably
6401 small improvement in executable run time.
6403 @item max-delay-slot-live-search
6404 When trying to fill delay slots, the maximum number of instructions to
6405 consider when searching for a block with valid live register
6406 information. Increasing this arbitrarily chosen value means more
6407 aggressive optimization, increasing the compile time. This parameter
6408 should be removed when the delay slot code is rewritten to maintain the
6411 @item max-gcse-memory
6412 The approximate maximum amount of memory that will be allocated in
6413 order to perform the global common subexpression elimination
6414 optimization. If more memory than specified is required, the
6415 optimization will not be done.
6417 @item max-gcse-passes
6418 The maximum number of passes of GCSE to run. The default is 1.
6420 @item max-pending-list-length
6421 The maximum number of pending dependencies scheduling will allow
6422 before flushing the current state and starting over. Large functions
6423 with few branches or calls can create excessively large lists which
6424 needlessly consume memory and resources.
6426 @item max-inline-insns-single
6427 Several parameters control the tree inliner used in gcc.
6428 This number sets the maximum number of instructions (counted in GCC's
6429 internal representation) in a single function that the tree inliner
6430 will consider for inlining. This only affects functions declared
6431 inline and methods implemented in a class declaration (C++).
6432 The default value is 450.
6434 @item max-inline-insns-auto
6435 When you use @option{-finline-functions} (included in @option{-O3}),
6436 a lot of functions that would otherwise not be considered for inlining
6437 by the compiler will be investigated. To those functions, a different
6438 (more restrictive) limit compared to functions declared inline can
6440 The default value is 90.
6442 @item large-function-insns
6443 The limit specifying really large functions. For functions larger than this
6444 limit after inlining inlining is constrained by
6445 @option{--param large-function-growth}. This parameter is useful primarily
6446 to avoid extreme compilation time caused by non-linear algorithms used by the
6448 This parameter is ignored when @option{-funit-at-a-time} is not used.
6449 The default value is 2700.
6451 @item large-function-growth
6452 Specifies maximal growth of large function caused by inlining in percents.
6453 This parameter is ignored when @option{-funit-at-a-time} is not used.
6454 The default value is 100 which limits large function growth to 2.0 times
6457 @item large-unit-insns
6458 The limit specifying large translation unit. Growth caused by inlining of
6459 units larger than this limit is limited by @option{--param inline-unit-growth}.
6460 For small units this might be too tight (consider unit consisting of function A
6461 that is inline and B that just calls A three time. If B is small relative to
6462 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6463 large units consisting of small inlininable functions however the overall unit
6464 growth limit is needed to avoid exponential explosion of code size. Thus for
6465 smaller units, the size is increased to @option{--param large-unit-insns}
6466 before applying @option{--param inline-unit-growth}. The default is 10000
6468 @item inline-unit-growth
6469 Specifies maximal overall growth of the compilation unit caused by inlining.
6470 This parameter is ignored when @option{-funit-at-a-time} is not used.
6471 The default value is 30 which limits unit growth to 1.3 times the original
6474 @item large-stack-frame
6475 The limit specifying large stack frames. While inlining the algorithm is trying
6476 to not grow past this limit too much. Default value is 256 bytes.
6478 @item large-stack-frame-growth
6479 Specifies maximal growth of large stack frames caused by inlining in percents.
6480 The default value is 1000 which limits large stack frame growth to 11 times
6483 @item max-inline-insns-recursive
6484 @itemx max-inline-insns-recursive-auto
6485 Specifies maximum number of instructions out-of-line copy of self recursive inline
6486 function can grow into by performing recursive inlining.
6488 For functions declared inline @option{--param max-inline-insns-recursive} is
6489 taken into account. For function not declared inline, recursive inlining
6490 happens only when @option{-finline-functions} (included in @option{-O3}) is
6491 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6492 default value is 450.
6494 @item max-inline-recursive-depth
6495 @itemx max-inline-recursive-depth-auto
6496 Specifies maximum recursion depth used by the recursive inlining.
6498 For functions declared inline @option{--param max-inline-recursive-depth} is
6499 taken into account. For function not declared inline, recursive inlining
6500 happens only when @option{-finline-functions} (included in @option{-O3}) is
6501 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6502 default value is 450.
6504 @item min-inline-recursive-probability
6505 Recursive inlining is profitable only for function having deep recursion
6506 in average and can hurt for function having little recursion depth by
6507 increasing the prologue size or complexity of function body to other
6510 When profile feedback is available (see @option{-fprofile-generate}) the actual
6511 recursion depth can be guessed from probability that function will recurse via
6512 given call expression. This parameter limits inlining only to call expression
6513 whose probability exceeds given threshold (in percents). The default value is
6516 @item inline-call-cost
6517 Specify cost of call instruction relative to simple arithmetics operations
6518 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6519 functions and at the same time increases size of leaf function that is believed to
6520 reduce function size by being inlined. In effect it increases amount of
6521 inlining for code having large abstraction penalty (many functions that just
6522 pass the arguments to other functions) and decrease inlining for code with low
6523 abstraction penalty. The default value is 16.
6525 @item min-vect-loop-bound
6526 The minimum number of iterations under which a loop will not get vectorized
6527 when @option{-ftree-vectorize} is used. The number of iterations after
6528 vectorization needs to be greater than the value specified by this option
6529 to allow vectorization. The default value is 0.
6531 @item max-unrolled-insns
6532 The maximum number of instructions that a loop should have if that loop
6533 is unrolled, and if the loop is unrolled, it determines how many times
6534 the loop code is unrolled.
6536 @item max-average-unrolled-insns
6537 The maximum number of instructions biased by probabilities of their execution
6538 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6539 it determines how many times the loop code is unrolled.
6541 @item max-unroll-times
6542 The maximum number of unrollings of a single loop.
6544 @item max-peeled-insns
6545 The maximum number of instructions that a loop should have if that loop
6546 is peeled, and if the loop is peeled, it determines how many times
6547 the loop code is peeled.
6549 @item max-peel-times
6550 The maximum number of peelings of a single loop.
6552 @item max-completely-peeled-insns
6553 The maximum number of insns of a completely peeled loop.
6555 @item max-completely-peel-times
6556 The maximum number of iterations of a loop to be suitable for complete peeling.
6558 @item max-unswitch-insns
6559 The maximum number of insns of an unswitched loop.
6561 @item max-unswitch-level
6562 The maximum number of branches unswitched in a single loop.
6565 The minimum cost of an expensive expression in the loop invariant motion.
6567 @item iv-consider-all-candidates-bound
6568 Bound on number of candidates for induction variables below that
6569 all candidates are considered for each use in induction variable
6570 optimizations. Only the most relevant candidates are considered
6571 if there are more candidates, to avoid quadratic time complexity.
6573 @item iv-max-considered-uses
6574 The induction variable optimizations give up on loops that contain more
6575 induction variable uses.
6577 @item iv-always-prune-cand-set-bound
6578 If number of candidates in the set is smaller than this value,
6579 we always try to remove unnecessary ivs from the set during its
6580 optimization when a new iv is added to the set.
6582 @item scev-max-expr-size
6583 Bound on size of expressions used in the scalar evolutions analyzer.
6584 Large expressions slow the analyzer.
6586 @item omega-max-vars
6587 The maximum number of variables in an Omega constraint system.
6588 The default value is 128.
6590 @item omega-max-geqs
6591 The maximum number of inequalities in an Omega constraint system.
6592 The default value is 256.
6595 The maximum number of equalities in an Omega constraint system.
6596 The default value is 128.
6598 @item omega-max-wild-cards
6599 The maximum number of wildcard variables that the Omega solver will
6600 be able to insert. The default value is 18.
6602 @item omega-hash-table-size
6603 The size of the hash table in the Omega solver. The default value is
6606 @item omega-max-keys
6607 The maximal number of keys used by the Omega solver. The default
6610 @item omega-eliminate-redundant-constraints
6611 When set to 1, use expensive methods to eliminate all redundant
6612 constraints. The default value is 0.
6614 @item vect-max-version-checks
6615 The maximum number of runtime checks that can be performed when doing
6616 loop versioning in the vectorizer. See option ftree-vect-loop-version
6617 for more information.
6619 @item max-iterations-to-track
6621 The maximum number of iterations of a loop the brute force algorithm
6622 for analysis of # of iterations of the loop tries to evaluate.
6624 @item hot-bb-count-fraction
6625 Select fraction of the maximal count of repetitions of basic block in program
6626 given basic block needs to have to be considered hot.
6628 @item hot-bb-frequency-fraction
6629 Select fraction of the maximal frequency of executions of basic block in
6630 function given basic block needs to have to be considered hot
6632 @item max-predicted-iterations
6633 The maximum number of loop iterations we predict statically. This is useful
6634 in cases where function contain single loop with known bound and other loop
6635 with unknown. We predict the known number of iterations correctly, while
6636 the unknown number of iterations average to roughly 10. This means that the
6637 loop without bounds would appear artificially cold relative to the other one.
6639 @item tracer-dynamic-coverage
6640 @itemx tracer-dynamic-coverage-feedback
6642 This value is used to limit superblock formation once the given percentage of
6643 executed instructions is covered. This limits unnecessary code size
6646 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6647 feedback is available. The real profiles (as opposed to statically estimated
6648 ones) are much less balanced allowing the threshold to be larger value.
6650 @item tracer-max-code-growth
6651 Stop tail duplication once code growth has reached given percentage. This is
6652 rather hokey argument, as most of the duplicates will be eliminated later in
6653 cross jumping, so it may be set to much higher values than is the desired code
6656 @item tracer-min-branch-ratio
6658 Stop reverse growth when the reverse probability of best edge is less than this
6659 threshold (in percent).
6661 @item tracer-min-branch-ratio
6662 @itemx tracer-min-branch-ratio-feedback
6664 Stop forward growth if the best edge do have probability lower than this
6667 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6668 compilation for profile feedback and one for compilation without. The value
6669 for compilation with profile feedback needs to be more conservative (higher) in
6670 order to make tracer effective.
6672 @item max-cse-path-length
6674 Maximum number of basic blocks on path that cse considers. The default is 10.
6677 The maximum instructions CSE process before flushing. The default is 1000.
6679 @item max-aliased-vops
6681 Maximum number of virtual operands per function allowed to represent
6682 aliases before triggering the alias partitioning heuristic. Alias
6683 partitioning reduces compile times and memory consumption needed for
6684 aliasing at the expense of precision loss in alias information. The
6685 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
6688 Notice that if a function contains more memory statements than the
6689 value of this parameter, it is not really possible to achieve this
6690 reduction. In this case, the compiler will use the number of memory
6691 statements as the value for @option{max-aliased-vops}.
6693 @item avg-aliased-vops
6695 Average number of virtual operands per statement allowed to represent
6696 aliases before triggering the alias partitioning heuristic. This
6697 works in conjunction with @option{max-aliased-vops}. If a function
6698 contains more than @option{max-aliased-vops} virtual operators, then
6699 memory symbols will be grouped into memory partitions until either the
6700 total number of virtual operators is below @option{max-aliased-vops}
6701 or the average number of virtual operators per memory statement is
6702 below @option{avg-aliased-vops}. The default value for this parameter
6703 is 1 for -O1 and -O2, and 3 for -O3.
6705 @item ggc-min-expand
6707 GCC uses a garbage collector to manage its own memory allocation. This
6708 parameter specifies the minimum percentage by which the garbage
6709 collector's heap should be allowed to expand between collections.
6710 Tuning this may improve compilation speed; it has no effect on code
6713 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6714 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6715 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6716 GCC is not able to calculate RAM on a particular platform, the lower
6717 bound of 30% is used. Setting this parameter and
6718 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6719 every opportunity. This is extremely slow, but can be useful for
6722 @item ggc-min-heapsize
6724 Minimum size of the garbage collector's heap before it begins bothering
6725 to collect garbage. The first collection occurs after the heap expands
6726 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6727 tuning this may improve compilation speed, and has no effect on code
6730 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6731 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6732 with a lower bound of 4096 (four megabytes) and an upper bound of
6733 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6734 particular platform, the lower bound is used. Setting this parameter
6735 very large effectively disables garbage collection. Setting this
6736 parameter and @option{ggc-min-expand} to zero causes a full collection
6737 to occur at every opportunity.
6739 @item max-reload-search-insns
6740 The maximum number of instruction reload should look backward for equivalent
6741 register. Increasing values mean more aggressive optimization, making the
6742 compile time increase with probably slightly better performance. The default
6745 @item max-cselib-memory-locations
6746 The maximum number of memory locations cselib should take into account.
6747 Increasing values mean more aggressive optimization, making the compile time
6748 increase with probably slightly better performance. The default value is 500.
6750 @item max-flow-memory-locations
6751 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6752 The default value is 100.
6754 @item reorder-blocks-duplicate
6755 @itemx reorder-blocks-duplicate-feedback
6757 Used by basic block reordering pass to decide whether to use unconditional
6758 branch or duplicate the code on its destination. Code is duplicated when its
6759 estimated size is smaller than this value multiplied by the estimated size of
6760 unconditional jump in the hot spots of the program.
6762 The @option{reorder-block-duplicate-feedback} is used only when profile
6763 feedback is available and may be set to higher values than
6764 @option{reorder-block-duplicate} since information about the hot spots is more
6767 @item max-sched-ready-insns
6768 The maximum number of instructions ready to be issued the scheduler should
6769 consider at any given time during the first scheduling pass. Increasing
6770 values mean more thorough searches, making the compilation time increase
6771 with probably little benefit. The default value is 100.
6773 @item max-sched-region-blocks
6774 The maximum number of blocks in a region to be considered for
6775 interblock scheduling. The default value is 10.
6777 @item max-sched-region-insns
6778 The maximum number of insns in a region to be considered for
6779 interblock scheduling. The default value is 100.
6782 The minimum probability (in percents) of reaching a source block
6783 for interblock speculative scheduling. The default value is 40.
6785 @item max-sched-extend-regions-iters
6786 The maximum number of iterations through CFG to extend regions.
6787 0 - disable region extension,
6788 N - do at most N iterations.
6789 The default value is 0.
6791 @item max-sched-insn-conflict-delay
6792 The maximum conflict delay for an insn to be considered for speculative motion.
6793 The default value is 3.
6795 @item sched-spec-prob-cutoff
6796 The minimal probability of speculation success (in percents), so that
6797 speculative insn will be scheduled.
6798 The default value is 40.
6800 @item max-last-value-rtl
6802 The maximum size measured as number of RTLs that can be recorded in an expression
6803 in combiner for a pseudo register as last known value of that register. The default
6806 @item integer-share-limit
6807 Small integer constants can use a shared data structure, reducing the
6808 compiler's memory usage and increasing its speed. This sets the maximum
6809 value of a shared integer constant's. The default value is 256.
6811 @item min-virtual-mappings
6812 Specifies the minimum number of virtual mappings in the incremental
6813 SSA updater that should be registered to trigger the virtual mappings
6814 heuristic defined by virtual-mappings-ratio. The default value is
6817 @item virtual-mappings-ratio
6818 If the number of virtual mappings is virtual-mappings-ratio bigger
6819 than the number of virtual symbols to be updated, then the incremental
6820 SSA updater switches to a full update for those symbols. The default
6823 @item ssp-buffer-size
6824 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6825 protection when @option{-fstack-protection} is used.
6827 @item max-jump-thread-duplication-stmts
6828 Maximum number of statements allowed in a block that needs to be
6829 duplicated when threading jumps.
6831 @item max-fields-for-field-sensitive
6832 Maximum number of fields in a structure we will treat in
6833 a field sensitive manner during pointer analysis.
6835 @item prefetch-latency
6836 Estimate on average number of instructions that are executed before
6837 prefetch finishes. The distance we prefetch ahead is proportional
6838 to this constant. Increasing this number may also lead to less
6839 streams being prefetched (see @option{simultaneous-prefetches}).
6841 @item simultaneous-prefetches
6842 Maximum number of prefetches that can run at the same time.
6844 @item l1-cache-line-size
6845 The size of cache line in L1 cache, in bytes.
6848 The number of cache lines in L1 cache.
6850 @item verify-canonical-types
6851 Whether the compiler should verify the ``canonical'' types used for
6852 type equality comparisons within the C++ and Objective-C++ front
6853 ends. Set to 1 (the default when GCC is configured with
6854 --enable-checking) to enable verification, 0 to disable verification
6855 (the default when GCC is configured with --disable-checking).
6860 @node Preprocessor Options
6861 @section Options Controlling the Preprocessor
6862 @cindex preprocessor options
6863 @cindex options, preprocessor
6865 These options control the C preprocessor, which is run on each C source
6866 file before actual compilation.
6868 If you use the @option{-E} option, nothing is done except preprocessing.
6869 Some of these options make sense only together with @option{-E} because
6870 they cause the preprocessor output to be unsuitable for actual
6875 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6876 and pass @var{option} directly through to the preprocessor. If
6877 @var{option} contains commas, it is split into multiple options at the
6878 commas. However, many options are modified, translated or interpreted
6879 by the compiler driver before being passed to the preprocessor, and
6880 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6881 interface is undocumented and subject to change, so whenever possible
6882 you should avoid using @option{-Wp} and let the driver handle the
6885 @item -Xpreprocessor @var{option}
6886 @opindex preprocessor
6887 Pass @var{option} as an option to the preprocessor. You can use this to
6888 supply system-specific preprocessor options which GCC does not know how to
6891 If you want to pass an option that takes an argument, you must use
6892 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6895 @include cppopts.texi
6897 @node Assembler Options
6898 @section Passing Options to the Assembler
6900 @c prevent bad page break with this line
6901 You can pass options to the assembler.
6904 @item -Wa,@var{option}
6906 Pass @var{option} as an option to the assembler. If @var{option}
6907 contains commas, it is split into multiple options at the commas.
6909 @item -Xassembler @var{option}
6911 Pass @var{option} as an option to the assembler. You can use this to
6912 supply system-specific assembler options which GCC does not know how to
6915 If you want to pass an option that takes an argument, you must use
6916 @option{-Xassembler} twice, once for the option and once for the argument.
6921 @section Options for Linking
6922 @cindex link options
6923 @cindex options, linking
6925 These options come into play when the compiler links object files into
6926 an executable output file. They are meaningless if the compiler is
6927 not doing a link step.
6931 @item @var{object-file-name}
6932 A file name that does not end in a special recognized suffix is
6933 considered to name an object file or library. (Object files are
6934 distinguished from libraries by the linker according to the file
6935 contents.) If linking is done, these object files are used as input
6944 If any of these options is used, then the linker is not run, and
6945 object file names should not be used as arguments. @xref{Overall
6949 @item -l@var{library}
6950 @itemx -l @var{library}
6952 Search the library named @var{library} when linking. (The second
6953 alternative with the library as a separate argument is only for
6954 POSIX compliance and is not recommended.)
6956 It makes a difference where in the command you write this option; the
6957 linker searches and processes libraries and object files in the order they
6958 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6959 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6960 to functions in @samp{z}, those functions may not be loaded.
6962 The linker searches a standard list of directories for the library,
6963 which is actually a file named @file{lib@var{library}.a}. The linker
6964 then uses this file as if it had been specified precisely by name.
6966 The directories searched include several standard system directories
6967 plus any that you specify with @option{-L}.
6969 Normally the files found this way are library files---archive files
6970 whose members are object files. The linker handles an archive file by
6971 scanning through it for members which define symbols that have so far
6972 been referenced but not defined. But if the file that is found is an
6973 ordinary object file, it is linked in the usual fashion. The only
6974 difference between using an @option{-l} option and specifying a file name
6975 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6976 and searches several directories.
6980 You need this special case of the @option{-l} option in order to
6981 link an Objective-C or Objective-C++ program.
6984 @opindex nostartfiles
6985 Do not use the standard system startup files when linking.
6986 The standard system libraries are used normally, unless @option{-nostdlib}
6987 or @option{-nodefaultlibs} is used.
6989 @item -nodefaultlibs
6990 @opindex nodefaultlibs
6991 Do not use the standard system libraries when linking.
6992 Only the libraries you specify will be passed to the linker.
6993 The standard startup files are used normally, unless @option{-nostartfiles}
6994 is used. The compiler may generate calls to @code{memcmp},
6995 @code{memset}, @code{memcpy} and @code{memmove}.
6996 These entries are usually resolved by entries in
6997 libc. These entry points should be supplied through some other
6998 mechanism when this option is specified.
7002 Do not use the standard system startup files or libraries when linking.
7003 No startup files and only the libraries you specify will be passed to
7004 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7005 @code{memcpy} and @code{memmove}.
7006 These entries are usually resolved by entries in
7007 libc. These entry points should be supplied through some other
7008 mechanism when this option is specified.
7010 @cindex @option{-lgcc}, use with @option{-nostdlib}
7011 @cindex @option{-nostdlib} and unresolved references
7012 @cindex unresolved references and @option{-nostdlib}
7013 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7014 @cindex @option{-nodefaultlibs} and unresolved references
7015 @cindex unresolved references and @option{-nodefaultlibs}
7016 One of the standard libraries bypassed by @option{-nostdlib} and
7017 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7018 that GCC uses to overcome shortcomings of particular machines, or special
7019 needs for some languages.
7020 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7021 Collection (GCC) Internals},
7022 for more discussion of @file{libgcc.a}.)
7023 In most cases, you need @file{libgcc.a} even when you want to avoid
7024 other standard libraries. In other words, when you specify @option{-nostdlib}
7025 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7026 This ensures that you have no unresolved references to internal GCC
7027 library subroutines. (For example, @samp{__main}, used to ensure C++
7028 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7029 GNU Compiler Collection (GCC) Internals}.)
7033 Produce a position independent executable on targets which support it.
7034 For predictable results, you must also specify the same set of options
7035 that were used to generate code (@option{-fpie}, @option{-fPIE},
7036 or model suboptions) when you specify this option.
7040 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7041 that support it. This instructs the linker to add all symbols, not
7042 only used ones, to the dynamic symbol table. This option is needed
7043 for some uses of @code{dlopen} or to allow obtaining backtraces
7044 from within a program.
7048 Remove all symbol table and relocation information from the executable.
7052 On systems that support dynamic linking, this prevents linking with the shared
7053 libraries. On other systems, this option has no effect.
7057 Produce a shared object which can then be linked with other objects to
7058 form an executable. Not all systems support this option. For predictable
7059 results, you must also specify the same set of options that were used to
7060 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7061 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7062 needs to build supplementary stub code for constructors to work. On
7063 multi-libbed systems, @samp{gcc -shared} must select the correct support
7064 libraries to link against. Failing to supply the correct flags may lead
7065 to subtle defects. Supplying them in cases where they are not necessary
7068 @item -shared-libgcc
7069 @itemx -static-libgcc
7070 @opindex shared-libgcc
7071 @opindex static-libgcc
7072 On systems that provide @file{libgcc} as a shared library, these options
7073 force the use of either the shared or static version respectively.
7074 If no shared version of @file{libgcc} was built when the compiler was
7075 configured, these options have no effect.
7077 There are several situations in which an application should use the
7078 shared @file{libgcc} instead of the static version. The most common
7079 of these is when the application wishes to throw and catch exceptions
7080 across different shared libraries. In that case, each of the libraries
7081 as well as the application itself should use the shared @file{libgcc}.
7083 Therefore, the G++ and GCJ drivers automatically add
7084 @option{-shared-libgcc} whenever you build a shared library or a main
7085 executable, because C++ and Java programs typically use exceptions, so
7086 this is the right thing to do.
7088 If, instead, you use the GCC driver to create shared libraries, you may
7089 find that they will not always be linked with the shared @file{libgcc}.
7090 If GCC finds, at its configuration time, that you have a non-GNU linker
7091 or a GNU linker that does not support option @option{--eh-frame-hdr},
7092 it will link the shared version of @file{libgcc} into shared libraries
7093 by default. Otherwise, it will take advantage of the linker and optimize
7094 away the linking with the shared version of @file{libgcc}, linking with
7095 the static version of libgcc by default. This allows exceptions to
7096 propagate through such shared libraries, without incurring relocation
7097 costs at library load time.
7099 However, if a library or main executable is supposed to throw or catch
7100 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7101 for the languages used in the program, or using the option
7102 @option{-shared-libgcc}, such that it is linked with the shared
7107 Bind references to global symbols when building a shared object. Warn
7108 about any unresolved references (unless overridden by the link editor
7109 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7112 @item -Xlinker @var{option}
7114 Pass @var{option} as an option to the linker. You can use this to
7115 supply system-specific linker options which GCC does not know how to
7118 If you want to pass an option that takes an argument, you must use
7119 @option{-Xlinker} twice, once for the option and once for the argument.
7120 For example, to pass @option{-assert definitions}, you must write
7121 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7122 @option{-Xlinker "-assert definitions"}, because this passes the entire
7123 string as a single argument, which is not what the linker expects.
7125 @item -Wl,@var{option}
7127 Pass @var{option} as an option to the linker. If @var{option} contains
7128 commas, it is split into multiple options at the commas.
7130 @item -u @var{symbol}
7132 Pretend the symbol @var{symbol} is undefined, to force linking of
7133 library modules to define it. You can use @option{-u} multiple times with
7134 different symbols to force loading of additional library modules.
7137 @node Directory Options
7138 @section Options for Directory Search
7139 @cindex directory options
7140 @cindex options, directory search
7143 These options specify directories to search for header files, for
7144 libraries and for parts of the compiler:
7149 Add the directory @var{dir} to the head of the list of directories to be
7150 searched for header files. This can be used to override a system header
7151 file, substituting your own version, since these directories are
7152 searched before the system header file directories. However, you should
7153 not use this option to add directories that contain vendor-supplied
7154 system header files (use @option{-isystem} for that). If you use more than
7155 one @option{-I} option, the directories are scanned in left-to-right
7156 order; the standard system directories come after.
7158 If a standard system include directory, or a directory specified with
7159 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7160 option will be ignored. The directory will still be searched but as a
7161 system directory at its normal position in the system include chain.
7162 This is to ensure that GCC's procedure to fix buggy system headers and
7163 the ordering for the include_next directive are not inadvertently changed.
7164 If you really need to change the search order for system directories,
7165 use the @option{-nostdinc} and/or @option{-isystem} options.
7167 @item -iquote@var{dir}
7169 Add the directory @var{dir} to the head of the list of directories to
7170 be searched for header files only for the case of @samp{#include
7171 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7172 otherwise just like @option{-I}.
7176 Add directory @var{dir} to the list of directories to be searched
7179 @item -B@var{prefix}
7181 This option specifies where to find the executables, libraries,
7182 include files, and data files of the compiler itself.
7184 The compiler driver program runs one or more of the subprograms
7185 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7186 @var{prefix} as a prefix for each program it tries to run, both with and
7187 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7189 For each subprogram to be run, the compiler driver first tries the
7190 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7191 was not specified, the driver tries two standard prefixes, which are
7192 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7193 those results in a file name that is found, the unmodified program
7194 name is searched for using the directories specified in your
7195 @env{PATH} environment variable.
7197 The compiler will check to see if the path provided by the @option{-B}
7198 refers to a directory, and if necessary it will add a directory
7199 separator character at the end of the path.
7201 @option{-B} prefixes that effectively specify directory names also apply
7202 to libraries in the linker, because the compiler translates these
7203 options into @option{-L} options for the linker. They also apply to
7204 includes files in the preprocessor, because the compiler translates these
7205 options into @option{-isystem} options for the preprocessor. In this case,
7206 the compiler appends @samp{include} to the prefix.
7208 The run-time support file @file{libgcc.a} can also be searched for using
7209 the @option{-B} prefix, if needed. If it is not found there, the two
7210 standard prefixes above are tried, and that is all. The file is left
7211 out of the link if it is not found by those means.
7213 Another way to specify a prefix much like the @option{-B} prefix is to use
7214 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7217 As a special kludge, if the path provided by @option{-B} is
7218 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7219 9, then it will be replaced by @file{[dir/]include}. This is to help
7220 with boot-strapping the compiler.
7222 @item -specs=@var{file}
7224 Process @var{file} after the compiler reads in the standard @file{specs}
7225 file, in order to override the defaults that the @file{gcc} driver
7226 program uses when determining what switches to pass to @file{cc1},
7227 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7228 @option{-specs=@var{file}} can be specified on the command line, and they
7229 are processed in order, from left to right.
7231 @item --sysroot=@var{dir}
7233 Use @var{dir} as the logical root directory for headers and libraries.
7234 For example, if the compiler would normally search for headers in
7235 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7236 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7238 If you use both this option and the @option{-isysroot} option, then
7239 the @option{--sysroot} option will apply to libraries, but the
7240 @option{-isysroot} option will apply to header files.
7242 The GNU linker (beginning with version 2.16) has the necessary support
7243 for this option. If your linker does not support this option, the
7244 header file aspect of @option{--sysroot} will still work, but the
7245 library aspect will not.
7249 This option has been deprecated. Please use @option{-iquote} instead for
7250 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7251 Any directories you specify with @option{-I} options before the @option{-I-}
7252 option are searched only for the case of @samp{#include "@var{file}"};
7253 they are not searched for @samp{#include <@var{file}>}.
7255 If additional directories are specified with @option{-I} options after
7256 the @option{-I-}, these directories are searched for all @samp{#include}
7257 directives. (Ordinarily @emph{all} @option{-I} directories are used
7260 In addition, the @option{-I-} option inhibits the use of the current
7261 directory (where the current input file came from) as the first search
7262 directory for @samp{#include "@var{file}"}. There is no way to
7263 override this effect of @option{-I-}. With @option{-I.} you can specify
7264 searching the directory which was current when the compiler was
7265 invoked. That is not exactly the same as what the preprocessor does
7266 by default, but it is often satisfactory.
7268 @option{-I-} does not inhibit the use of the standard system directories
7269 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7276 @section Specifying subprocesses and the switches to pass to them
7279 @command{gcc} is a driver program. It performs its job by invoking a
7280 sequence of other programs to do the work of compiling, assembling and
7281 linking. GCC interprets its command-line parameters and uses these to
7282 deduce which programs it should invoke, and which command-line options
7283 it ought to place on their command lines. This behavior is controlled
7284 by @dfn{spec strings}. In most cases there is one spec string for each
7285 program that GCC can invoke, but a few programs have multiple spec
7286 strings to control their behavior. The spec strings built into GCC can
7287 be overridden by using the @option{-specs=} command-line switch to specify
7290 @dfn{Spec files} are plaintext files that are used to construct spec
7291 strings. They consist of a sequence of directives separated by blank
7292 lines. The type of directive is determined by the first non-whitespace
7293 character on the line and it can be one of the following:
7296 @item %@var{command}
7297 Issues a @var{command} to the spec file processor. The commands that can
7301 @item %include <@var{file}>
7303 Search for @var{file} and insert its text at the current point in the
7306 @item %include_noerr <@var{file}>
7307 @cindex %include_noerr
7308 Just like @samp{%include}, but do not generate an error message if the include
7309 file cannot be found.
7311 @item %rename @var{old_name} @var{new_name}
7313 Rename the spec string @var{old_name} to @var{new_name}.
7317 @item *[@var{spec_name}]:
7318 This tells the compiler to create, override or delete the named spec
7319 string. All lines after this directive up to the next directive or
7320 blank line are considered to be the text for the spec string. If this
7321 results in an empty string then the spec will be deleted. (Or, if the
7322 spec did not exist, then nothing will happened.) Otherwise, if the spec
7323 does not currently exist a new spec will be created. If the spec does
7324 exist then its contents will be overridden by the text of this
7325 directive, unless the first character of that text is the @samp{+}
7326 character, in which case the text will be appended to the spec.
7328 @item [@var{suffix}]:
7329 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7330 and up to the next directive or blank line are considered to make up the
7331 spec string for the indicated suffix. When the compiler encounters an
7332 input file with the named suffix, it will processes the spec string in
7333 order to work out how to compile that file. For example:
7340 This says that any input file whose name ends in @samp{.ZZ} should be
7341 passed to the program @samp{z-compile}, which should be invoked with the
7342 command-line switch @option{-input} and with the result of performing the
7343 @samp{%i} substitution. (See below.)
7345 As an alternative to providing a spec string, the text that follows a
7346 suffix directive can be one of the following:
7349 @item @@@var{language}
7350 This says that the suffix is an alias for a known @var{language}. This is
7351 similar to using the @option{-x} command-line switch to GCC to specify a
7352 language explicitly. For example:
7359 Says that .ZZ files are, in fact, C++ source files.
7362 This causes an error messages saying:
7365 @var{name} compiler not installed on this system.
7369 GCC already has an extensive list of suffixes built into it.
7370 This directive will add an entry to the end of the list of suffixes, but
7371 since the list is searched from the end backwards, it is effectively
7372 possible to override earlier entries using this technique.
7376 GCC has the following spec strings built into it. Spec files can
7377 override these strings or create their own. Note that individual
7378 targets can also add their own spec strings to this list.
7381 asm Options to pass to the assembler
7382 asm_final Options to pass to the assembler post-processor
7383 cpp Options to pass to the C preprocessor
7384 cc1 Options to pass to the C compiler
7385 cc1plus Options to pass to the C++ compiler
7386 endfile Object files to include at the end of the link
7387 link Options to pass to the linker
7388 lib Libraries to include on the command line to the linker
7389 libgcc Decides which GCC support library to pass to the linker
7390 linker Sets the name of the linker
7391 predefines Defines to be passed to the C preprocessor
7392 signed_char Defines to pass to CPP to say whether @code{char} is signed
7394 startfile Object files to include at the start of the link
7397 Here is a small example of a spec file:
7403 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7406 This example renames the spec called @samp{lib} to @samp{old_lib} and
7407 then overrides the previous definition of @samp{lib} with a new one.
7408 The new definition adds in some extra command-line options before
7409 including the text of the old definition.
7411 @dfn{Spec strings} are a list of command-line options to be passed to their
7412 corresponding program. In addition, the spec strings can contain
7413 @samp{%}-prefixed sequences to substitute variable text or to
7414 conditionally insert text into the command line. Using these constructs
7415 it is possible to generate quite complex command lines.
7417 Here is a table of all defined @samp{%}-sequences for spec
7418 strings. Note that spaces are not generated automatically around the
7419 results of expanding these sequences. Therefore you can concatenate them
7420 together or combine them with constant text in a single argument.
7424 Substitute one @samp{%} into the program name or argument.
7427 Substitute the name of the input file being processed.
7430 Substitute the basename of the input file being processed.
7431 This is the substring up to (and not including) the last period
7432 and not including the directory.
7435 This is the same as @samp{%b}, but include the file suffix (text after
7439 Marks the argument containing or following the @samp{%d} as a
7440 temporary file name, so that that file will be deleted if GCC exits
7441 successfully. Unlike @samp{%g}, this contributes no text to the
7444 @item %g@var{suffix}
7445 Substitute a file name that has suffix @var{suffix} and is chosen
7446 once per compilation, and mark the argument in the same way as
7447 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7448 name is now chosen in a way that is hard to predict even when previously
7449 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7450 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7451 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7452 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7453 was simply substituted with a file name chosen once per compilation,
7454 without regard to any appended suffix (which was therefore treated
7455 just like ordinary text), making such attacks more likely to succeed.
7457 @item %u@var{suffix}
7458 Like @samp{%g}, but generates a new temporary file name even if
7459 @samp{%u@var{suffix}} was already seen.
7461 @item %U@var{suffix}
7462 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7463 new one if there is no such last file name. In the absence of any
7464 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7465 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7466 would involve the generation of two distinct file names, one
7467 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7468 simply substituted with a file name chosen for the previous @samp{%u},
7469 without regard to any appended suffix.
7471 @item %j@var{suffix}
7472 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7473 writable, and if save-temps is off; otherwise, substitute the name
7474 of a temporary file, just like @samp{%u}. This temporary file is not
7475 meant for communication between processes, but rather as a junk
7478 @item %|@var{suffix}
7479 @itemx %m@var{suffix}
7480 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7481 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7482 all. These are the two most common ways to instruct a program that it
7483 should read from standard input or write to standard output. If you
7484 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7485 construct: see for example @file{f/lang-specs.h}.
7487 @item %.@var{SUFFIX}
7488 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7489 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7490 terminated by the next space or %.
7493 Marks the argument containing or following the @samp{%w} as the
7494 designated output file of this compilation. This puts the argument
7495 into the sequence of arguments that @samp{%o} will substitute later.
7498 Substitutes the names of all the output files, with spaces
7499 automatically placed around them. You should write spaces
7500 around the @samp{%o} as well or the results are undefined.
7501 @samp{%o} is for use in the specs for running the linker.
7502 Input files whose names have no recognized suffix are not compiled
7503 at all, but they are included among the output files, so they will
7507 Substitutes the suffix for object files. Note that this is
7508 handled specially when it immediately follows @samp{%g, %u, or %U},
7509 because of the need for those to form complete file names. The
7510 handling is such that @samp{%O} is treated exactly as if it had already
7511 been substituted, except that @samp{%g, %u, and %U} do not currently
7512 support additional @var{suffix} characters following @samp{%O} as they would
7513 following, for example, @samp{.o}.
7516 Substitutes the standard macro predefinitions for the
7517 current target machine. Use this when running @code{cpp}.
7520 Like @samp{%p}, but puts @samp{__} before and after the name of each
7521 predefined macro, except for macros that start with @samp{__} or with
7522 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7526 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7527 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7528 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7529 and @option{-imultilib} as necessary.
7532 Current argument is the name of a library or startup file of some sort.
7533 Search for that file in a standard list of directories and substitute
7534 the full name found.
7537 Print @var{str} as an error message. @var{str} is terminated by a newline.
7538 Use this when inconsistent options are detected.
7541 Substitute the contents of spec string @var{name} at this point.
7544 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7546 @item %x@{@var{option}@}
7547 Accumulate an option for @samp{%X}.
7550 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7554 Output the accumulated assembler options specified by @option{-Wa}.
7557 Output the accumulated preprocessor options specified by @option{-Wp}.
7560 Process the @code{asm} spec. This is used to compute the
7561 switches to be passed to the assembler.
7564 Process the @code{asm_final} spec. This is a spec string for
7565 passing switches to an assembler post-processor, if such a program is
7569 Process the @code{link} spec. This is the spec for computing the
7570 command line passed to the linker. Typically it will make use of the
7571 @samp{%L %G %S %D and %E} sequences.
7574 Dump out a @option{-L} option for each directory that GCC believes might
7575 contain startup files. If the target supports multilibs then the
7576 current multilib directory will be prepended to each of these paths.
7579 Process the @code{lib} spec. This is a spec string for deciding which
7580 libraries should be included on the command line to the linker.
7583 Process the @code{libgcc} spec. This is a spec string for deciding
7584 which GCC support library should be included on the command line to the linker.
7587 Process the @code{startfile} spec. This is a spec for deciding which
7588 object files should be the first ones passed to the linker. Typically
7589 this might be a file named @file{crt0.o}.
7592 Process the @code{endfile} spec. This is a spec string that specifies
7593 the last object files that will be passed to the linker.
7596 Process the @code{cpp} spec. This is used to construct the arguments
7597 to be passed to the C preprocessor.
7600 Process the @code{cc1} spec. This is used to construct the options to be
7601 passed to the actual C compiler (@samp{cc1}).
7604 Process the @code{cc1plus} spec. This is used to construct the options to be
7605 passed to the actual C++ compiler (@samp{cc1plus}).
7608 Substitute the variable part of a matched option. See below.
7609 Note that each comma in the substituted string is replaced by
7613 Remove all occurrences of @code{-S} from the command line. Note---this
7614 command is position dependent. @samp{%} commands in the spec string
7615 before this one will see @code{-S}, @samp{%} commands in the spec string
7616 after this one will not.
7618 @item %:@var{function}(@var{args})
7619 Call the named function @var{function}, passing it @var{args}.
7620 @var{args} is first processed as a nested spec string, then split
7621 into an argument vector in the usual fashion. The function returns
7622 a string which is processed as if it had appeared literally as part
7623 of the current spec.
7625 The following built-in spec functions are provided:
7629 The @code{getenv} spec function takes two arguments: an environment
7630 variable name and a string. If the environment variable is not
7631 defined, a fatal error is issued. Otherwise, the return value is the
7632 value of the environment variable concatenated with the string. For
7633 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
7636 %:getenv(TOPDIR /include)
7639 expands to @file{/path/to/top/include}.
7641 @item @code{if-exists}
7642 The @code{if-exists} spec function takes one argument, an absolute
7643 pathname to a file. If the file exists, @code{if-exists} returns the
7644 pathname. Here is a small example of its usage:
7648 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7651 @item @code{if-exists-else}
7652 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7653 spec function, except that it takes two arguments. The first argument is
7654 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7655 returns the pathname. If it does not exist, it returns the second argument.
7656 This way, @code{if-exists-else} can be used to select one file or another,
7657 based on the existence of the first. Here is a small example of its usage:
7661 crt0%O%s %:if-exists(crti%O%s) \
7662 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7665 @item @code{replace-outfile}
7666 The @code{replace-outfile} spec function takes two arguments. It looks for the
7667 first argument in the outfiles array and replaces it with the second argument. Here
7668 is a small example of its usage:
7671 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7674 @item @code{print-asm-header}
7675 The @code{print-asm-header} function takes no arguments and simply
7676 prints a banner like:
7682 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
7685 It is used to separate compiler options from assembler options
7686 in the @option{--target-help} output.
7690 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7691 If that switch was not specified, this substitutes nothing. Note that
7692 the leading dash is omitted when specifying this option, and it is
7693 automatically inserted if the substitution is performed. Thus the spec
7694 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7695 and would output the command line option @option{-foo}.
7697 @item %W@{@code{S}@}
7698 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7701 @item %@{@code{S}*@}
7702 Substitutes all the switches specified to GCC whose names start
7703 with @code{-S}, but which also take an argument. This is used for
7704 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7705 GCC considers @option{-o foo} as being
7706 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7707 text, including the space. Thus two arguments would be generated.
7709 @item %@{@code{S}*&@code{T}*@}
7710 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7711 (the order of @code{S} and @code{T} in the spec is not significant).
7712 There can be any number of ampersand-separated variables; for each the
7713 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7715 @item %@{@code{S}:@code{X}@}
7716 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7718 @item %@{!@code{S}:@code{X}@}
7719 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7721 @item %@{@code{S}*:@code{X}@}
7722 Substitutes @code{X} if one or more switches whose names start with
7723 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7724 once, no matter how many such switches appeared. However, if @code{%*}
7725 appears somewhere in @code{X}, then @code{X} will be substituted once
7726 for each matching switch, with the @code{%*} replaced by the part of
7727 that switch that matched the @code{*}.
7729 @item %@{.@code{S}:@code{X}@}
7730 Substitutes @code{X}, if processing a file with suffix @code{S}.
7732 @item %@{!.@code{S}:@code{X}@}
7733 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7735 @item %@{,@code{S}:@code{X}@}
7736 Substitutes @code{X}, if processing a file for language @code{S}.
7738 @item %@{!,@code{S}:@code{X}@}
7739 Substitutes @code{X}, if not processing a file for language @code{S}.
7741 @item %@{@code{S}|@code{P}:@code{X}@}
7742 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
7743 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
7744 @code{*} sequences as well, although they have a stronger binding than
7745 the @samp{|}. If @code{%*} appears in @code{X}, all of the
7746 alternatives must be starred, and only the first matching alternative
7749 For example, a spec string like this:
7752 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7755 will output the following command-line options from the following input
7756 command-line options:
7761 -d fred.c -foo -baz -boggle
7762 -d jim.d -bar -baz -boggle
7765 @item %@{S:X; T:Y; :D@}
7767 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7768 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7769 be as many clauses as you need. This may be combined with @code{.},
7770 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
7775 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7776 construct may contain other nested @samp{%} constructs or spaces, or
7777 even newlines. They are processed as usual, as described above.
7778 Trailing white space in @code{X} is ignored. White space may also
7779 appear anywhere on the left side of the colon in these constructs,
7780 except between @code{.} or @code{*} and the corresponding word.
7782 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7783 handled specifically in these constructs. If another value of
7784 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7785 @option{-W} switch is found later in the command line, the earlier
7786 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7787 just one letter, which passes all matching options.
7789 The character @samp{|} at the beginning of the predicate text is used to
7790 indicate that a command should be piped to the following command, but
7791 only if @option{-pipe} is specified.
7793 It is built into GCC which switches take arguments and which do not.
7794 (You might think it would be useful to generalize this to allow each
7795 compiler's spec to say which switches take arguments. But this cannot
7796 be done in a consistent fashion. GCC cannot even decide which input
7797 files have been specified without knowing which switches take arguments,
7798 and it must know which input files to compile in order to tell which
7801 GCC also knows implicitly that arguments starting in @option{-l} are to be
7802 treated as compiler output files, and passed to the linker in their
7803 proper position among the other output files.
7805 @c man begin OPTIONS
7807 @node Target Options
7808 @section Specifying Target Machine and Compiler Version
7809 @cindex target options
7810 @cindex cross compiling
7811 @cindex specifying machine version
7812 @cindex specifying compiler version and target machine
7813 @cindex compiler version, specifying
7814 @cindex target machine, specifying
7816 The usual way to run GCC is to run the executable called @file{gcc}, or
7817 @file{<machine>-gcc} when cross-compiling, or
7818 @file{<machine>-gcc-<version>} to run a version other than the one that
7819 was installed last. Sometimes this is inconvenient, so GCC provides
7820 options that will switch to another cross-compiler or version.
7823 @item -b @var{machine}
7825 The argument @var{machine} specifies the target machine for compilation.
7827 The value to use for @var{machine} is the same as was specified as the
7828 machine type when configuring GCC as a cross-compiler. For
7829 example, if a cross-compiler was configured with @samp{configure
7830 arm-elf}, meaning to compile for an arm processor with elf binaries,
7831 then you would specify @option{-b arm-elf} to run that cross compiler.
7832 Because there are other options beginning with @option{-b}, the
7833 configuration must contain a hyphen.
7835 @item -V @var{version}
7837 The argument @var{version} specifies which version of GCC to run.
7838 This is useful when multiple versions are installed. For example,
7839 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7842 The @option{-V} and @option{-b} options work by running the
7843 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7844 use them if you can just run that directly.
7846 @node Submodel Options
7847 @section Hardware Models and Configurations
7848 @cindex submodel options
7849 @cindex specifying hardware config
7850 @cindex hardware models and configurations, specifying
7851 @cindex machine dependent options
7853 Earlier we discussed the standard option @option{-b} which chooses among
7854 different installed compilers for completely different target
7855 machines, such as VAX vs.@: 68000 vs.@: 80386.
7857 In addition, each of these target machine types can have its own
7858 special options, starting with @samp{-m}, to choose among various
7859 hardware models or configurations---for example, 68010 vs 68020,
7860 floating coprocessor or none. A single installed version of the
7861 compiler can compile for any model or configuration, according to the
7864 Some configurations of the compiler also support additional special
7865 options, usually for compatibility with other compilers on the same
7868 @c This list is ordered alphanumerically by subsection name.
7869 @c It should be the same order and spelling as these options are listed
7870 @c in Machine Dependent Options
7876 * Blackfin Options::
7880 * DEC Alpha Options::
7881 * DEC Alpha/VMS Options::
7883 * GNU/Linux Options::
7886 * i386 and x86-64 Options::
7899 * RS/6000 and PowerPC Options::
7900 * S/390 and zSeries Options::
7905 * System V Options::
7906 * TMS320C3x/C4x Options::
7911 * Xstormy16 Options::
7917 @subsection ARC Options
7920 These options are defined for ARC implementations:
7925 Compile code for little endian mode. This is the default.
7929 Compile code for big endian mode.
7932 @opindex mmangle-cpu
7933 Prepend the name of the cpu to all public symbol names.
7934 In multiple-processor systems, there are many ARC variants with different
7935 instruction and register set characteristics. This flag prevents code
7936 compiled for one cpu to be linked with code compiled for another.
7937 No facility exists for handling variants that are ``almost identical''.
7938 This is an all or nothing option.
7940 @item -mcpu=@var{cpu}
7942 Compile code for ARC variant @var{cpu}.
7943 Which variants are supported depend on the configuration.
7944 All variants support @option{-mcpu=base}, this is the default.
7946 @item -mtext=@var{text-section}
7947 @itemx -mdata=@var{data-section}
7948 @itemx -mrodata=@var{readonly-data-section}
7952 Put functions, data, and readonly data in @var{text-section},
7953 @var{data-section}, and @var{readonly-data-section} respectively
7954 by default. This can be overridden with the @code{section} attribute.
7955 @xref{Variable Attributes}.
7960 @subsection ARM Options
7963 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7967 @item -mabi=@var{name}
7969 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7970 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7973 @opindex mapcs-frame
7974 Generate a stack frame that is compliant with the ARM Procedure Call
7975 Standard for all functions, even if this is not strictly necessary for
7976 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7977 with this option will cause the stack frames not to be generated for
7978 leaf functions. The default is @option{-mno-apcs-frame}.
7982 This is a synonym for @option{-mapcs-frame}.
7985 @c not currently implemented
7986 @item -mapcs-stack-check
7987 @opindex mapcs-stack-check
7988 Generate code to check the amount of stack space available upon entry to
7989 every function (that actually uses some stack space). If there is
7990 insufficient space available then either the function
7991 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7992 called, depending upon the amount of stack space required. The run time
7993 system is required to provide these functions. The default is
7994 @option{-mno-apcs-stack-check}, since this produces smaller code.
7996 @c not currently implemented
7998 @opindex mapcs-float
7999 Pass floating point arguments using the float point registers. This is
8000 one of the variants of the APCS@. This option is recommended if the
8001 target hardware has a floating point unit or if a lot of floating point
8002 arithmetic is going to be performed by the code. The default is
8003 @option{-mno-apcs-float}, since integer only code is slightly increased in
8004 size if @option{-mapcs-float} is used.
8006 @c not currently implemented
8007 @item -mapcs-reentrant
8008 @opindex mapcs-reentrant
8009 Generate reentrant, position independent code. The default is
8010 @option{-mno-apcs-reentrant}.
8013 @item -mthumb-interwork
8014 @opindex mthumb-interwork
8015 Generate code which supports calling between the ARM and Thumb
8016 instruction sets. Without this option the two instruction sets cannot
8017 be reliably used inside one program. The default is
8018 @option{-mno-thumb-interwork}, since slightly larger code is generated
8019 when @option{-mthumb-interwork} is specified.
8021 @item -mno-sched-prolog
8022 @opindex mno-sched-prolog
8023 Prevent the reordering of instructions in the function prolog, or the
8024 merging of those instruction with the instructions in the function's
8025 body. This means that all functions will start with a recognizable set
8026 of instructions (or in fact one of a choice from a small set of
8027 different function prologues), and this information can be used to
8028 locate the start if functions inside an executable piece of code. The
8029 default is @option{-msched-prolog}.
8032 @opindex mhard-float
8033 Generate output containing floating point instructions. This is the
8037 @opindex msoft-float
8038 Generate output containing library calls for floating point.
8039 @strong{Warning:} the requisite libraries are not available for all ARM
8040 targets. Normally the facilities of the machine's usual C compiler are
8041 used, but this cannot be done directly in cross-compilation. You must make
8042 your own arrangements to provide suitable library functions for
8045 @option{-msoft-float} changes the calling convention in the output file;
8046 therefore, it is only useful if you compile @emph{all} of a program with
8047 this option. In particular, you need to compile @file{libgcc.a}, the
8048 library that comes with GCC, with @option{-msoft-float} in order for
8051 @item -mfloat-abi=@var{name}
8053 Specifies which ABI to use for floating point values. Permissible values
8054 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8056 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8057 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8058 of floating point instructions, but still uses the soft-float calling
8061 @item -mlittle-endian
8062 @opindex mlittle-endian
8063 Generate code for a processor running in little-endian mode. This is
8064 the default for all standard configurations.
8067 @opindex mbig-endian
8068 Generate code for a processor running in big-endian mode; the default is
8069 to compile code for a little-endian processor.
8071 @item -mwords-little-endian
8072 @opindex mwords-little-endian
8073 This option only applies when generating code for big-endian processors.
8074 Generate code for a little-endian word order but a big-endian byte
8075 order. That is, a byte order of the form @samp{32107654}. Note: this
8076 option should only be used if you require compatibility with code for
8077 big-endian ARM processors generated by versions of the compiler prior to
8080 @item -mcpu=@var{name}
8082 This specifies the name of the target ARM processor. GCC uses this name
8083 to determine what kind of instructions it can emit when generating
8084 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8085 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8086 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8087 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8088 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8089 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8090 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8091 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8092 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8093 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8094 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8095 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8096 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8097 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8098 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
8099 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8101 @itemx -mtune=@var{name}
8103 This option is very similar to the @option{-mcpu=} option, except that
8104 instead of specifying the actual target processor type, and hence
8105 restricting which instructions can be used, it specifies that GCC should
8106 tune the performance of the code as if the target were of the type
8107 specified in this option, but still choosing the instructions that it
8108 will generate based on the cpu specified by a @option{-mcpu=} option.
8109 For some ARM implementations better performance can be obtained by using
8112 @item -march=@var{name}
8114 This specifies the name of the target ARM architecture. GCC uses this
8115 name to determine what kind of instructions it can emit when generating
8116 assembly code. This option can be used in conjunction with or instead
8117 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8118 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8119 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8120 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
8121 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
8123 @item -mfpu=@var{name}
8124 @itemx -mfpe=@var{number}
8125 @itemx -mfp=@var{number}
8129 This specifies what floating point hardware (or hardware emulation) is
8130 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8131 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8132 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8133 with older versions of GCC@.
8135 If @option{-msoft-float} is specified this specifies the format of
8136 floating point values.
8138 @item -mstructure-size-boundary=@var{n}
8139 @opindex mstructure-size-boundary
8140 The size of all structures and unions will be rounded up to a multiple
8141 of the number of bits set by this option. Permissible values are 8, 32
8142 and 64. The default value varies for different toolchains. For the COFF
8143 targeted toolchain the default value is 8. A value of 64 is only allowed
8144 if the underlying ABI supports it.
8146 Specifying the larger number can produce faster, more efficient code, but
8147 can also increase the size of the program. Different values are potentially
8148 incompatible. Code compiled with one value cannot necessarily expect to
8149 work with code or libraries compiled with another value, if they exchange
8150 information using structures or unions.
8152 @item -mabort-on-noreturn
8153 @opindex mabort-on-noreturn
8154 Generate a call to the function @code{abort} at the end of a
8155 @code{noreturn} function. It will be executed if the function tries to
8159 @itemx -mno-long-calls
8160 @opindex mlong-calls
8161 @opindex mno-long-calls
8162 Tells the compiler to perform function calls by first loading the
8163 address of the function into a register and then performing a subroutine
8164 call on this register. This switch is needed if the target function
8165 will lie outside of the 64 megabyte addressing range of the offset based
8166 version of subroutine call instruction.
8168 Even if this switch is enabled, not all function calls will be turned
8169 into long calls. The heuristic is that static functions, functions
8170 which have the @samp{short-call} attribute, functions that are inside
8171 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8172 definitions have already been compiled within the current compilation
8173 unit, will not be turned into long calls. The exception to this rule is
8174 that weak function definitions, functions with the @samp{long-call}
8175 attribute or the @samp{section} attribute, and functions that are within
8176 the scope of a @samp{#pragma long_calls} directive, will always be
8177 turned into long calls.
8179 This feature is not enabled by default. Specifying
8180 @option{-mno-long-calls} will restore the default behavior, as will
8181 placing the function calls within the scope of a @samp{#pragma
8182 long_calls_off} directive. Note these switches have no effect on how
8183 the compiler generates code to handle function calls via function
8186 @item -mnop-fun-dllimport
8187 @opindex mnop-fun-dllimport
8188 Disable support for the @code{dllimport} attribute.
8190 @item -msingle-pic-base
8191 @opindex msingle-pic-base
8192 Treat the register used for PIC addressing as read-only, rather than
8193 loading it in the prologue for each function. The run-time system is
8194 responsible for initializing this register with an appropriate value
8195 before execution begins.
8197 @item -mpic-register=@var{reg}
8198 @opindex mpic-register
8199 Specify the register to be used for PIC addressing. The default is R10
8200 unless stack-checking is enabled, when R9 is used.
8202 @item -mcirrus-fix-invalid-insns
8203 @opindex mcirrus-fix-invalid-insns
8204 @opindex mno-cirrus-fix-invalid-insns
8205 Insert NOPs into the instruction stream to in order to work around
8206 problems with invalid Maverick instruction combinations. This option
8207 is only valid if the @option{-mcpu=ep9312} option has been used to
8208 enable generation of instructions for the Cirrus Maverick floating
8209 point co-processor. This option is not enabled by default, since the
8210 problem is only present in older Maverick implementations. The default
8211 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8214 @item -mpoke-function-name
8215 @opindex mpoke-function-name
8216 Write the name of each function into the text section, directly
8217 preceding the function prologue. The generated code is similar to this:
8221 .ascii "arm_poke_function_name", 0
8224 .word 0xff000000 + (t1 - t0)
8225 arm_poke_function_name
8227 stmfd sp!, @{fp, ip, lr, pc@}
8231 When performing a stack backtrace, code can inspect the value of
8232 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8233 location @code{pc - 12} and the top 8 bits are set, then we know that
8234 there is a function name embedded immediately preceding this location
8235 and has length @code{((pc[-3]) & 0xff000000)}.
8239 Generate code for the Thumb instruction set. The default is to
8240 use the 32-bit ARM instruction set.
8241 This option automatically enables either 16-bit Thumb-1 or
8242 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8243 and @option{-march=@var{name}} options.
8246 @opindex mtpcs-frame
8247 Generate a stack frame that is compliant with the Thumb Procedure Call
8248 Standard for all non-leaf functions. (A leaf function is one that does
8249 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8251 @item -mtpcs-leaf-frame
8252 @opindex mtpcs-leaf-frame
8253 Generate a stack frame that is compliant with the Thumb Procedure Call
8254 Standard for all leaf functions. (A leaf function is one that does
8255 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8257 @item -mcallee-super-interworking
8258 @opindex mcallee-super-interworking
8259 Gives all externally visible functions in the file being compiled an ARM
8260 instruction set header which switches to Thumb mode before executing the
8261 rest of the function. This allows these functions to be called from
8262 non-interworking code.
8264 @item -mcaller-super-interworking
8265 @opindex mcaller-super-interworking
8266 Allows calls via function pointers (including virtual functions) to
8267 execute correctly regardless of whether the target code has been
8268 compiled for interworking or not. There is a small overhead in the cost
8269 of executing a function pointer if this option is enabled.
8271 @item -mtp=@var{name}
8273 Specify the access model for the thread local storage pointer. The valid
8274 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8275 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8276 (supported in the arm6k architecture), and @option{auto}, which uses the
8277 best available method for the selected processor. The default setting is
8283 @subsection AVR Options
8286 These options are defined for AVR implementations:
8289 @item -mmcu=@var{mcu}
8291 Specify ATMEL AVR instruction set or MCU type.
8293 Instruction set avr1 is for the minimal AVR core, not supported by the C
8294 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8295 attiny11, attiny12, attiny15, attiny28).
8297 Instruction set avr2 (default) is for the classic AVR core with up to
8298 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8299 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8300 at90c8534, at90s8535).
8302 Instruction set avr3 is for the classic AVR core with up to 128K program
8303 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8305 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8306 memory space (MCU types: atmega8, atmega83, atmega85).
8308 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8309 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8310 atmega64, atmega128, at43usb355, at94k).
8314 Output instruction sizes to the asm file.
8316 @item -minit-stack=@var{N}
8317 @opindex minit-stack
8318 Specify the initial stack address, which may be a symbol or numeric value,
8319 @samp{__stack} is the default.
8321 @item -mno-interrupts
8322 @opindex mno-interrupts
8323 Generated code is not compatible with hardware interrupts.
8324 Code size will be smaller.
8326 @item -mcall-prologues
8327 @opindex mcall-prologues
8328 Functions prologues/epilogues expanded as call to appropriate
8329 subroutines. Code size will be smaller.
8331 @item -mno-tablejump
8332 @opindex mno-tablejump
8333 Do not generate tablejump insns which sometimes increase code size.
8336 @opindex mtiny-stack
8337 Change only the low 8 bits of the stack pointer.
8341 Assume int to be 8 bit integer. This affects the sizes of all types: A
8342 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8343 and long long will be 4 bytes. Please note that this option does not
8344 comply to the C standards, but it will provide you with smaller code
8348 @node Blackfin Options
8349 @subsection Blackfin Options
8350 @cindex Blackfin Options
8353 @item -mcpu=@var{cpu}
8355 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8356 can be one of @samp{bf531}, @samp{bf532}, @samp{bf533},
8357 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf561}.
8358 Without this option, @samp{bf532} is used as the processor by default.
8359 The corresponding predefined processor macros for @var{cpu} is to
8360 be defined. For the @samp{bfin-elf} toolchain, this causes the hardware
8361 BSP provided by libgloss to be linked in if @samp{-msim} is not given.
8362 Support for @samp{bf561} is incomplete; only the processor macro is defined.
8366 Specifies that the program will be run on the simulator. This causes
8367 the simulator BSP provided by libgloss to be linked in. This option
8368 has effect only for @samp{bfin-elf} toolchain.
8370 @item -momit-leaf-frame-pointer
8371 @opindex momit-leaf-frame-pointer
8372 Don't keep the frame pointer in a register for leaf functions. This
8373 avoids the instructions to save, set up and restore frame pointers and
8374 makes an extra register available in leaf functions. The option
8375 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8376 which might make debugging harder.
8378 @item -mspecld-anomaly
8379 @opindex mspecld-anomaly
8380 When enabled, the compiler will ensure that the generated code does not
8381 contain speculative loads after jump instructions. This option is enabled
8384 @item -mno-specld-anomaly
8385 @opindex mno-specld-anomaly
8386 Don't generate extra code to prevent speculative loads from occurring.
8388 @item -mcsync-anomaly
8389 @opindex mcsync-anomaly
8390 When enabled, the compiler will ensure that the generated code does not
8391 contain CSYNC or SSYNC instructions too soon after conditional branches.
8392 This option is enabled by default.
8394 @item -mno-csync-anomaly
8395 @opindex mno-csync-anomaly
8396 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8397 occurring too soon after a conditional branch.
8401 When enabled, the compiler is free to take advantage of the knowledge that
8402 the entire program fits into the low 64k of memory.
8405 @opindex mno-low-64k
8406 Assume that the program is arbitrarily large. This is the default.
8408 @item -mstack-check-l1
8409 @opindex mstack-check-l1
8410 Do stack checking using information placed into L1 scratchpad memory by the
8413 @item -mid-shared-library
8414 @opindex mid-shared-library
8415 Generate code that supports shared libraries via the library ID method.
8416 This allows for execute in place and shared libraries in an environment
8417 without virtual memory management. This option implies @option{-fPIC}.
8419 @item -mno-id-shared-library
8420 @opindex mno-id-shared-library
8421 Generate code that doesn't assume ID based shared libraries are being used.
8422 This is the default.
8424 @item -mleaf-id-shared-library
8425 @opindex mleaf-id-shared-library
8426 Generate code that supports shared libraries via the library ID method,
8427 but assumes that this library or executable won't link against any other
8428 ID shared libraries. That allows the compiler to use faster code for jumps
8431 @item -mno-leaf-id-shared-library
8432 @opindex mno-leaf-id-shared-library
8433 Do not assume that the code being compiled won't link against any ID shared
8434 libraries. Slower code will be generated for jump and call insns.
8436 @item -mshared-library-id=n
8437 @opindex mshared-library-id
8438 Specified the identification number of the ID based shared library being
8439 compiled. Specifying a value of 0 will generate more compact code, specifying
8440 other values will force the allocation of that number to the current
8441 library but is no more space or time efficient than omitting this option.
8445 Generate code that allows the data segment to be located in a different
8446 area of memory from the text segment. This allows for execute in place in
8447 an environment without virtual memory management by eliminating relocations
8448 against the text section.
8451 @opindex mno-sep-data
8452 Generate code that assumes that the data segment follows the text segment.
8453 This is the default.
8456 @itemx -mno-long-calls
8457 @opindex mlong-calls
8458 @opindex mno-long-calls
8459 Tells the compiler to perform function calls by first loading the
8460 address of the function into a register and then performing a subroutine
8461 call on this register. This switch is needed if the target function
8462 will lie outside of the 24 bit addressing range of the offset based
8463 version of subroutine call instruction.
8465 This feature is not enabled by default. Specifying
8466 @option{-mno-long-calls} will restore the default behavior. Note these
8467 switches have no effect on how the compiler generates code to handle
8468 function calls via function pointers.
8472 @subsection CRIS Options
8473 @cindex CRIS Options
8475 These options are defined specifically for the CRIS ports.
8478 @item -march=@var{architecture-type}
8479 @itemx -mcpu=@var{architecture-type}
8482 Generate code for the specified architecture. The choices for
8483 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8484 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8485 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8488 @item -mtune=@var{architecture-type}
8490 Tune to @var{architecture-type} everything applicable about the generated
8491 code, except for the ABI and the set of available instructions. The
8492 choices for @var{architecture-type} are the same as for
8493 @option{-march=@var{architecture-type}}.
8495 @item -mmax-stack-frame=@var{n}
8496 @opindex mmax-stack-frame
8497 Warn when the stack frame of a function exceeds @var{n} bytes.
8499 @item -melinux-stacksize=@var{n}
8500 @opindex melinux-stacksize
8501 Only available with the @samp{cris-axis-aout} target. Arranges for
8502 indications in the program to the kernel loader that the stack of the
8503 program should be set to @var{n} bytes.
8509 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8510 @option{-march=v3} and @option{-march=v8} respectively.
8512 @item -mmul-bug-workaround
8513 @itemx -mno-mul-bug-workaround
8514 @opindex mmul-bug-workaround
8515 @opindex mno-mul-bug-workaround
8516 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8517 models where it applies. This option is active by default.
8521 Enable CRIS-specific verbose debug-related information in the assembly
8522 code. This option also has the effect to turn off the @samp{#NO_APP}
8523 formatted-code indicator to the assembler at the beginning of the
8528 Do not use condition-code results from previous instruction; always emit
8529 compare and test instructions before use of condition codes.
8531 @item -mno-side-effects
8532 @opindex mno-side-effects
8533 Do not emit instructions with side-effects in addressing modes other than
8537 @itemx -mno-stack-align
8539 @itemx -mno-data-align
8540 @itemx -mconst-align
8541 @itemx -mno-const-align
8542 @opindex mstack-align
8543 @opindex mno-stack-align
8544 @opindex mdata-align
8545 @opindex mno-data-align
8546 @opindex mconst-align
8547 @opindex mno-const-align
8548 These options (no-options) arranges (eliminate arrangements) for the
8549 stack-frame, individual data and constants to be aligned for the maximum
8550 single data access size for the chosen CPU model. The default is to
8551 arrange for 32-bit alignment. ABI details such as structure layout are
8552 not affected by these options.
8560 Similar to the stack- data- and const-align options above, these options
8561 arrange for stack-frame, writable data and constants to all be 32-bit,
8562 16-bit or 8-bit aligned. The default is 32-bit alignment.
8564 @item -mno-prologue-epilogue
8565 @itemx -mprologue-epilogue
8566 @opindex mno-prologue-epilogue
8567 @opindex mprologue-epilogue
8568 With @option{-mno-prologue-epilogue}, the normal function prologue and
8569 epilogue that sets up the stack-frame are omitted and no return
8570 instructions or return sequences are generated in the code. Use this
8571 option only together with visual inspection of the compiled code: no
8572 warnings or errors are generated when call-saved registers must be saved,
8573 or storage for local variable needs to be allocated.
8579 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8580 instruction sequences that load addresses for functions from the PLT part
8581 of the GOT rather than (traditional on other architectures) calls to the
8582 PLT@. The default is @option{-mgotplt}.
8586 Legacy no-op option only recognized with the cris-axis-aout target.
8590 Legacy no-op option only recognized with the cris-axis-elf and
8591 cris-axis-linux-gnu targets.
8595 Only recognized with the cris-axis-aout target, where it selects a
8596 GNU/linux-like multilib, include files and instruction set for
8601 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8605 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8606 to link with input-output functions from a simulator library. Code,
8607 initialized data and zero-initialized data are allocated consecutively.
8611 Like @option{-sim}, but pass linker options to locate initialized data at
8612 0x40000000 and zero-initialized data at 0x80000000.
8616 @subsection CRX Options
8619 These options are defined specifically for the CRX ports.
8625 Enable the use of multiply-accumulate instructions. Disabled by default.
8629 Push instructions will be used to pass outgoing arguments when functions
8630 are called. Enabled by default.
8633 @node Darwin Options
8634 @subsection Darwin Options
8635 @cindex Darwin options
8637 These options are defined for all architectures running the Darwin operating
8640 FSF GCC on Darwin does not create ``fat'' object files; it will create
8641 an object file for the single architecture that it was built to
8642 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8643 @option{-arch} options are used; it does so by running the compiler or
8644 linker multiple times and joining the results together with
8647 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8648 @samp{i686}) is determined by the flags that specify the ISA
8649 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8650 @option{-force_cpusubtype_ALL} option can be used to override this.
8652 The Darwin tools vary in their behavior when presented with an ISA
8653 mismatch. The assembler, @file{as}, will only permit instructions to
8654 be used that are valid for the subtype of the file it is generating,
8655 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8656 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8657 and print an error if asked to create a shared library with a less
8658 restrictive subtype than its input files (for instance, trying to put
8659 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8660 for executables, @file{ld}, will quietly give the executable the most
8661 restrictive subtype of any of its input files.
8666 Add the framework directory @var{dir} to the head of the list of
8667 directories to be searched for header files. These directories are
8668 interleaved with those specified by @option{-I} options and are
8669 scanned in a left-to-right order.
8671 A framework directory is a directory with frameworks in it. A
8672 framework is a directory with a @samp{"Headers"} and/or
8673 @samp{"PrivateHeaders"} directory contained directly in it that ends
8674 in @samp{".framework"}. The name of a framework is the name of this
8675 directory excluding the @samp{".framework"}. Headers associated with
8676 the framework are found in one of those two directories, with
8677 @samp{"Headers"} being searched first. A subframework is a framework
8678 directory that is in a framework's @samp{"Frameworks"} directory.
8679 Includes of subframework headers can only appear in a header of a
8680 framework that contains the subframework, or in a sibling subframework
8681 header. Two subframeworks are siblings if they occur in the same
8682 framework. A subframework should not have the same name as a
8683 framework, a warning will be issued if this is violated. Currently a
8684 subframework cannot have subframeworks, in the future, the mechanism
8685 may be extended to support this. The standard frameworks can be found
8686 in @samp{"/System/Library/Frameworks"} and
8687 @samp{"/Library/Frameworks"}. An example include looks like
8688 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8689 the name of the framework and header.h is found in the
8690 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8692 @item -iframework@var{dir}
8694 Like @option{-F} except the directory is a treated as a system
8695 directory. The main difference between this @option{-iframework} and
8696 @option{-F} is that with @option{-iframework} the compiler does not
8697 warn about constructs contained within header files found via
8698 @var{dir}. This option is valid only for the C family of languages.
8702 Emit debugging information for symbols that are used. For STABS
8703 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8704 This is by default ON@.
8708 Emit debugging information for all symbols and types.
8710 @item -mmacosx-version-min=@var{version}
8711 The earliest version of MacOS X that this executable will run on
8712 is @var{version}. Typical values of @var{version} include @code{10.1},
8713 @code{10.2}, and @code{10.3.9}.
8715 The default for this option is to make choices that seem to be most
8720 Enable kernel development mode. The @option{-mkernel} option sets
8721 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8722 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8723 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8724 applicable. This mode also sets @option{-mno-altivec},
8725 @option{-msoft-float}, @option{-fno-builtin} and
8726 @option{-mlong-branch} for PowerPC targets.
8728 @item -mone-byte-bool
8729 @opindex mone-byte-bool
8730 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8731 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8732 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8733 option has no effect on x86.
8735 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8736 to generate code that is not binary compatible with code generated
8737 without that switch. Using this switch may require recompiling all
8738 other modules in a program, including system libraries. Use this
8739 switch to conform to a non-default data model.
8741 @item -mfix-and-continue
8742 @itemx -ffix-and-continue
8743 @itemx -findirect-data
8744 @opindex mfix-and-continue
8745 @opindex ffix-and-continue
8746 @opindex findirect-data
8747 Generate code suitable for fast turn around development. Needed to
8748 enable gdb to dynamically load @code{.o} files into already running
8749 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8750 are provided for backwards compatibility.
8754 Loads all members of static archive libraries.
8755 See man ld(1) for more information.
8757 @item -arch_errors_fatal
8758 @opindex arch_errors_fatal
8759 Cause the errors having to do with files that have the wrong architecture
8763 @opindex bind_at_load
8764 Causes the output file to be marked such that the dynamic linker will
8765 bind all undefined references when the file is loaded or launched.
8769 Produce a Mach-o bundle format file.
8770 See man ld(1) for more information.
8772 @item -bundle_loader @var{executable}
8773 @opindex bundle_loader
8774 This option specifies the @var{executable} that will be loading the build
8775 output file being linked. See man ld(1) for more information.
8779 When passed this option, GCC will produce a dynamic library instead of
8780 an executable when linking, using the Darwin @file{libtool} command.
8782 @item -force_cpusubtype_ALL
8783 @opindex force_cpusubtype_ALL
8784 This causes GCC's output file to have the @var{ALL} subtype, instead of
8785 one controlled by the @option{-mcpu} or @option{-march} option.
8787 @item -allowable_client @var{client_name}
8789 @itemx -compatibility_version
8790 @itemx -current_version
8792 @itemx -dependency-file
8794 @itemx -dylinker_install_name
8796 @itemx -exported_symbols_list
8798 @itemx -flat_namespace
8799 @itemx -force_flat_namespace
8800 @itemx -headerpad_max_install_names
8803 @itemx -install_name
8804 @itemx -keep_private_externs
8805 @itemx -multi_module
8806 @itemx -multiply_defined
8807 @itemx -multiply_defined_unused
8809 @itemx -no_dead_strip_inits_and_terms
8810 @itemx -nofixprebinding
8813 @itemx -noseglinkedit
8814 @itemx -pagezero_size
8816 @itemx -prebind_all_twolevel_modules
8817 @itemx -private_bundle
8818 @itemx -read_only_relocs
8820 @itemx -sectobjectsymbols
8824 @itemx -sectobjectsymbols
8827 @itemx -segs_read_only_addr
8828 @itemx -segs_read_write_addr
8829 @itemx -seg_addr_table
8830 @itemx -seg_addr_table_filename
8833 @itemx -segs_read_only_addr
8834 @itemx -segs_read_write_addr
8835 @itemx -single_module
8838 @itemx -sub_umbrella
8839 @itemx -twolevel_namespace
8842 @itemx -unexported_symbols_list
8843 @itemx -weak_reference_mismatches
8846 @opindex allowable_client
8847 @opindex client_name
8848 @opindex compatibility_version
8849 @opindex current_version
8851 @opindex dependency-file
8853 @opindex dylinker_install_name
8855 @opindex exported_symbols_list
8857 @opindex flat_namespace
8858 @opindex force_flat_namespace
8859 @opindex headerpad_max_install_names
8862 @opindex install_name
8863 @opindex keep_private_externs
8864 @opindex multi_module
8865 @opindex multiply_defined
8866 @opindex multiply_defined_unused
8868 @opindex no_dead_strip_inits_and_terms
8869 @opindex nofixprebinding
8870 @opindex nomultidefs
8872 @opindex noseglinkedit
8873 @opindex pagezero_size
8875 @opindex prebind_all_twolevel_modules
8876 @opindex private_bundle
8877 @opindex read_only_relocs
8879 @opindex sectobjectsymbols
8883 @opindex sectobjectsymbols
8886 @opindex segs_read_only_addr
8887 @opindex segs_read_write_addr
8888 @opindex seg_addr_table
8889 @opindex seg_addr_table_filename
8890 @opindex seglinkedit
8892 @opindex segs_read_only_addr
8893 @opindex segs_read_write_addr
8894 @opindex single_module
8896 @opindex sub_library
8897 @opindex sub_umbrella
8898 @opindex twolevel_namespace
8901 @opindex unexported_symbols_list
8902 @opindex weak_reference_mismatches
8903 @opindex whatsloaded
8905 These options are passed to the Darwin linker. The Darwin linker man page
8906 describes them in detail.
8909 @node DEC Alpha Options
8910 @subsection DEC Alpha Options
8912 These @samp{-m} options are defined for the DEC Alpha implementations:
8915 @item -mno-soft-float
8917 @opindex mno-soft-float
8918 @opindex msoft-float
8919 Use (do not use) the hardware floating-point instructions for
8920 floating-point operations. When @option{-msoft-float} is specified,
8921 functions in @file{libgcc.a} will be used to perform floating-point
8922 operations. Unless they are replaced by routines that emulate the
8923 floating-point operations, or compiled in such a way as to call such
8924 emulations routines, these routines will issue floating-point
8925 operations. If you are compiling for an Alpha without floating-point
8926 operations, you must ensure that the library is built so as not to call
8929 Note that Alpha implementations without floating-point operations are
8930 required to have floating-point registers.
8935 @opindex mno-fp-regs
8936 Generate code that uses (does not use) the floating-point register set.
8937 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8938 register set is not used, floating point operands are passed in integer
8939 registers as if they were integers and floating-point results are passed
8940 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8941 so any function with a floating-point argument or return value called by code
8942 compiled with @option{-mno-fp-regs} must also be compiled with that
8945 A typical use of this option is building a kernel that does not use,
8946 and hence need not save and restore, any floating-point registers.
8950 The Alpha architecture implements floating-point hardware optimized for
8951 maximum performance. It is mostly compliant with the IEEE floating
8952 point standard. However, for full compliance, software assistance is
8953 required. This option generates code fully IEEE compliant code
8954 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8955 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8956 defined during compilation. The resulting code is less efficient but is
8957 able to correctly support denormalized numbers and exceptional IEEE
8958 values such as not-a-number and plus/minus infinity. Other Alpha
8959 compilers call this option @option{-ieee_with_no_inexact}.
8961 @item -mieee-with-inexact
8962 @opindex mieee-with-inexact
8963 This is like @option{-mieee} except the generated code also maintains
8964 the IEEE @var{inexact-flag}. Turning on this option causes the
8965 generated code to implement fully-compliant IEEE math. In addition to
8966 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8967 macro. On some Alpha implementations the resulting code may execute
8968 significantly slower than the code generated by default. Since there is
8969 very little code that depends on the @var{inexact-flag}, you should
8970 normally not specify this option. Other Alpha compilers call this
8971 option @option{-ieee_with_inexact}.
8973 @item -mfp-trap-mode=@var{trap-mode}
8974 @opindex mfp-trap-mode
8975 This option controls what floating-point related traps are enabled.
8976 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8977 The trap mode can be set to one of four values:
8981 This is the default (normal) setting. The only traps that are enabled
8982 are the ones that cannot be disabled in software (e.g., division by zero
8986 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8990 Like @samp{u}, but the instructions are marked to be safe for software
8991 completion (see Alpha architecture manual for details).
8994 Like @samp{su}, but inexact traps are enabled as well.
8997 @item -mfp-rounding-mode=@var{rounding-mode}
8998 @opindex mfp-rounding-mode
8999 Selects the IEEE rounding mode. Other Alpha compilers call this option
9000 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9005 Normal IEEE rounding mode. Floating point numbers are rounded towards
9006 the nearest machine number or towards the even machine number in case
9010 Round towards minus infinity.
9013 Chopped rounding mode. Floating point numbers are rounded towards zero.
9016 Dynamic rounding mode. A field in the floating point control register
9017 (@var{fpcr}, see Alpha architecture reference manual) controls the
9018 rounding mode in effect. The C library initializes this register for
9019 rounding towards plus infinity. Thus, unless your program modifies the
9020 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9023 @item -mtrap-precision=@var{trap-precision}
9024 @opindex mtrap-precision
9025 In the Alpha architecture, floating point traps are imprecise. This
9026 means without software assistance it is impossible to recover from a
9027 floating trap and program execution normally needs to be terminated.
9028 GCC can generate code that can assist operating system trap handlers
9029 in determining the exact location that caused a floating point trap.
9030 Depending on the requirements of an application, different levels of
9031 precisions can be selected:
9035 Program precision. This option is the default and means a trap handler
9036 can only identify which program caused a floating point exception.
9039 Function precision. The trap handler can determine the function that
9040 caused a floating point exception.
9043 Instruction precision. The trap handler can determine the exact
9044 instruction that caused a floating point exception.
9047 Other Alpha compilers provide the equivalent options called
9048 @option{-scope_safe} and @option{-resumption_safe}.
9050 @item -mieee-conformant
9051 @opindex mieee-conformant
9052 This option marks the generated code as IEEE conformant. You must not
9053 use this option unless you also specify @option{-mtrap-precision=i} and either
9054 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9055 is to emit the line @samp{.eflag 48} in the function prologue of the
9056 generated assembly file. Under DEC Unix, this has the effect that
9057 IEEE-conformant math library routines will be linked in.
9059 @item -mbuild-constants
9060 @opindex mbuild-constants
9061 Normally GCC examines a 32- or 64-bit integer constant to
9062 see if it can construct it from smaller constants in two or three
9063 instructions. If it cannot, it will output the constant as a literal and
9064 generate code to load it from the data segment at runtime.
9066 Use this option to require GCC to construct @emph{all} integer constants
9067 using code, even if it takes more instructions (the maximum is six).
9069 You would typically use this option to build a shared library dynamic
9070 loader. Itself a shared library, it must relocate itself in memory
9071 before it can find the variables and constants in its own data segment.
9077 Select whether to generate code to be assembled by the vendor-supplied
9078 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9096 Indicate whether GCC should generate code to use the optional BWX,
9097 CIX, FIX and MAX instruction sets. The default is to use the instruction
9098 sets supported by the CPU type specified via @option{-mcpu=} option or that
9099 of the CPU on which GCC was built if none was specified.
9104 @opindex mfloat-ieee
9105 Generate code that uses (does not use) VAX F and G floating point
9106 arithmetic instead of IEEE single and double precision.
9108 @item -mexplicit-relocs
9109 @itemx -mno-explicit-relocs
9110 @opindex mexplicit-relocs
9111 @opindex mno-explicit-relocs
9112 Older Alpha assemblers provided no way to generate symbol relocations
9113 except via assembler macros. Use of these macros does not allow
9114 optimal instruction scheduling. GNU binutils as of version 2.12
9115 supports a new syntax that allows the compiler to explicitly mark
9116 which relocations should apply to which instructions. This option
9117 is mostly useful for debugging, as GCC detects the capabilities of
9118 the assembler when it is built and sets the default accordingly.
9122 @opindex msmall-data
9123 @opindex mlarge-data
9124 When @option{-mexplicit-relocs} is in effect, static data is
9125 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9126 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9127 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9128 16-bit relocations off of the @code{$gp} register. This limits the
9129 size of the small data area to 64KB, but allows the variables to be
9130 directly accessed via a single instruction.
9132 The default is @option{-mlarge-data}. With this option the data area
9133 is limited to just below 2GB@. Programs that require more than 2GB of
9134 data must use @code{malloc} or @code{mmap} to allocate the data in the
9135 heap instead of in the program's data segment.
9137 When generating code for shared libraries, @option{-fpic} implies
9138 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9142 @opindex msmall-text
9143 @opindex mlarge-text
9144 When @option{-msmall-text} is used, the compiler assumes that the
9145 code of the entire program (or shared library) fits in 4MB, and is
9146 thus reachable with a branch instruction. When @option{-msmall-data}
9147 is used, the compiler can assume that all local symbols share the
9148 same @code{$gp} value, and thus reduce the number of instructions
9149 required for a function call from 4 to 1.
9151 The default is @option{-mlarge-text}.
9153 @item -mcpu=@var{cpu_type}
9155 Set the instruction set and instruction scheduling parameters for
9156 machine type @var{cpu_type}. You can specify either the @samp{EV}
9157 style name or the corresponding chip number. GCC supports scheduling
9158 parameters for the EV4, EV5 and EV6 family of processors and will
9159 choose the default values for the instruction set from the processor
9160 you specify. If you do not specify a processor type, GCC will default
9161 to the processor on which the compiler was built.
9163 Supported values for @var{cpu_type} are
9169 Schedules as an EV4 and has no instruction set extensions.
9173 Schedules as an EV5 and has no instruction set extensions.
9177 Schedules as an EV5 and supports the BWX extension.
9182 Schedules as an EV5 and supports the BWX and MAX extensions.
9186 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9190 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9193 @item -mtune=@var{cpu_type}
9195 Set only the instruction scheduling parameters for machine type
9196 @var{cpu_type}. The instruction set is not changed.
9198 @item -mmemory-latency=@var{time}
9199 @opindex mmemory-latency
9200 Sets the latency the scheduler should assume for typical memory
9201 references as seen by the application. This number is highly
9202 dependent on the memory access patterns used by the application
9203 and the size of the external cache on the machine.
9205 Valid options for @var{time} are
9209 A decimal number representing clock cycles.
9215 The compiler contains estimates of the number of clock cycles for
9216 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9217 (also called Dcache, Scache, and Bcache), as well as to main memory.
9218 Note that L3 is only valid for EV5.
9223 @node DEC Alpha/VMS Options
9224 @subsection DEC Alpha/VMS Options
9226 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9229 @item -mvms-return-codes
9230 @opindex mvms-return-codes
9231 Return VMS condition codes from main. The default is to return POSIX
9232 style condition (e.g.@ error) codes.
9236 @subsection FRV Options
9243 Only use the first 32 general purpose registers.
9248 Use all 64 general purpose registers.
9253 Use only the first 32 floating point registers.
9258 Use all 64 floating point registers
9261 @opindex mhard-float
9263 Use hardware instructions for floating point operations.
9266 @opindex msoft-float
9268 Use library routines for floating point operations.
9273 Dynamically allocate condition code registers.
9278 Do not try to dynamically allocate condition code registers, only
9279 use @code{icc0} and @code{fcc0}.
9284 Change ABI to use double word insns.
9289 Do not use double word instructions.
9294 Use floating point double instructions.
9299 Do not use floating point double instructions.
9304 Use media instructions.
9309 Do not use media instructions.
9314 Use multiply and add/subtract instructions.
9319 Do not use multiply and add/subtract instructions.
9324 Select the FDPIC ABI, that uses function descriptors to represent
9325 pointers to functions. Without any PIC/PIE-related options, it
9326 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9327 assumes GOT entries and small data are within a 12-bit range from the
9328 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9329 are computed with 32 bits.
9332 @opindex minline-plt
9334 Enable inlining of PLT entries in function calls to functions that are
9335 not known to bind locally. It has no effect without @option{-mfdpic}.
9336 It's enabled by default if optimizing for speed and compiling for
9337 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9338 optimization option such as @option{-O3} or above is present in the
9344 Assume a large TLS segment when generating thread-local code.
9349 Do not assume a large TLS segment when generating thread-local code.
9354 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9355 that is known to be in read-only sections. It's enabled by default,
9356 except for @option{-fpic} or @option{-fpie}: even though it may help
9357 make the global offset table smaller, it trades 1 instruction for 4.
9358 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9359 one of which may be shared by multiple symbols, and it avoids the need
9360 for a GOT entry for the referenced symbol, so it's more likely to be a
9361 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9363 @item -multilib-library-pic
9364 @opindex multilib-library-pic
9366 Link with the (library, not FD) pic libraries. It's implied by
9367 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9368 @option{-fpic} without @option{-mfdpic}. You should never have to use
9374 Follow the EABI requirement of always creating a frame pointer whenever
9375 a stack frame is allocated. This option is enabled by default and can
9376 be disabled with @option{-mno-linked-fp}.
9379 @opindex mlong-calls
9381 Use indirect addressing to call functions outside the current
9382 compilation unit. This allows the functions to be placed anywhere
9383 within the 32-bit address space.
9385 @item -malign-labels
9386 @opindex malign-labels
9388 Try to align labels to an 8-byte boundary by inserting nops into the
9389 previous packet. This option only has an effect when VLIW packing
9390 is enabled. It doesn't create new packets; it merely adds nops to
9394 @opindex mlibrary-pic
9396 Generate position-independent EABI code.
9401 Use only the first four media accumulator registers.
9406 Use all eight media accumulator registers.
9411 Pack VLIW instructions.
9416 Do not pack VLIW instructions.
9421 Do not mark ABI switches in e_flags.
9426 Enable the use of conditional-move instructions (default).
9428 This switch is mainly for debugging the compiler and will likely be removed
9429 in a future version.
9431 @item -mno-cond-move
9432 @opindex mno-cond-move
9434 Disable the use of conditional-move instructions.
9436 This switch is mainly for debugging the compiler and will likely be removed
9437 in a future version.
9442 Enable the use of conditional set instructions (default).
9444 This switch is mainly for debugging the compiler and will likely be removed
9445 in a future version.
9450 Disable the use of conditional set instructions.
9452 This switch is mainly for debugging the compiler and will likely be removed
9453 in a future version.
9458 Enable the use of conditional execution (default).
9460 This switch is mainly for debugging the compiler and will likely be removed
9461 in a future version.
9463 @item -mno-cond-exec
9464 @opindex mno-cond-exec
9466 Disable the use of conditional execution.
9468 This switch is mainly for debugging the compiler and will likely be removed
9469 in a future version.
9472 @opindex mvliw-branch
9474 Run a pass to pack branches into VLIW instructions (default).
9476 This switch is mainly for debugging the compiler and will likely be removed
9477 in a future version.
9479 @item -mno-vliw-branch
9480 @opindex mno-vliw-branch
9482 Do not run a pass to pack branches into VLIW instructions.
9484 This switch is mainly for debugging the compiler and will likely be removed
9485 in a future version.
9487 @item -mmulti-cond-exec
9488 @opindex mmulti-cond-exec
9490 Enable optimization of @code{&&} and @code{||} in conditional execution
9493 This switch is mainly for debugging the compiler and will likely be removed
9494 in a future version.
9496 @item -mno-multi-cond-exec
9497 @opindex mno-multi-cond-exec
9499 Disable optimization of @code{&&} and @code{||} in conditional execution.
9501 This switch is mainly for debugging the compiler and will likely be removed
9502 in a future version.
9504 @item -mnested-cond-exec
9505 @opindex mnested-cond-exec
9507 Enable nested conditional execution optimizations (default).
9509 This switch is mainly for debugging the compiler and will likely be removed
9510 in a future version.
9512 @item -mno-nested-cond-exec
9513 @opindex mno-nested-cond-exec
9515 Disable nested conditional execution optimizations.
9517 This switch is mainly for debugging the compiler and will likely be removed
9518 in a future version.
9520 @item -moptimize-membar
9521 @opindex moptimize-membar
9523 This switch removes redundant @code{membar} instructions from the
9524 compiler generated code. It is enabled by default.
9526 @item -mno-optimize-membar
9527 @opindex mno-optimize-membar
9529 This switch disables the automatic removal of redundant @code{membar}
9530 instructions from the generated code.
9532 @item -mtomcat-stats
9533 @opindex mtomcat-stats
9535 Cause gas to print out tomcat statistics.
9537 @item -mcpu=@var{cpu}
9540 Select the processor type for which to generate code. Possible values are
9541 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9542 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9546 @node GNU/Linux Options
9547 @subsection GNU/Linux Options
9549 These @samp{-m} options are defined for GNU/Linux targets:
9554 Use the GNU C library instead of uClibc. This is the default except
9555 on @samp{*-*-linux-*uclibc*} targets.
9559 Use uClibc instead of the GNU C library. This is the default on
9560 @samp{*-*-linux-*uclibc*} targets.
9563 @node H8/300 Options
9564 @subsection H8/300 Options
9566 These @samp{-m} options are defined for the H8/300 implementations:
9571 Shorten some address references at link time, when possible; uses the
9572 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9573 ld, Using ld}, for a fuller description.
9577 Generate code for the H8/300H@.
9581 Generate code for the H8S@.
9585 Generate code for the H8S and H8/300H in the normal mode. This switch
9586 must be used either with @option{-mh} or @option{-ms}.
9590 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9594 Make @code{int} data 32 bits by default.
9598 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9599 The default for the H8/300H and H8S is to align longs and floats on 4
9601 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9602 This option has no effect on the H8/300.
9606 @subsection HPPA Options
9607 @cindex HPPA Options
9609 These @samp{-m} options are defined for the HPPA family of computers:
9612 @item -march=@var{architecture-type}
9614 Generate code for the specified architecture. The choices for
9615 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9616 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9617 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9618 architecture option for your machine. Code compiled for lower numbered
9619 architectures will run on higher numbered architectures, but not the
9623 @itemx -mpa-risc-1-1
9624 @itemx -mpa-risc-2-0
9625 @opindex mpa-risc-1-0
9626 @opindex mpa-risc-1-1
9627 @opindex mpa-risc-2-0
9628 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9631 @opindex mbig-switch
9632 Generate code suitable for big switch tables. Use this option only if
9633 the assembler/linker complain about out of range branches within a switch
9636 @item -mjump-in-delay
9637 @opindex mjump-in-delay
9638 Fill delay slots of function calls with unconditional jump instructions
9639 by modifying the return pointer for the function call to be the target
9640 of the conditional jump.
9642 @item -mdisable-fpregs
9643 @opindex mdisable-fpregs
9644 Prevent floating point registers from being used in any manner. This is
9645 necessary for compiling kernels which perform lazy context switching of
9646 floating point registers. If you use this option and attempt to perform
9647 floating point operations, the compiler will abort.
9649 @item -mdisable-indexing
9650 @opindex mdisable-indexing
9651 Prevent the compiler from using indexing address modes. This avoids some
9652 rather obscure problems when compiling MIG generated code under MACH@.
9654 @item -mno-space-regs
9655 @opindex mno-space-regs
9656 Generate code that assumes the target has no space registers. This allows
9657 GCC to generate faster indirect calls and use unscaled index address modes.
9659 Such code is suitable for level 0 PA systems and kernels.
9661 @item -mfast-indirect-calls
9662 @opindex mfast-indirect-calls
9663 Generate code that assumes calls never cross space boundaries. This
9664 allows GCC to emit code which performs faster indirect calls.
9666 This option will not work in the presence of shared libraries or nested
9669 @item -mfixed-range=@var{register-range}
9670 @opindex mfixed-range
9671 Generate code treating the given register range as fixed registers.
9672 A fixed register is one that the register allocator can not use. This is
9673 useful when compiling kernel code. A register range is specified as
9674 two registers separated by a dash. Multiple register ranges can be
9675 specified separated by a comma.
9677 @item -mlong-load-store
9678 @opindex mlong-load-store
9679 Generate 3-instruction load and store sequences as sometimes required by
9680 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9683 @item -mportable-runtime
9684 @opindex mportable-runtime
9685 Use the portable calling conventions proposed by HP for ELF systems.
9689 Enable the use of assembler directives only GAS understands.
9691 @item -mschedule=@var{cpu-type}
9693 Schedule code according to the constraints for the machine type
9694 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9695 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9696 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9697 proper scheduling option for your machine. The default scheduling is
9701 @opindex mlinker-opt
9702 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9703 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9704 linkers in which they give bogus error messages when linking some programs.
9707 @opindex msoft-float
9708 Generate output containing library calls for floating point.
9709 @strong{Warning:} the requisite libraries are not available for all HPPA
9710 targets. Normally the facilities of the machine's usual C compiler are
9711 used, but this cannot be done directly in cross-compilation. You must make
9712 your own arrangements to provide suitable library functions for
9713 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9714 does provide software floating point support.
9716 @option{-msoft-float} changes the calling convention in the output file;
9717 therefore, it is only useful if you compile @emph{all} of a program with
9718 this option. In particular, you need to compile @file{libgcc.a}, the
9719 library that comes with GCC, with @option{-msoft-float} in order for
9724 Generate the predefine, @code{_SIO}, for server IO@. The default is
9725 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9726 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9727 options are available under HP-UX and HI-UX@.
9731 Use GNU ld specific options. This passes @option{-shared} to ld when
9732 building a shared library. It is the default when GCC is configured,
9733 explicitly or implicitly, with the GNU linker. This option does not
9734 have any affect on which ld is called, it only changes what parameters
9735 are passed to that ld. The ld that is called is determined by the
9736 @option{--with-ld} configure option, GCC's program search path, and
9737 finally by the user's @env{PATH}. The linker used by GCC can be printed
9738 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9739 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9743 Use HP ld specific options. This passes @option{-b} to ld when building
9744 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9745 links. It is the default when GCC is configured, explicitly or
9746 implicitly, with the HP linker. This option does not have any affect on
9747 which ld is called, it only changes what parameters are passed to that
9748 ld. The ld that is called is determined by the @option{--with-ld}
9749 configure option, GCC's program search path, and finally by the user's
9750 @env{PATH}. The linker used by GCC can be printed using @samp{which
9751 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9752 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9755 @opindex mno-long-calls
9756 Generate code that uses long call sequences. This ensures that a call
9757 is always able to reach linker generated stubs. The default is to generate
9758 long calls only when the distance from the call site to the beginning
9759 of the function or translation unit, as the case may be, exceeds a
9760 predefined limit set by the branch type being used. The limits for
9761 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9762 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9765 Distances are measured from the beginning of functions when using the
9766 @option{-ffunction-sections} option, or when using the @option{-mgas}
9767 and @option{-mno-portable-runtime} options together under HP-UX with
9770 It is normally not desirable to use this option as it will degrade
9771 performance. However, it may be useful in large applications,
9772 particularly when partial linking is used to build the application.
9774 The types of long calls used depends on the capabilities of the
9775 assembler and linker, and the type of code being generated. The
9776 impact on systems that support long absolute calls, and long pic
9777 symbol-difference or pc-relative calls should be relatively small.
9778 However, an indirect call is used on 32-bit ELF systems in pic code
9779 and it is quite long.
9781 @item -munix=@var{unix-std}
9783 Generate compiler predefines and select a startfile for the specified
9784 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9785 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9786 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9787 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9788 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9791 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9792 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9793 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9794 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9795 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9796 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9798 It is @emph{important} to note that this option changes the interfaces
9799 for various library routines. It also affects the operational behavior
9800 of the C library. Thus, @emph{extreme} care is needed in using this
9803 Library code that is intended to operate with more than one UNIX
9804 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9805 as appropriate. Most GNU software doesn't provide this capability.
9809 Suppress the generation of link options to search libdld.sl when the
9810 @option{-static} option is specified on HP-UX 10 and later.
9814 The HP-UX implementation of setlocale in libc has a dependency on
9815 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9816 when the @option{-static} option is specified, special link options
9817 are needed to resolve this dependency.
9819 On HP-UX 10 and later, the GCC driver adds the necessary options to
9820 link with libdld.sl when the @option{-static} option is specified.
9821 This causes the resulting binary to be dynamic. On the 64-bit port,
9822 the linkers generate dynamic binaries by default in any case. The
9823 @option{-nolibdld} option can be used to prevent the GCC driver from
9824 adding these link options.
9828 Add support for multithreading with the @dfn{dce thread} library
9829 under HP-UX@. This option sets flags for both the preprocessor and
9833 @node i386 and x86-64 Options
9834 @subsection Intel 386 and AMD x86-64 Options
9835 @cindex i386 Options
9836 @cindex x86-64 Options
9837 @cindex Intel 386 Options
9838 @cindex AMD x86-64 Options
9840 These @samp{-m} options are defined for the i386 and x86-64 family of
9844 @item -mtune=@var{cpu-type}
9846 Tune to @var{cpu-type} everything applicable about the generated code, except
9847 for the ABI and the set of available instructions. The choices for
9851 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9852 If you know the CPU on which your code will run, then you should use
9853 the corresponding @option{-mtune} option instead of
9854 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9855 of your application will have, then you should use this option.
9857 As new processors are deployed in the marketplace, the behavior of this
9858 option will change. Therefore, if you upgrade to a newer version of
9859 GCC, the code generated option will change to reflect the processors
9860 that were most common when that version of GCC was released.
9862 There is no @option{-march=generic} option because @option{-march}
9863 indicates the instruction set the compiler can use, and there is no
9864 generic instruction set applicable to all processors. In contrast,
9865 @option{-mtune} indicates the processor (or, in this case, collection of
9866 processors) for which the code is optimized.
9868 This selects the CPU to tune for at compilation time by determining
9869 the processor type of the compiling machine. Using @option{-mtune=native}
9870 will produce code optimized for the local machine under the constraints
9871 of the selected instruction set. Using @option{-march=native} will
9872 enable all instruction subsets supported by the local machine (hence
9873 the result might not run on different machines).
9875 Original Intel's i386 CPU@.
9877 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9879 Intel Pentium CPU with no MMX support.
9881 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9883 Intel PentiumPro CPU@.
9885 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9886 instruction set will be used, so the code will run on all i686 family chips.
9888 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9889 @item pentium3, pentium3m
9890 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9893 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9894 support. Used by Centrino notebooks.
9895 @item pentium4, pentium4m
9896 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9898 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9901 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9902 SSE2 and SSE3 instruction set support.
9904 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9905 instruction set support.
9907 AMD K6 CPU with MMX instruction set support.
9909 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9910 @item athlon, athlon-tbird
9911 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9913 @item athlon-4, athlon-xp, athlon-mp
9914 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9915 instruction set support.
9916 @item k8, opteron, athlon64, athlon-fx
9917 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9918 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9919 @item k8-sse3, opteron-sse3, athlon64-sse3
9920 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
9921 @item amdfam10, barcelona
9922 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
9923 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9924 instruction set extensions.)
9926 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9929 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9930 instruction set support.
9932 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9933 implemented for this chip.)
9935 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9936 implemented for this chip.)
9938 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9941 While picking a specific @var{cpu-type} will schedule things appropriately
9942 for that particular chip, the compiler will not generate any code that
9943 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9946 @item -march=@var{cpu-type}
9948 Generate instructions for the machine type @var{cpu-type}. The choices
9949 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9950 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9952 @item -mcpu=@var{cpu-type}
9954 A deprecated synonym for @option{-mtune}.
9956 @item -mfpmath=@var{unit}
9958 Generate floating point arithmetics for selected unit @var{unit}. The choices
9963 Use the standard 387 floating point coprocessor present majority of chips and
9964 emulated otherwise. Code compiled with this option will run almost everywhere.
9965 The temporary results are computed in 80bit precision instead of precision
9966 specified by the type resulting in slightly different results compared to most
9967 of other chips. See @option{-ffloat-store} for more detailed description.
9969 This is the default choice for i386 compiler.
9972 Use scalar floating point instructions present in the SSE instruction set.
9973 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9974 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9975 instruction set supports only single precision arithmetics, thus the double and
9976 extended precision arithmetics is still done using 387. Later version, present
9977 only in Pentium4 and the future AMD x86-64 chips supports double precision
9980 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9981 or @option{-msse2} switches to enable SSE extensions and make this option
9982 effective. For the x86-64 compiler, these extensions are enabled by default.
9984 The resulting code should be considerably faster in the majority of cases and avoid
9985 the numerical instability problems of 387 code, but may break some existing
9986 code that expects temporaries to be 80bit.
9988 This is the default choice for the x86-64 compiler.
9991 Attempt to utilize both instruction sets at once. This effectively double the
9992 amount of available registers and on chips with separate execution units for
9993 387 and SSE the execution resources too. Use this option with care, as it is
9994 still experimental, because the GCC register allocator does not model separate
9995 functional units well resulting in instable performance.
9998 @item -masm=@var{dialect}
9999 @opindex masm=@var{dialect}
10000 Output asm instructions using selected @var{dialect}. Supported
10001 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10002 not support @samp{intel}.
10005 @itemx -mno-ieee-fp
10007 @opindex mno-ieee-fp
10008 Control whether or not the compiler uses IEEE floating point
10009 comparisons. These handle correctly the case where the result of a
10010 comparison is unordered.
10013 @opindex msoft-float
10014 Generate output containing library calls for floating point.
10015 @strong{Warning:} the requisite libraries are not part of GCC@.
10016 Normally the facilities of the machine's usual C compiler are used, but
10017 this can't be done directly in cross-compilation. You must make your
10018 own arrangements to provide suitable library functions for
10021 On machines where a function returns floating point results in the 80387
10022 register stack, some floating point opcodes may be emitted even if
10023 @option{-msoft-float} is used.
10025 @item -mno-fp-ret-in-387
10026 @opindex mno-fp-ret-in-387
10027 Do not use the FPU registers for return values of functions.
10029 The usual calling convention has functions return values of types
10030 @code{float} and @code{double} in an FPU register, even if there
10031 is no FPU@. The idea is that the operating system should emulate
10034 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10035 in ordinary CPU registers instead.
10037 @item -mno-fancy-math-387
10038 @opindex mno-fancy-math-387
10039 Some 387 emulators do not support the @code{sin}, @code{cos} and
10040 @code{sqrt} instructions for the 387. Specify this option to avoid
10041 generating those instructions. This option is the default on FreeBSD,
10042 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10043 indicates that the target cpu will always have an FPU and so the
10044 instruction will not need emulation. As of revision 2.6.1, these
10045 instructions are not generated unless you also use the
10046 @option{-funsafe-math-optimizations} switch.
10048 @item -malign-double
10049 @itemx -mno-align-double
10050 @opindex malign-double
10051 @opindex mno-align-double
10052 Control whether GCC aligns @code{double}, @code{long double}, and
10053 @code{long long} variables on a two word boundary or a one word
10054 boundary. Aligning @code{double} variables on a two word boundary will
10055 produce code that runs somewhat faster on a @samp{Pentium} at the
10056 expense of more memory.
10058 On x86-64, @option{-malign-double} is enabled by default.
10060 @strong{Warning:} if you use the @option{-malign-double} switch,
10061 structures containing the above types will be aligned differently than
10062 the published application binary interface specifications for the 386
10063 and will not be binary compatible with structures in code compiled
10064 without that switch.
10066 @item -m96bit-long-double
10067 @itemx -m128bit-long-double
10068 @opindex m96bit-long-double
10069 @opindex m128bit-long-double
10070 These switches control the size of @code{long double} type. The i386
10071 application binary interface specifies the size to be 96 bits,
10072 so @option{-m96bit-long-double} is the default in 32 bit mode.
10074 Modern architectures (Pentium and newer) would prefer @code{long double}
10075 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10076 conforming to the ABI, this would not be possible. So specifying a
10077 @option{-m128bit-long-double} will align @code{long double}
10078 to a 16 byte boundary by padding the @code{long double} with an additional
10081 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10082 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10084 Notice that neither of these options enable any extra precision over the x87
10085 standard of 80 bits for a @code{long double}.
10087 @strong{Warning:} if you override the default value for your target ABI, the
10088 structures and arrays containing @code{long double} variables will change
10089 their size as well as function calling convention for function taking
10090 @code{long double} will be modified. Hence they will not be binary
10091 compatible with arrays or structures in code compiled without that switch.
10093 @item -mmlarge-data-threshold=@var{number}
10094 @opindex mlarge-data-threshold=@var{number}
10095 When @option{-mcmodel=medium} is specified, the data greater than
10096 @var{threshold} are placed in large data section. This value must be the
10097 same across all object linked into the binary and defaults to 65535.
10101 Use a different function-calling convention, in which functions that
10102 take a fixed number of arguments return with the @code{ret} @var{num}
10103 instruction, which pops their arguments while returning. This saves one
10104 instruction in the caller since there is no need to pop the arguments
10107 You can specify that an individual function is called with this calling
10108 sequence with the function attribute @samp{stdcall}. You can also
10109 override the @option{-mrtd} option by using the function attribute
10110 @samp{cdecl}. @xref{Function Attributes}.
10112 @strong{Warning:} this calling convention is incompatible with the one
10113 normally used on Unix, so you cannot use it if you need to call
10114 libraries compiled with the Unix compiler.
10116 Also, you must provide function prototypes for all functions that
10117 take variable numbers of arguments (including @code{printf});
10118 otherwise incorrect code will be generated for calls to those
10121 In addition, seriously incorrect code will result if you call a
10122 function with too many arguments. (Normally, extra arguments are
10123 harmlessly ignored.)
10125 @item -mregparm=@var{num}
10127 Control how many registers are used to pass integer arguments. By
10128 default, no registers are used to pass arguments, and at most 3
10129 registers can be used. You can control this behavior for a specific
10130 function by using the function attribute @samp{regparm}.
10131 @xref{Function Attributes}.
10133 @strong{Warning:} if you use this switch, and
10134 @var{num} is nonzero, then you must build all modules with the same
10135 value, including any libraries. This includes the system libraries and
10139 @opindex msseregparm
10140 Use SSE register passing conventions for float and double arguments
10141 and return values. You can control this behavior for a specific
10142 function by using the function attribute @samp{sseregparm}.
10143 @xref{Function Attributes}.
10145 @strong{Warning:} if you use this switch then you must build all
10146 modules with the same value, including any libraries. This includes
10147 the system libraries and startup modules.
10156 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10157 is specified, the significands of results of floating-point operations are
10158 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10159 significands of results of floating-point operations to 53 bits (double
10160 precision) and @option{-mpc80} rounds the significands of results of
10161 floating-point operations to 64 bits (extended double precision), which is
10162 the default. When this option is used, floating-point operations in higher
10163 precisions are not available to the programmer without setting the FPU
10164 control word explicitly.
10166 Setting the rounding of floating-point operations to less than the default
10167 80 bits can speed some programs by 2% or more. Note that some mathematical
10168 libraries assume that extended precision (80 bit) floating-point operations
10169 are enabled by default; routines in such libraries could suffer significant
10170 loss of accuracy, typically through so-called "catastrophic cancellation",
10171 when this option is used to set the precision to less than extended precision.
10173 @item -mstackrealign
10174 @opindex mstackrealign
10175 Realign the stack at entry. On the Intel x86, the
10176 @option{-mstackrealign} option will generate an alternate prologue and
10177 epilogue that realigns the runtime stack. This supports mixing legacy
10178 codes that keep a 4-byte aligned stack with modern codes that keep a
10179 16-byte stack for SSE compatibility. The alternate prologue and
10180 epilogue are slower and bigger than the regular ones, and the
10181 alternate prologue requires an extra scratch register; this lowers the
10182 number of registers available if used in conjunction with the
10183 @code{regparm} attribute. The @option{-mstackrealign} option is
10184 incompatible with the nested function prologue; this is considered a
10185 hard error. See also the attribute @code{force_align_arg_pointer},
10186 applicable to individual functions.
10188 @item -mpreferred-stack-boundary=@var{num}
10189 @opindex mpreferred-stack-boundary
10190 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10191 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10192 the default is 4 (16 bytes or 128 bits).
10194 On Pentium and PentiumPro, @code{double} and @code{long double} values
10195 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10196 suffer significant run time performance penalties. On Pentium III, the
10197 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10198 properly if it is not 16 byte aligned.
10200 To ensure proper alignment of this values on the stack, the stack boundary
10201 must be as aligned as that required by any value stored on the stack.
10202 Further, every function must be generated such that it keeps the stack
10203 aligned. Thus calling a function compiled with a higher preferred
10204 stack boundary from a function compiled with a lower preferred stack
10205 boundary will most likely misalign the stack. It is recommended that
10206 libraries that use callbacks always use the default setting.
10208 This extra alignment does consume extra stack space, and generally
10209 increases code size. Code that is sensitive to stack space usage, such
10210 as embedded systems and operating system kernels, may want to reduce the
10211 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10237 These switches enable or disable the use of instructions in the MMX,
10238 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM or 3DNow! extended instruction sets.
10239 These extensions are also available as built-in functions: see
10240 @ref{X86 Built-in Functions}, for details of the functions enabled and
10241 disabled by these switches.
10243 To have SSE/SSE2 instructions generated automatically from floating-point
10244 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10246 These options will enable GCC to use these extended instructions in
10247 generated code, even without @option{-mfpmath=sse}. Applications which
10248 perform runtime CPU detection must compile separate files for each
10249 supported architecture, using the appropriate flags. In particular,
10250 the file containing the CPU detection code should be compiled without
10255 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10256 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10257 data types. This is useful for high resolution counters that could be updated
10258 by multiple processors (or cores). This instruction is generated as part of
10259 atomic built-in functions: see @ref{Atomic Builtins} for details.
10263 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10264 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10265 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10266 SAHF are load and store instructions, respectively, for certain status flags.
10267 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10268 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10271 @itemx -mno-push-args
10272 @opindex mpush-args
10273 @opindex mno-push-args
10274 Use PUSH operations to store outgoing parameters. This method is shorter
10275 and usually equally fast as method using SUB/MOV operations and is enabled
10276 by default. In some cases disabling it may improve performance because of
10277 improved scheduling and reduced dependencies.
10279 @item -maccumulate-outgoing-args
10280 @opindex maccumulate-outgoing-args
10281 If enabled, the maximum amount of space required for outgoing arguments will be
10282 computed in the function prologue. This is faster on most modern CPUs
10283 because of reduced dependencies, improved scheduling and reduced stack usage
10284 when preferred stack boundary is not equal to 2. The drawback is a notable
10285 increase in code size. This switch implies @option{-mno-push-args}.
10289 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10290 on thread-safe exception handling must compile and link all code with the
10291 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10292 @option{-D_MT}; when linking, it links in a special thread helper library
10293 @option{-lmingwthrd} which cleans up per thread exception handling data.
10295 @item -mno-align-stringops
10296 @opindex mno-align-stringops
10297 Do not align destination of inlined string operations. This switch reduces
10298 code size and improves performance in case the destination is already aligned,
10299 but GCC doesn't know about it.
10301 @item -minline-all-stringops
10302 @opindex minline-all-stringops
10303 By default GCC inlines string operations only when destination is known to be
10304 aligned at least to 4 byte boundary. This enables more inlining, increase code
10305 size, but may improve performance of code that depends on fast memcpy, strlen
10306 and memset for short lengths.
10308 @item -minline-stringops-dynamically
10309 @opindex minline-stringops-dynamically
10310 For string operation of unknown size, inline runtime checks so for small
10311 blocks inline code is used, while for large blocks library call is used.
10313 @item -mstringop-strategy=@var{alg}
10314 @opindex mstringop-strategy=@var{alg}
10315 Overwrite internal decision heuristic about particular algorithm to inline
10316 string operation with. The allowed values are @code{rep_byte},
10317 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10318 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10319 expanding inline loop, @code{libcall} for always expanding library call.
10321 @item -momit-leaf-frame-pointer
10322 @opindex momit-leaf-frame-pointer
10323 Don't keep the frame pointer in a register for leaf functions. This
10324 avoids the instructions to save, set up and restore frame pointers and
10325 makes an extra register available in leaf functions. The option
10326 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10327 which might make debugging harder.
10329 @item -mtls-direct-seg-refs
10330 @itemx -mno-tls-direct-seg-refs
10331 @opindex mtls-direct-seg-refs
10332 Controls whether TLS variables may be accessed with offsets from the
10333 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10334 or whether the thread base pointer must be added. Whether or not this
10335 is legal depends on the operating system, and whether it maps the
10336 segment to cover the entire TLS area.
10338 For systems that use GNU libc, the default is on.
10341 These @samp{-m} switches are supported in addition to the above
10342 on AMD x86-64 processors in 64-bit environments.
10349 Generate code for a 32-bit or 64-bit environment.
10350 The 32-bit environment sets int, long and pointer to 32 bits and
10351 generates code that runs on any i386 system.
10352 The 64-bit environment sets int to 32 bits and long and pointer
10353 to 64 bits and generates code for AMD's x86-64 architecture. For
10354 darwin only the -m64 option turns off the @option{-fno-pic} and
10355 @option{-mdynamic-no-pic} options.
10357 @item -mno-red-zone
10358 @opindex no-red-zone
10359 Do not use a so called red zone for x86-64 code. The red zone is mandated
10360 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10361 stack pointer that will not be modified by signal or interrupt handlers
10362 and therefore can be used for temporary data without adjusting the stack
10363 pointer. The flag @option{-mno-red-zone} disables this red zone.
10365 @item -mcmodel=small
10366 @opindex mcmodel=small
10367 Generate code for the small code model: the program and its symbols must
10368 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10369 Programs can be statically or dynamically linked. This is the default
10372 @item -mcmodel=kernel
10373 @opindex mcmodel=kernel
10374 Generate code for the kernel code model. The kernel runs in the
10375 negative 2 GB of the address space.
10376 This model has to be used for Linux kernel code.
10378 @item -mcmodel=medium
10379 @opindex mcmodel=medium
10380 Generate code for the medium model: The program is linked in the lower 2
10381 GB of the address space but symbols can be located anywhere in the
10382 address space. Programs can be statically or dynamically linked, but
10383 building of shared libraries are not supported with the medium model.
10385 @item -mcmodel=large
10386 @opindex mcmodel=large
10387 Generate code for the large model: This model makes no assumptions
10388 about addresses and sizes of sections.
10391 @node IA-64 Options
10392 @subsection IA-64 Options
10393 @cindex IA-64 Options
10395 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10399 @opindex mbig-endian
10400 Generate code for a big endian target. This is the default for HP-UX@.
10402 @item -mlittle-endian
10403 @opindex mlittle-endian
10404 Generate code for a little endian target. This is the default for AIX5
10410 @opindex mno-gnu-as
10411 Generate (or don't) code for the GNU assembler. This is the default.
10412 @c Also, this is the default if the configure option @option{--with-gnu-as}
10418 @opindex mno-gnu-ld
10419 Generate (or don't) code for the GNU linker. This is the default.
10420 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10425 Generate code that does not use a global pointer register. The result
10426 is not position independent code, and violates the IA-64 ABI@.
10428 @item -mvolatile-asm-stop
10429 @itemx -mno-volatile-asm-stop
10430 @opindex mvolatile-asm-stop
10431 @opindex mno-volatile-asm-stop
10432 Generate (or don't) a stop bit immediately before and after volatile asm
10435 @item -mregister-names
10436 @itemx -mno-register-names
10437 @opindex mregister-names
10438 @opindex mno-register-names
10439 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10440 the stacked registers. This may make assembler output more readable.
10446 Disable (or enable) optimizations that use the small data section. This may
10447 be useful for working around optimizer bugs.
10449 @item -mconstant-gp
10450 @opindex mconstant-gp
10451 Generate code that uses a single constant global pointer value. This is
10452 useful when compiling kernel code.
10456 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10457 This is useful when compiling firmware code.
10459 @item -minline-float-divide-min-latency
10460 @opindex minline-float-divide-min-latency
10461 Generate code for inline divides of floating point values
10462 using the minimum latency algorithm.
10464 @item -minline-float-divide-max-throughput
10465 @opindex minline-float-divide-max-throughput
10466 Generate code for inline divides of floating point values
10467 using the maximum throughput algorithm.
10469 @item -minline-int-divide-min-latency
10470 @opindex minline-int-divide-min-latency
10471 Generate code for inline divides of integer values
10472 using the minimum latency algorithm.
10474 @item -minline-int-divide-max-throughput
10475 @opindex minline-int-divide-max-throughput
10476 Generate code for inline divides of integer values
10477 using the maximum throughput algorithm.
10479 @item -minline-sqrt-min-latency
10480 @opindex minline-sqrt-min-latency
10481 Generate code for inline square roots
10482 using the minimum latency algorithm.
10484 @item -minline-sqrt-max-throughput
10485 @opindex minline-sqrt-max-throughput
10486 Generate code for inline square roots
10487 using the maximum throughput algorithm.
10489 @item -mno-dwarf2-asm
10490 @itemx -mdwarf2-asm
10491 @opindex mno-dwarf2-asm
10492 @opindex mdwarf2-asm
10493 Don't (or do) generate assembler code for the DWARF2 line number debugging
10494 info. This may be useful when not using the GNU assembler.
10496 @item -mearly-stop-bits
10497 @itemx -mno-early-stop-bits
10498 @opindex mearly-stop-bits
10499 @opindex mno-early-stop-bits
10500 Allow stop bits to be placed earlier than immediately preceding the
10501 instruction that triggered the stop bit. This can improve instruction
10502 scheduling, but does not always do so.
10504 @item -mfixed-range=@var{register-range}
10505 @opindex mfixed-range
10506 Generate code treating the given register range as fixed registers.
10507 A fixed register is one that the register allocator can not use. This is
10508 useful when compiling kernel code. A register range is specified as
10509 two registers separated by a dash. Multiple register ranges can be
10510 specified separated by a comma.
10512 @item -mtls-size=@var{tls-size}
10514 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
10517 @item -mtune=@var{cpu-type}
10519 Tune the instruction scheduling for a particular CPU, Valid values are
10520 itanium, itanium1, merced, itanium2, and mckinley.
10526 Add support for multithreading using the POSIX threads library. This
10527 option sets flags for both the preprocessor and linker. It does
10528 not affect the thread safety of object code produced by the compiler or
10529 that of libraries supplied with it. These are HP-UX specific flags.
10535 Generate code for a 32-bit or 64-bit environment.
10536 The 32-bit environment sets int, long and pointer to 32 bits.
10537 The 64-bit environment sets int to 32 bits and long and pointer
10538 to 64 bits. These are HP-UX specific flags.
10540 @item -mno-sched-br-data-spec
10541 @itemx -msched-br-data-spec
10542 @opindex mno-sched-br-data-spec
10543 @opindex msched-br-data-spec
10544 (Dis/En)able data speculative scheduling before reload.
10545 This will result in generation of the ld.a instructions and
10546 the corresponding check instructions (ld.c / chk.a).
10547 The default is 'disable'.
10549 @item -msched-ar-data-spec
10550 @itemx -mno-sched-ar-data-spec
10551 @opindex msched-ar-data-spec
10552 @opindex mno-sched-ar-data-spec
10553 (En/Dis)able data speculative scheduling after reload.
10554 This will result in generation of the ld.a instructions and
10555 the corresponding check instructions (ld.c / chk.a).
10556 The default is 'enable'.
10558 @item -mno-sched-control-spec
10559 @itemx -msched-control-spec
10560 @opindex mno-sched-control-spec
10561 @opindex msched-control-spec
10562 (Dis/En)able control speculative scheduling. This feature is
10563 available only during region scheduling (i.e. before reload).
10564 This will result in generation of the ld.s instructions and
10565 the corresponding check instructions chk.s .
10566 The default is 'disable'.
10568 @item -msched-br-in-data-spec
10569 @itemx -mno-sched-br-in-data-spec
10570 @opindex msched-br-in-data-spec
10571 @opindex mno-sched-br-in-data-spec
10572 (En/Dis)able speculative scheduling of the instructions that
10573 are dependent on the data speculative loads before reload.
10574 This is effective only with @option{-msched-br-data-spec} enabled.
10575 The default is 'enable'.
10577 @item -msched-ar-in-data-spec
10578 @itemx -mno-sched-ar-in-data-spec
10579 @opindex msched-ar-in-data-spec
10580 @opindex mno-sched-ar-in-data-spec
10581 (En/Dis)able speculative scheduling of the instructions that
10582 are dependent on the data speculative loads after reload.
10583 This is effective only with @option{-msched-ar-data-spec} enabled.
10584 The default is 'enable'.
10586 @item -msched-in-control-spec
10587 @itemx -mno-sched-in-control-spec
10588 @opindex msched-in-control-spec
10589 @opindex mno-sched-in-control-spec
10590 (En/Dis)able speculative scheduling of the instructions that
10591 are dependent on the control speculative loads.
10592 This is effective only with @option{-msched-control-spec} enabled.
10593 The default is 'enable'.
10596 @itemx -mno-sched-ldc
10597 @opindex msched-ldc
10598 @opindex mno-sched-ldc
10599 (En/Dis)able use of simple data speculation checks ld.c .
10600 If disabled, only chk.a instructions will be emitted to check
10601 data speculative loads.
10602 The default is 'enable'.
10604 @item -mno-sched-control-ldc
10605 @itemx -msched-control-ldc
10606 @opindex mno-sched-control-ldc
10607 @opindex msched-control-ldc
10608 (Dis/En)able use of ld.c instructions to check control speculative loads.
10609 If enabled, in case of control speculative load with no speculatively
10610 scheduled dependent instructions this load will be emitted as ld.sa and
10611 ld.c will be used to check it.
10612 The default is 'disable'.
10614 @item -mno-sched-spec-verbose
10615 @itemx -msched-spec-verbose
10616 @opindex mno-sched-spec-verbose
10617 @opindex msched-spec-verbose
10618 (Dis/En)able printing of the information about speculative motions.
10620 @item -mno-sched-prefer-non-data-spec-insns
10621 @itemx -msched-prefer-non-data-spec-insns
10622 @opindex mno-sched-prefer-non-data-spec-insns
10623 @opindex msched-prefer-non-data-spec-insns
10624 If enabled, data speculative instructions will be chosen for schedule
10625 only if there are no other choices at the moment. This will make
10626 the use of the data speculation much more conservative.
10627 The default is 'disable'.
10629 @item -mno-sched-prefer-non-control-spec-insns
10630 @itemx -msched-prefer-non-control-spec-insns
10631 @opindex mno-sched-prefer-non-control-spec-insns
10632 @opindex msched-prefer-non-control-spec-insns
10633 If enabled, control speculative instructions will be chosen for schedule
10634 only if there are no other choices at the moment. This will make
10635 the use of the control speculation much more conservative.
10636 The default is 'disable'.
10638 @item -mno-sched-count-spec-in-critical-path
10639 @itemx -msched-count-spec-in-critical-path
10640 @opindex mno-sched-count-spec-in-critical-path
10641 @opindex msched-count-spec-in-critical-path
10642 If enabled, speculative dependencies will be considered during
10643 computation of the instructions priorities. This will make the use of the
10644 speculation a bit more conservative.
10645 The default is 'disable'.
10650 @subsection M32C Options
10651 @cindex M32C options
10654 @item -mcpu=@var{name}
10656 Select the CPU for which code is generated. @var{name} may be one of
10657 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10658 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10659 the M32C/80 series.
10663 Specifies that the program will be run on the simulator. This causes
10664 an alternate runtime library to be linked in which supports, for
10665 example, file I/O. You must not use this option when generating
10666 programs that will run on real hardware; you must provide your own
10667 runtime library for whatever I/O functions are needed.
10669 @item -memregs=@var{number}
10671 Specifies the number of memory-based pseudo-registers GCC will use
10672 during code generation. These pseudo-registers will be used like real
10673 registers, so there is a tradeoff between GCC's ability to fit the
10674 code into available registers, and the performance penalty of using
10675 memory instead of registers. Note that all modules in a program must
10676 be compiled with the same value for this option. Because of that, you
10677 must not use this option with the default runtime libraries gcc
10682 @node M32R/D Options
10683 @subsection M32R/D Options
10684 @cindex M32R/D options
10686 These @option{-m} options are defined for Renesas M32R/D architectures:
10691 Generate code for the M32R/2@.
10695 Generate code for the M32R/X@.
10699 Generate code for the M32R@. This is the default.
10701 @item -mmodel=small
10702 @opindex mmodel=small
10703 Assume all objects live in the lower 16MB of memory (so that their addresses
10704 can be loaded with the @code{ld24} instruction), and assume all subroutines
10705 are reachable with the @code{bl} instruction.
10706 This is the default.
10708 The addressability of a particular object can be set with the
10709 @code{model} attribute.
10711 @item -mmodel=medium
10712 @opindex mmodel=medium
10713 Assume objects may be anywhere in the 32-bit address space (the compiler
10714 will generate @code{seth/add3} instructions to load their addresses), and
10715 assume all subroutines are reachable with the @code{bl} instruction.
10717 @item -mmodel=large
10718 @opindex mmodel=large
10719 Assume objects may be anywhere in the 32-bit address space (the compiler
10720 will generate @code{seth/add3} instructions to load their addresses), and
10721 assume subroutines may not be reachable with the @code{bl} instruction
10722 (the compiler will generate the much slower @code{seth/add3/jl}
10723 instruction sequence).
10726 @opindex msdata=none
10727 Disable use of the small data area. Variables will be put into
10728 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10729 @code{section} attribute has been specified).
10730 This is the default.
10732 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10733 Objects may be explicitly put in the small data area with the
10734 @code{section} attribute using one of these sections.
10736 @item -msdata=sdata
10737 @opindex msdata=sdata
10738 Put small global and static data in the small data area, but do not
10739 generate special code to reference them.
10742 @opindex msdata=use
10743 Put small global and static data in the small data area, and generate
10744 special instructions to reference them.
10748 @cindex smaller data references
10749 Put global and static objects less than or equal to @var{num} bytes
10750 into the small data or bss sections instead of the normal data or bss
10751 sections. The default value of @var{num} is 8.
10752 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10753 for this option to have any effect.
10755 All modules should be compiled with the same @option{-G @var{num}} value.
10756 Compiling with different values of @var{num} may or may not work; if it
10757 doesn't the linker will give an error message---incorrect code will not be
10762 Makes the M32R specific code in the compiler display some statistics
10763 that might help in debugging programs.
10765 @item -malign-loops
10766 @opindex malign-loops
10767 Align all loops to a 32-byte boundary.
10769 @item -mno-align-loops
10770 @opindex mno-align-loops
10771 Do not enforce a 32-byte alignment for loops. This is the default.
10773 @item -missue-rate=@var{number}
10774 @opindex missue-rate=@var{number}
10775 Issue @var{number} instructions per cycle. @var{number} can only be 1
10778 @item -mbranch-cost=@var{number}
10779 @opindex mbranch-cost=@var{number}
10780 @var{number} can only be 1 or 2. If it is 1 then branches will be
10781 preferred over conditional code, if it is 2, then the opposite will
10784 @item -mflush-trap=@var{number}
10785 @opindex mflush-trap=@var{number}
10786 Specifies the trap number to use to flush the cache. The default is
10787 12. Valid numbers are between 0 and 15 inclusive.
10789 @item -mno-flush-trap
10790 @opindex mno-flush-trap
10791 Specifies that the cache cannot be flushed by using a trap.
10793 @item -mflush-func=@var{name}
10794 @opindex mflush-func=@var{name}
10795 Specifies the name of the operating system function to call to flush
10796 the cache. The default is @emph{_flush_cache}, but a function call
10797 will only be used if a trap is not available.
10799 @item -mno-flush-func
10800 @opindex mno-flush-func
10801 Indicates that there is no OS function for flushing the cache.
10805 @node M680x0 Options
10806 @subsection M680x0 Options
10807 @cindex M680x0 options
10809 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
10810 The default settings depend on which architecture was selected when
10811 the compiler was configured; the defaults for the most common choices
10815 @item -march=@var{arch}
10817 Generate code for a specific M680x0 or ColdFire instruction set
10818 architecture. Permissible values of @var{arch} for M680x0
10819 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
10820 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
10821 architectures are selected according to Freescale's ISA classification
10822 and the permissible values are: @samp{isaa}, @samp{isaaplus},
10823 @samp{isab} and @samp{isac}.
10825 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
10826 code for a ColdFire target. The @var{arch} in this macro is one of the
10827 @option{-march} arguments given above.
10829 When used together, @option{-march} and @option{-mtune} select code
10830 that runs on a family of similar processors but that is optimized
10831 for a particular microarchitecture.
10833 @item -mcpu=@var{cpu}
10835 Generate code for a specific M680x0 or ColdFire processor.
10836 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
10837 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
10838 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
10839 below, which also classifies the CPUs into families:
10841 @multitable @columnfractions 0.20 0.80
10842 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
10843 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
10844 @item @samp{5206e} @tab @samp{5206e}
10845 @item @samp{5208} @tab @samp{5207} @samp{5208}
10846 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
10847 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
10848 @item @samp{5216} @tab @samp{5214} @samp{5216}
10849 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
10850 @item @samp{5225} @tab @samp{5224} @samp{5225}
10851 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
10852 @item @samp{5249} @tab @samp{5249}
10853 @item @samp{5250} @tab @samp{5250}
10854 @item @samp{5271} @tab @samp{5270} @samp{5271}
10855 @item @samp{5272} @tab @samp{5272}
10856 @item @samp{5275} @tab @samp{5274} @samp{5275}
10857 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
10858 @item @samp{5307} @tab @samp{5307}
10859 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
10860 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
10861 @item @samp{5407} @tab @samp{5407}
10862 @item @samp{5475} @tab @samp{5470} @samp{5471} @samp{5472} @samp{5473} @samp{5474} @samp{5475} @samp{547x} @samp{5480} @samp{5481} @samp{5482} @samp{5483} @samp{5484} @samp{5485}
10865 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
10866 @var{arch} is compatible with @var{cpu}. Other combinations of
10867 @option{-mcpu} and @option{-march} are rejected.
10869 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
10870 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
10871 where the value of @var{family} is given by the table above.
10873 @item -mtune=@var{tune}
10875 Tune the code for a particular microarchitecture, within the
10876 constraints set by @option{-march} and @option{-mcpu}.
10877 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
10878 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
10879 and @samp{cpu32}. The ColdFire microarchitectures
10880 are: @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
10882 You can also use @option{-mtune=68020-40} for code that needs
10883 to run relatively well on 68020, 68030 and 68040 targets.
10884 @option{-mtune=68020-60} is similar but includes 68060 targets
10885 as well. These two options select the same tuning decisions as
10886 @option{-m68020-40} and @option{-m68020-60} respectively.
10888 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
10889 when tuning for 680x0 architecture @var{arch}. It also defines
10890 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
10891 option is used. If gcc is tuning for a range of architectures,
10892 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
10893 it defines the macros for every architecture in the range.
10895 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
10896 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
10897 of the arguments given above.
10903 Generate output for a 68000. This is the default
10904 when the compiler is configured for 68000-based systems.
10905 It is equivalent to @option{-march=68000}.
10907 Use this option for microcontrollers with a 68000 or EC000 core,
10908 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10912 Generate output for a 68010. This is the default
10913 when the compiler is configured for 68010-based systems.
10914 It is equivalent to @option{-march=68010}.
10920 Generate output for a 68020. This is the default
10921 when the compiler is configured for 68020-based systems.
10922 It is equivalent to @option{-march=68020}.
10926 Generate output for a 68030. This is the default when the compiler is
10927 configured for 68030-based systems. It is equivalent to
10928 @option{-march=68030}.
10932 Generate output for a 68040. This is the default when the compiler is
10933 configured for 68040-based systems. It is equivalent to
10934 @option{-march=68040}.
10936 This option inhibits the use of 68881/68882 instructions that have to be
10937 emulated by software on the 68040. Use this option if your 68040 does not
10938 have code to emulate those instructions.
10942 Generate output for a 68060. This is the default when the compiler is
10943 configured for 68060-based systems. It is equivalent to
10944 @option{-march=68060}.
10946 This option inhibits the use of 68020 and 68881/68882 instructions that
10947 have to be emulated by software on the 68060. Use this option if your 68060
10948 does not have code to emulate those instructions.
10952 Generate output for a CPU32. This is the default
10953 when the compiler is configured for CPU32-based systems.
10954 It is equivalent to @option{-march=cpu32}.
10956 Use this option for microcontrollers with a
10957 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10958 68336, 68340, 68341, 68349 and 68360.
10962 Generate output for a 520X ColdFire CPU. This is the default
10963 when the compiler is configured for 520X-based systems.
10964 It is equivalent to @option{-mcpu=5206}, and is now deprecated
10965 in favor of that option.
10967 Use this option for microcontroller with a 5200 core, including
10968 the MCF5202, MCF5203, MCF5204 and MCF5206.
10972 Generate output for a 5206e ColdFire CPU. The option is now
10973 deprecated in favor of the equivalent @option{-mcpu=5206e}.
10977 Generate output for a member of the ColdFire 528X family.
10978 The option is now deprecated in favor of the equivalent
10979 @option{-mcpu=528x}.
10983 Generate output for a ColdFire 5307 CPU. The option is now deprecated
10984 in favor of the equivalent @option{-mcpu=5307}.
10988 Generate output for a ColdFire 5407 CPU. The option is now deprecated
10989 in favor of the equivalent @option{-mcpu=5407}.
10993 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
10994 This includes use of hardware floating point instructions.
10995 The option is equivalent to @option{-mcpu=547x}, and is now
10996 deprecated in favor of that option.
11000 Generate output for a 68040, without using any of the new instructions.
11001 This results in code which can run relatively efficiently on either a
11002 68020/68881 or a 68030 or a 68040. The generated code does use the
11003 68881 instructions that are emulated on the 68040.
11005 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11009 Generate output for a 68060, without using any of the new instructions.
11010 This results in code which can run relatively efficiently on either a
11011 68020/68881 or a 68030 or a 68040. The generated code does use the
11012 68881 instructions that are emulated on the 68060.
11014 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11018 @opindex mhard-float
11020 Generate floating-point instructions. This is the default for 68020
11021 and above, and for ColdFire devices that have an FPU. It defines the
11022 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11023 on ColdFire targets.
11026 @opindex msoft-float
11027 Do not generate floating-point instructions; use library calls instead.
11028 This is the default for 68000, 68010, and 68832 targets. It is also
11029 the default for ColdFire devices that have no FPU.
11035 Generate (do not generate) ColdFire hardware divide and remainder
11036 instructions. If @option{-march} is used without @option{-mcpu},
11037 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11038 architectures. Otherwise, the default is taken from the target CPU
11039 (either the default CPU, or the one specified by @option{-mcpu}). For
11040 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11041 @option{-mcpu=5206e}.
11043 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11047 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11048 Additionally, parameters passed on the stack are also aligned to a
11049 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11053 Do not consider type @code{int} to be 16 bits wide. This is the default.
11056 @itemx -mno-bitfield
11057 @opindex mnobitfield
11058 @opindex mno-bitfield
11059 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11060 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11064 Do use the bit-field instructions. The @option{-m68020} option implies
11065 @option{-mbitfield}. This is the default if you use a configuration
11066 designed for a 68020.
11070 Use a different function-calling convention, in which functions
11071 that take a fixed number of arguments return with the @code{rtd}
11072 instruction, which pops their arguments while returning. This
11073 saves one instruction in the caller since there is no need to pop
11074 the arguments there.
11076 This calling convention is incompatible with the one normally
11077 used on Unix, so you cannot use it if you need to call libraries
11078 compiled with the Unix compiler.
11080 Also, you must provide function prototypes for all functions that
11081 take variable numbers of arguments (including @code{printf});
11082 otherwise incorrect code will be generated for calls to those
11085 In addition, seriously incorrect code will result if you call a
11086 function with too many arguments. (Normally, extra arguments are
11087 harmlessly ignored.)
11089 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11090 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11094 Do not use the calling conventions selected by @option{-mrtd}.
11095 This is the default.
11098 @itemx -mno-align-int
11099 @opindex malign-int
11100 @opindex mno-align-int
11101 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11102 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11103 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11104 Aligning variables on 32-bit boundaries produces code that runs somewhat
11105 faster on processors with 32-bit busses at the expense of more memory.
11107 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11108 align structures containing the above types differently than
11109 most published application binary interface specifications for the m68k.
11113 Use the pc-relative addressing mode of the 68000 directly, instead of
11114 using a global offset table. At present, this option implies @option{-fpic},
11115 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11116 not presently supported with @option{-mpcrel}, though this could be supported for
11117 68020 and higher processors.
11119 @item -mno-strict-align
11120 @itemx -mstrict-align
11121 @opindex mno-strict-align
11122 @opindex mstrict-align
11123 Do not (do) assume that unaligned memory references will be handled by
11127 Generate code that allows the data segment to be located in a different
11128 area of memory from the text segment. This allows for execute in place in
11129 an environment without virtual memory management. This option implies
11132 @item -mno-sep-data
11133 Generate code that assumes that the data segment follows the text segment.
11134 This is the default.
11136 @item -mid-shared-library
11137 Generate code that supports shared libraries via the library ID method.
11138 This allows for execute in place and shared libraries in an environment
11139 without virtual memory management. This option implies @option{-fPIC}.
11141 @item -mno-id-shared-library
11142 Generate code that doesn't assume ID based shared libraries are being used.
11143 This is the default.
11145 @item -mshared-library-id=n
11146 Specified the identification number of the ID based shared library being
11147 compiled. Specifying a value of 0 will generate more compact code, specifying
11148 other values will force the allocation of that number to the current
11149 library but is no more space or time efficient than omitting this option.
11153 @node M68hc1x Options
11154 @subsection M68hc1x Options
11155 @cindex M68hc1x options
11157 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11158 microcontrollers. The default values for these options depends on
11159 which style of microcontroller was selected when the compiler was configured;
11160 the defaults for the most common choices are given below.
11167 Generate output for a 68HC11. This is the default
11168 when the compiler is configured for 68HC11-based systems.
11174 Generate output for a 68HC12. This is the default
11175 when the compiler is configured for 68HC12-based systems.
11181 Generate output for a 68HCS12.
11183 @item -mauto-incdec
11184 @opindex mauto-incdec
11185 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11192 Enable the use of 68HC12 min and max instructions.
11195 @itemx -mno-long-calls
11196 @opindex mlong-calls
11197 @opindex mno-long-calls
11198 Treat all calls as being far away (near). If calls are assumed to be
11199 far away, the compiler will use the @code{call} instruction to
11200 call a function and the @code{rtc} instruction for returning.
11204 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11206 @item -msoft-reg-count=@var{count}
11207 @opindex msoft-reg-count
11208 Specify the number of pseudo-soft registers which are used for the
11209 code generation. The maximum number is 32. Using more pseudo-soft
11210 register may or may not result in better code depending on the program.
11211 The default is 4 for 68HC11 and 2 for 68HC12.
11215 @node MCore Options
11216 @subsection MCore Options
11217 @cindex MCore options
11219 These are the @samp{-m} options defined for the Motorola M*Core
11225 @itemx -mno-hardlit
11227 @opindex mno-hardlit
11228 Inline constants into the code stream if it can be done in two
11229 instructions or less.
11235 Use the divide instruction. (Enabled by default).
11237 @item -mrelax-immediate
11238 @itemx -mno-relax-immediate
11239 @opindex mrelax-immediate
11240 @opindex mno-relax-immediate
11241 Allow arbitrary sized immediates in bit operations.
11243 @item -mwide-bitfields
11244 @itemx -mno-wide-bitfields
11245 @opindex mwide-bitfields
11246 @opindex mno-wide-bitfields
11247 Always treat bit-fields as int-sized.
11249 @item -m4byte-functions
11250 @itemx -mno-4byte-functions
11251 @opindex m4byte-functions
11252 @opindex mno-4byte-functions
11253 Force all functions to be aligned to a four byte boundary.
11255 @item -mcallgraph-data
11256 @itemx -mno-callgraph-data
11257 @opindex mcallgraph-data
11258 @opindex mno-callgraph-data
11259 Emit callgraph information.
11262 @itemx -mno-slow-bytes
11263 @opindex mslow-bytes
11264 @opindex mno-slow-bytes
11265 Prefer word access when reading byte quantities.
11267 @item -mlittle-endian
11268 @itemx -mbig-endian
11269 @opindex mlittle-endian
11270 @opindex mbig-endian
11271 Generate code for a little endian target.
11277 Generate code for the 210 processor.
11281 @subsection MIPS Options
11282 @cindex MIPS options
11288 Generate big-endian code.
11292 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11295 @item -march=@var{arch}
11297 Generate code that will run on @var{arch}, which can be the name of a
11298 generic MIPS ISA, or the name of a particular processor.
11300 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11301 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11302 The processor names are:
11303 @samp{4kc}, @samp{4km}, @samp{4kp},
11304 @samp{4kec}, @samp{4kem}, @samp{4kep},
11305 @samp{5kc}, @samp{5kf},
11307 @samp{24kc}, @samp{24kf}, @samp{24kx},
11308 @samp{24kec}, @samp{24kef}, @samp{24kex},
11309 @samp{34kc}, @samp{34kf}, @samp{34kx},
11310 @samp{74kc}, @samp{74kf}, @samp{74kx},
11313 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11314 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11315 @samp{rm7000}, @samp{rm9000},
11318 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11319 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11320 The special value @samp{from-abi} selects the
11321 most compatible architecture for the selected ABI (that is,
11322 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11324 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11325 (for example, @samp{-march=r2k}). Prefixes are optional, and
11326 @samp{vr} may be written @samp{r}.
11328 GCC defines two macros based on the value of this option. The first
11329 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11330 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11331 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11332 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11333 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11335 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11336 above. In other words, it will have the full prefix and will not
11337 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11338 the macro names the resolved architecture (either @samp{"mips1"} or
11339 @samp{"mips3"}). It names the default architecture when no
11340 @option{-march} option is given.
11342 @item -mtune=@var{arch}
11344 Optimize for @var{arch}. Among other things, this option controls
11345 the way instructions are scheduled, and the perceived cost of arithmetic
11346 operations. The list of @var{arch} values is the same as for
11349 When this option is not used, GCC will optimize for the processor
11350 specified by @option{-march}. By using @option{-march} and
11351 @option{-mtune} together, it is possible to generate code that will
11352 run on a family of processors, but optimize the code for one
11353 particular member of that family.
11355 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11356 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11357 @samp{-march} ones described above.
11361 Equivalent to @samp{-march=mips1}.
11365 Equivalent to @samp{-march=mips2}.
11369 Equivalent to @samp{-march=mips3}.
11373 Equivalent to @samp{-march=mips4}.
11377 Equivalent to @samp{-march=mips32}.
11381 Equivalent to @samp{-march=mips32r2}.
11385 Equivalent to @samp{-march=mips64}.
11390 @opindex mno-mips16
11391 Generate (do not generate) MIPS16 code. If GCC is targetting a
11392 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11404 Generate code for the given ABI@.
11406 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11407 generates 64-bit code when you select a 64-bit architecture, but you
11408 can use @option{-mgp32} to get 32-bit code instead.
11410 For information about the O64 ABI, see
11411 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11413 GCC supports a variant of the o32 ABI in which floating-point registers
11414 are 64 rather than 32 bits wide. You can select this combination with
11415 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11416 and @samp{mfhc1} instructions and is therefore only supported for
11417 MIPS32R2 processors.
11419 The register assignments for arguments and return values remain the
11420 same, but each scalar value is passed in a single 64-bit register
11421 rather than a pair of 32-bit registers. For example, scalar
11422 floating-point values are returned in @samp{$f0} only, not a
11423 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11424 remains the same, but all 64 bits are saved.
11427 @itemx -mno-abicalls
11429 @opindex mno-abicalls
11430 Generate (do not generate) code that is suitable for SVR4-style
11431 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11436 Generate (do not generate) code that is fully position-independent,
11437 and that can therefore be linked into shared libraries. This option
11438 only affects @option{-mabicalls}.
11440 All @option{-mabicalls} code has traditionally been position-independent,
11441 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11442 as an extension, the GNU toolchain allows executables to use absolute
11443 accesses for locally-binding symbols. It can also use shorter GP
11444 initialization sequences and generate direct calls to locally-defined
11445 functions. This mode is selected by @option{-mno-shared}.
11447 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11448 objects that can only be linked by the GNU linker. However, the option
11449 does not affect the ABI of the final executable; it only affects the ABI
11450 of relocatable objects. Using @option{-mno-shared} will generally make
11451 executables both smaller and quicker.
11453 @option{-mshared} is the default.
11459 Lift (do not lift) the usual restrictions on the size of the global
11462 GCC normally uses a single instruction to load values from the GOT@.
11463 While this is relatively efficient, it will only work if the GOT
11464 is smaller than about 64k. Anything larger will cause the linker
11465 to report an error such as:
11467 @cindex relocation truncated to fit (MIPS)
11469 relocation truncated to fit: R_MIPS_GOT16 foobar
11472 If this happens, you should recompile your code with @option{-mxgot}.
11473 It should then work with very large GOTs, although it will also be
11474 less efficient, since it will take three instructions to fetch the
11475 value of a global symbol.
11477 Note that some linkers can create multiple GOTs. If you have such a
11478 linker, you should only need to use @option{-mxgot} when a single object
11479 file accesses more than 64k's worth of GOT entries. Very few do.
11481 These options have no effect unless GCC is generating position
11486 Assume that general-purpose registers are 32 bits wide.
11490 Assume that general-purpose registers are 64 bits wide.
11494 Assume that floating-point registers are 32 bits wide.
11498 Assume that floating-point registers are 64 bits wide.
11501 @opindex mhard-float
11502 Use floating-point coprocessor instructions.
11505 @opindex msoft-float
11506 Do not use floating-point coprocessor instructions. Implement
11507 floating-point calculations using library calls instead.
11509 @item -msingle-float
11510 @opindex msingle-float
11511 Assume that the floating-point coprocessor only supports single-precision
11514 @item -mdouble-float
11515 @opindex mdouble-float
11516 Assume that the floating-point coprocessor supports double-precision
11517 operations. This is the default.
11523 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
11529 Use (do not use) the MIPS DSP ASE REV 2. @xref{MIPS DSP Built-in Functions}.
11530 The option @option{-mdspr2} implies @option{-mdsp}.
11532 @item -mpaired-single
11533 @itemx -mno-paired-single
11534 @opindex mpaired-single
11535 @opindex mno-paired-single
11536 Use (do not use) paired-single floating-point instructions.
11537 @xref{MIPS Paired-Single Support}. This option can only be used
11538 when generating 64-bit code and requires hardware floating-point
11539 support to be enabled.
11545 Use (do not use) MIPS Digital Media Extension instructions.
11546 This option can only be used when generating 64-bit code and requires
11547 hardware floating-point support to be enabled.
11552 @opindex mno-mips3d
11553 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
11554 The option @option{-mips3d} implies @option{-mpaired-single}.
11560 Use (do not use) MT Multithreading instructions.
11564 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
11565 an explanation of the default and the way that the pointer size is
11570 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
11572 The default size of @code{int}s, @code{long}s and pointers depends on
11573 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
11574 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
11575 32-bit @code{long}s. Pointers are the same size as @code{long}s,
11576 or the same size as integer registers, whichever is smaller.
11582 Assume (do not assume) that all symbols have 32-bit values, regardless
11583 of the selected ABI@. This option is useful in combination with
11584 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
11585 to generate shorter and faster references to symbolic addresses.
11589 @cindex smaller data references (MIPS)
11590 @cindex gp-relative references (MIPS)
11591 Put global and static items less than or equal to @var{num} bytes into
11592 the small data or bss section instead of the normal data or bss section.
11593 This allows the data to be accessed using a single instruction.
11595 All modules should be compiled with the same @option{-G @var{num}}
11598 @item -membedded-data
11599 @itemx -mno-embedded-data
11600 @opindex membedded-data
11601 @opindex mno-embedded-data
11602 Allocate variables to the read-only data section first if possible, then
11603 next in the small data section if possible, otherwise in data. This gives
11604 slightly slower code than the default, but reduces the amount of RAM required
11605 when executing, and thus may be preferred for some embedded systems.
11607 @item -muninit-const-in-rodata
11608 @itemx -mno-uninit-const-in-rodata
11609 @opindex muninit-const-in-rodata
11610 @opindex mno-uninit-const-in-rodata
11611 Put uninitialized @code{const} variables in the read-only data section.
11612 This option is only meaningful in conjunction with @option{-membedded-data}.
11614 @item -msplit-addresses
11615 @itemx -mno-split-addresses
11616 @opindex msplit-addresses
11617 @opindex mno-split-addresses
11618 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
11619 relocation operators. This option has been superseded by
11620 @option{-mexplicit-relocs} but is retained for backwards compatibility.
11622 @item -mexplicit-relocs
11623 @itemx -mno-explicit-relocs
11624 @opindex mexplicit-relocs
11625 @opindex mno-explicit-relocs
11626 Use (do not use) assembler relocation operators when dealing with symbolic
11627 addresses. The alternative, selected by @option{-mno-explicit-relocs},
11628 is to use assembler macros instead.
11630 @option{-mexplicit-relocs} is the default if GCC was configured
11631 to use an assembler that supports relocation operators.
11633 @item -mcheck-zero-division
11634 @itemx -mno-check-zero-division
11635 @opindex mcheck-zero-division
11636 @opindex mno-check-zero-division
11637 Trap (do not trap) on integer division by zero.
11639 The default is @option{-mcheck-zero-division}.
11641 @item -mdivide-traps
11642 @itemx -mdivide-breaks
11643 @opindex mdivide-traps
11644 @opindex mdivide-breaks
11645 MIPS systems check for division by zero by generating either a
11646 conditional trap or a break instruction. Using traps results in
11647 smaller code, but is only supported on MIPS II and later. Also, some
11648 versions of the Linux kernel have a bug that prevents trap from
11649 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
11650 allow conditional traps on architectures that support them and
11651 @option{-mdivide-breaks} to force the use of breaks.
11653 The default is usually @option{-mdivide-traps}, but this can be
11654 overridden at configure time using @option{--with-divide=breaks}.
11655 Divide-by-zero checks can be completely disabled using
11656 @option{-mno-check-zero-division}.
11661 @opindex mno-memcpy
11662 Force (do not force) the use of @code{memcpy()} for non-trivial block
11663 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
11664 most constant-sized copies.
11667 @itemx -mno-long-calls
11668 @opindex mlong-calls
11669 @opindex mno-long-calls
11670 Disable (do not disable) use of the @code{jal} instruction. Calling
11671 functions using @code{jal} is more efficient but requires the caller
11672 and callee to be in the same 256 megabyte segment.
11674 This option has no effect on abicalls code. The default is
11675 @option{-mno-long-calls}.
11681 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
11682 instructions, as provided by the R4650 ISA@.
11685 @itemx -mno-fused-madd
11686 @opindex mfused-madd
11687 @opindex mno-fused-madd
11688 Enable (disable) use of the floating point multiply-accumulate
11689 instructions, when they are available. The default is
11690 @option{-mfused-madd}.
11692 When multiply-accumulate instructions are used, the intermediate
11693 product is calculated to infinite precision and is not subject to
11694 the FCSR Flush to Zero bit. This may be undesirable in some
11699 Tell the MIPS assembler to not run its preprocessor over user
11700 assembler files (with a @samp{.s} suffix) when assembling them.
11703 @itemx -mno-fix-r4000
11704 @opindex mfix-r4000
11705 @opindex mno-fix-r4000
11706 Work around certain R4000 CPU errata:
11709 A double-word or a variable shift may give an incorrect result if executed
11710 immediately after starting an integer division.
11712 A double-word or a variable shift may give an incorrect result if executed
11713 while an integer multiplication is in progress.
11715 An integer division may give an incorrect result if started in a delay slot
11716 of a taken branch or a jump.
11720 @itemx -mno-fix-r4400
11721 @opindex mfix-r4400
11722 @opindex mno-fix-r4400
11723 Work around certain R4400 CPU errata:
11726 A double-word or a variable shift may give an incorrect result if executed
11727 immediately after starting an integer division.
11731 @itemx -mno-fix-vr4120
11732 @opindex mfix-vr4120
11733 Work around certain VR4120 errata:
11736 @code{dmultu} does not always produce the correct result.
11738 @code{div} and @code{ddiv} do not always produce the correct result if one
11739 of the operands is negative.
11741 The workarounds for the division errata rely on special functions in
11742 @file{libgcc.a}. At present, these functions are only provided by
11743 the @code{mips64vr*-elf} configurations.
11745 Other VR4120 errata require a nop to be inserted between certain pairs of
11746 instructions. These errata are handled by the assembler, not by GCC itself.
11749 @opindex mfix-vr4130
11750 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
11751 workarounds are implemented by the assembler rather than by GCC,
11752 although GCC will avoid using @code{mflo} and @code{mfhi} if the
11753 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
11754 instructions are available instead.
11757 @itemx -mno-fix-sb1
11759 Work around certain SB-1 CPU core errata.
11760 (This flag currently works around the SB-1 revision 2
11761 ``F1'' and ``F2'' floating point errata.)
11763 @item -mflush-func=@var{func}
11764 @itemx -mno-flush-func
11765 @opindex mflush-func
11766 Specifies the function to call to flush the I and D caches, or to not
11767 call any such function. If called, the function must take the same
11768 arguments as the common @code{_flush_func()}, that is, the address of the
11769 memory range for which the cache is being flushed, the size of the
11770 memory range, and the number 3 (to flush both caches). The default
11771 depends on the target GCC was configured for, but commonly is either
11772 @samp{_flush_func} or @samp{__cpu_flush}.
11774 @item -mbranch-likely
11775 @itemx -mno-branch-likely
11776 @opindex mbranch-likely
11777 @opindex mno-branch-likely
11778 Enable or disable use of Branch Likely instructions, regardless of the
11779 default for the selected architecture. By default, Branch Likely
11780 instructions may be generated if they are supported by the selected
11781 architecture. An exception is for the MIPS32 and MIPS64 architectures
11782 and processors which implement those architectures; for those, Branch
11783 Likely instructions will not be generated by default because the MIPS32
11784 and MIPS64 architectures specifically deprecate their use.
11786 @item -mfp-exceptions
11787 @itemx -mno-fp-exceptions
11788 @opindex mfp-exceptions
11789 Specifies whether FP exceptions are enabled. This affects how we schedule
11790 FP instructions for some processors. The default is that FP exceptions are
11793 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11794 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11797 @item -mvr4130-align
11798 @itemx -mno-vr4130-align
11799 @opindex mvr4130-align
11800 The VR4130 pipeline is two-way superscalar, but can only issue two
11801 instructions together if the first one is 8-byte aligned. When this
11802 option is enabled, GCC will align pairs of instructions that it
11803 thinks should execute in parallel.
11805 This option only has an effect when optimizing for the VR4130.
11806 It normally makes code faster, but at the expense of making it bigger.
11807 It is enabled by default at optimization level @option{-O3}.
11811 @subsection MMIX Options
11812 @cindex MMIX Options
11814 These options are defined for the MMIX:
11818 @itemx -mno-libfuncs
11820 @opindex mno-libfuncs
11821 Specify that intrinsic library functions are being compiled, passing all
11822 values in registers, no matter the size.
11825 @itemx -mno-epsilon
11827 @opindex mno-epsilon
11828 Generate floating-point comparison instructions that compare with respect
11829 to the @code{rE} epsilon register.
11831 @item -mabi=mmixware
11833 @opindex mabi-mmixware
11835 Generate code that passes function parameters and return values that (in
11836 the called function) are seen as registers @code{$0} and up, as opposed to
11837 the GNU ABI which uses global registers @code{$231} and up.
11839 @item -mzero-extend
11840 @itemx -mno-zero-extend
11841 @opindex mzero-extend
11842 @opindex mno-zero-extend
11843 When reading data from memory in sizes shorter than 64 bits, use (do not
11844 use) zero-extending load instructions by default, rather than
11845 sign-extending ones.
11848 @itemx -mno-knuthdiv
11850 @opindex mno-knuthdiv
11851 Make the result of a division yielding a remainder have the same sign as
11852 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
11853 remainder follows the sign of the dividend. Both methods are
11854 arithmetically valid, the latter being almost exclusively used.
11856 @item -mtoplevel-symbols
11857 @itemx -mno-toplevel-symbols
11858 @opindex mtoplevel-symbols
11859 @opindex mno-toplevel-symbols
11860 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
11861 code can be used with the @code{PREFIX} assembly directive.
11865 Generate an executable in the ELF format, rather than the default
11866 @samp{mmo} format used by the @command{mmix} simulator.
11868 @item -mbranch-predict
11869 @itemx -mno-branch-predict
11870 @opindex mbranch-predict
11871 @opindex mno-branch-predict
11872 Use (do not use) the probable-branch instructions, when static branch
11873 prediction indicates a probable branch.
11875 @item -mbase-addresses
11876 @itemx -mno-base-addresses
11877 @opindex mbase-addresses
11878 @opindex mno-base-addresses
11879 Generate (do not generate) code that uses @emph{base addresses}. Using a
11880 base address automatically generates a request (handled by the assembler
11881 and the linker) for a constant to be set up in a global register. The
11882 register is used for one or more base address requests within the range 0
11883 to 255 from the value held in the register. The generally leads to short
11884 and fast code, but the number of different data items that can be
11885 addressed is limited. This means that a program that uses lots of static
11886 data may require @option{-mno-base-addresses}.
11888 @item -msingle-exit
11889 @itemx -mno-single-exit
11890 @opindex msingle-exit
11891 @opindex mno-single-exit
11892 Force (do not force) generated code to have a single exit point in each
11896 @node MN10300 Options
11897 @subsection MN10300 Options
11898 @cindex MN10300 options
11900 These @option{-m} options are defined for Matsushita MN10300 architectures:
11905 Generate code to avoid bugs in the multiply instructions for the MN10300
11906 processors. This is the default.
11908 @item -mno-mult-bug
11909 @opindex mno-mult-bug
11910 Do not generate code to avoid bugs in the multiply instructions for the
11911 MN10300 processors.
11915 Generate code which uses features specific to the AM33 processor.
11919 Do not generate code which uses features specific to the AM33 processor. This
11922 @item -mreturn-pointer-on-d0
11923 @opindex mreturn-pointer-on-d0
11924 When generating a function which returns a pointer, return the pointer
11925 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11926 only in a0, and attempts to call such functions without a prototype
11927 would result in errors. Note that this option is on by default; use
11928 @option{-mno-return-pointer-on-d0} to disable it.
11932 Do not link in the C run-time initialization object file.
11936 Indicate to the linker that it should perform a relaxation optimization pass
11937 to shorten branches, calls and absolute memory addresses. This option only
11938 has an effect when used on the command line for the final link step.
11940 This option makes symbolic debugging impossible.
11944 @subsection MT Options
11947 These @option{-m} options are defined for Morpho MT architectures:
11951 @item -march=@var{cpu-type}
11953 Generate code that will run on @var{cpu-type}, which is the name of a system
11954 representing a certain processor type. Possible values for
11955 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11956 @samp{ms1-16-003} and @samp{ms2}.
11958 When this option is not used, the default is @option{-march=ms1-16-002}.
11962 Use byte loads and stores when generating code.
11966 Do not use byte loads and stores when generating code.
11970 Use simulator runtime
11974 Do not link in the C run-time initialization object file
11975 @file{crti.o}. Other run-time initialization and termination files
11976 such as @file{startup.o} and @file{exit.o} are still included on the
11977 linker command line.
11981 @node PDP-11 Options
11982 @subsection PDP-11 Options
11983 @cindex PDP-11 Options
11985 These options are defined for the PDP-11:
11990 Use hardware FPP floating point. This is the default. (FIS floating
11991 point on the PDP-11/40 is not supported.)
11994 @opindex msoft-float
11995 Do not use hardware floating point.
11999 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12003 Return floating-point results in memory. This is the default.
12007 Generate code for a PDP-11/40.
12011 Generate code for a PDP-11/45. This is the default.
12015 Generate code for a PDP-11/10.
12017 @item -mbcopy-builtin
12018 @opindex bcopy-builtin
12019 Use inline @code{movmemhi} patterns for copying memory. This is the
12024 Do not use inline @code{movmemhi} patterns for copying memory.
12030 Use 16-bit @code{int}. This is the default.
12036 Use 32-bit @code{int}.
12039 @itemx -mno-float32
12041 @opindex mno-float32
12042 Use 64-bit @code{float}. This is the default.
12045 @itemx -mno-float64
12047 @opindex mno-float64
12048 Use 32-bit @code{float}.
12052 Use @code{abshi2} pattern. This is the default.
12056 Do not use @code{abshi2} pattern.
12058 @item -mbranch-expensive
12059 @opindex mbranch-expensive
12060 Pretend that branches are expensive. This is for experimenting with
12061 code generation only.
12063 @item -mbranch-cheap
12064 @opindex mbranch-cheap
12065 Do not pretend that branches are expensive. This is the default.
12069 Generate code for a system with split I&D@.
12073 Generate code for a system without split I&D@. This is the default.
12077 Use Unix assembler syntax. This is the default when configured for
12078 @samp{pdp11-*-bsd}.
12082 Use DEC assembler syntax. This is the default when configured for any
12083 PDP-11 target other than @samp{pdp11-*-bsd}.
12086 @node PowerPC Options
12087 @subsection PowerPC Options
12088 @cindex PowerPC options
12090 These are listed under @xref{RS/6000 and PowerPC Options}.
12092 @node RS/6000 and PowerPC Options
12093 @subsection IBM RS/6000 and PowerPC Options
12094 @cindex RS/6000 and PowerPC Options
12095 @cindex IBM RS/6000 and PowerPC Options
12097 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12104 @itemx -mno-powerpc
12105 @itemx -mpowerpc-gpopt
12106 @itemx -mno-powerpc-gpopt
12107 @itemx -mpowerpc-gfxopt
12108 @itemx -mno-powerpc-gfxopt
12110 @itemx -mno-powerpc64
12114 @itemx -mno-popcntb
12126 @opindex mno-power2
12128 @opindex mno-powerpc
12129 @opindex mpowerpc-gpopt
12130 @opindex mno-powerpc-gpopt
12131 @opindex mpowerpc-gfxopt
12132 @opindex mno-powerpc-gfxopt
12133 @opindex mpowerpc64
12134 @opindex mno-powerpc64
12138 @opindex mno-popcntb
12144 @opindex mno-mfpgpr
12147 GCC supports two related instruction set architectures for the
12148 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12149 instructions supported by the @samp{rios} chip set used in the original
12150 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12151 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12152 the IBM 4xx, 6xx, and follow-on microprocessors.
12154 Neither architecture is a subset of the other. However there is a
12155 large common subset of instructions supported by both. An MQ
12156 register is included in processors supporting the POWER architecture.
12158 You use these options to specify which instructions are available on the
12159 processor you are using. The default value of these options is
12160 determined when configuring GCC@. Specifying the
12161 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12162 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12163 rather than the options listed above.
12165 The @option{-mpower} option allows GCC to generate instructions that
12166 are found only in the POWER architecture and to use the MQ register.
12167 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12168 to generate instructions that are present in the POWER2 architecture but
12169 not the original POWER architecture.
12171 The @option{-mpowerpc} option allows GCC to generate instructions that
12172 are found only in the 32-bit subset of the PowerPC architecture.
12173 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12174 GCC to use the optional PowerPC architecture instructions in the
12175 General Purpose group, including floating-point square root. Specifying
12176 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12177 use the optional PowerPC architecture instructions in the Graphics
12178 group, including floating-point select.
12180 The @option{-mmfcrf} option allows GCC to generate the move from
12181 condition register field instruction implemented on the POWER4
12182 processor and other processors that support the PowerPC V2.01
12184 The @option{-mpopcntb} option allows GCC to generate the popcount and
12185 double precision FP reciprocal estimate instruction implemented on the
12186 POWER5 processor and other processors that support the PowerPC V2.02
12188 The @option{-mfprnd} option allows GCC to generate the FP round to
12189 integer instructions implemented on the POWER5+ processor and other
12190 processors that support the PowerPC V2.03 architecture.
12191 The @option{-mcmpb} option allows GCC to generate the compare bytes
12192 instruction implemented on the POWER6 processor and other processors
12193 that support the PowerPC V2.05 architecture.
12194 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12195 general purpose register instructions implemented on the POWER6X
12196 processor and other processors that support the extended PowerPC V2.05
12198 The @option{-mdfp} option allows GCC to generate the decimal floating
12199 point instructions implemented on some POWER processors.
12201 The @option{-mpowerpc64} option allows GCC to generate the additional
12202 64-bit instructions that are found in the full PowerPC64 architecture
12203 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12204 @option{-mno-powerpc64}.
12206 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12207 will use only the instructions in the common subset of both
12208 architectures plus some special AIX common-mode calls, and will not use
12209 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12210 permits GCC to use any instruction from either architecture and to
12211 allow use of the MQ register; specify this for the Motorola MPC601.
12213 @item -mnew-mnemonics
12214 @itemx -mold-mnemonics
12215 @opindex mnew-mnemonics
12216 @opindex mold-mnemonics
12217 Select which mnemonics to use in the generated assembler code. With
12218 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12219 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12220 assembler mnemonics defined for the POWER architecture. Instructions
12221 defined in only one architecture have only one mnemonic; GCC uses that
12222 mnemonic irrespective of which of these options is specified.
12224 GCC defaults to the mnemonics appropriate for the architecture in
12225 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
12226 value of these option. Unless you are building a cross-compiler, you
12227 should normally not specify either @option{-mnew-mnemonics} or
12228 @option{-mold-mnemonics}, but should instead accept the default.
12230 @item -mcpu=@var{cpu_type}
12232 Set architecture type, register usage, choice of mnemonics, and
12233 instruction scheduling parameters for machine type @var{cpu_type}.
12234 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
12235 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
12236 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
12237 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
12238 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
12239 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
12240 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
12241 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
12242 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
12243 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
12245 @option{-mcpu=common} selects a completely generic processor. Code
12246 generated under this option will run on any POWER or PowerPC processor.
12247 GCC will use only the instructions in the common subset of both
12248 architectures, and will not use the MQ register. GCC assumes a generic
12249 processor model for scheduling purposes.
12251 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
12252 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
12253 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
12254 types, with an appropriate, generic processor model assumed for
12255 scheduling purposes.
12257 The other options specify a specific processor. Code generated under
12258 those options will run best on that processor, and may not run at all on
12261 The @option{-mcpu} options automatically enable or disable the
12264 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
12265 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
12266 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
12268 The particular options set for any particular CPU will vary between
12269 compiler versions, depending on what setting seems to produce optimal
12270 code for that CPU; it doesn't necessarily reflect the actual hardware's
12271 capabilities. If you wish to set an individual option to a particular
12272 value, you may specify it after the @option{-mcpu} option, like
12273 @samp{-mcpu=970 -mno-altivec}.
12275 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
12276 not enabled or disabled by the @option{-mcpu} option at present because
12277 AIX does not have full support for these options. You may still
12278 enable or disable them individually if you're sure it'll work in your
12281 @item -mtune=@var{cpu_type}
12283 Set the instruction scheduling parameters for machine type
12284 @var{cpu_type}, but do not set the architecture type, register usage, or
12285 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
12286 values for @var{cpu_type} are used for @option{-mtune} as for
12287 @option{-mcpu}. If both are specified, the code generated will use the
12288 architecture, registers, and mnemonics set by @option{-mcpu}, but the
12289 scheduling parameters set by @option{-mtune}.
12295 Generate code to compute division as reciprocal estimate and iterative
12296 refinement, creating opportunities for increased throughput. This
12297 feature requires: optional PowerPC Graphics instruction set for single
12298 precision and FRE instruction for double precision, assuming divides
12299 cannot generate user-visible traps, and the domain values not include
12300 Infinities, denormals or zero denominator.
12303 @itemx -mno-altivec
12305 @opindex mno-altivec
12306 Generate code that uses (does not use) AltiVec instructions, and also
12307 enable the use of built-in functions that allow more direct access to
12308 the AltiVec instruction set. You may also need to set
12309 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
12315 @opindex mno-vrsave
12316 Generate VRSAVE instructions when generating AltiVec code.
12319 @opindex msecure-plt
12320 Generate code that allows ld and ld.so to build executables and shared
12321 libraries with non-exec .plt and .got sections. This is a PowerPC
12322 32-bit SYSV ABI option.
12326 Generate code that uses a BSS .plt section that ld.so fills in, and
12327 requires .plt and .got sections that are both writable and executable.
12328 This is a PowerPC 32-bit SYSV ABI option.
12334 This switch enables or disables the generation of ISEL instructions.
12336 @item -misel=@var{yes/no}
12337 This switch has been deprecated. Use @option{-misel} and
12338 @option{-mno-isel} instead.
12344 This switch enables or disables the generation of SPE simd
12347 @item -mspe=@var{yes/no}
12348 This option has been deprecated. Use @option{-mspe} and
12349 @option{-mno-spe} instead.
12351 @item -mfloat-gprs=@var{yes/single/double/no}
12352 @itemx -mfloat-gprs
12353 @opindex mfloat-gprs
12354 This switch enables or disables the generation of floating point
12355 operations on the general purpose registers for architectures that
12358 The argument @var{yes} or @var{single} enables the use of
12359 single-precision floating point operations.
12361 The argument @var{double} enables the use of single and
12362 double-precision floating point operations.
12364 The argument @var{no} disables floating point operations on the
12365 general purpose registers.
12367 This option is currently only available on the MPC854x.
12373 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12374 targets (including GNU/Linux). The 32-bit environment sets int, long
12375 and pointer to 32 bits and generates code that runs on any PowerPC
12376 variant. The 64-bit environment sets int to 32 bits and long and
12377 pointer to 64 bits, and generates code for PowerPC64, as for
12378 @option{-mpowerpc64}.
12381 @itemx -mno-fp-in-toc
12382 @itemx -mno-sum-in-toc
12383 @itemx -mminimal-toc
12385 @opindex mno-fp-in-toc
12386 @opindex mno-sum-in-toc
12387 @opindex mminimal-toc
12388 Modify generation of the TOC (Table Of Contents), which is created for
12389 every executable file. The @option{-mfull-toc} option is selected by
12390 default. In that case, GCC will allocate at least one TOC entry for
12391 each unique non-automatic variable reference in your program. GCC
12392 will also place floating-point constants in the TOC@. However, only
12393 16,384 entries are available in the TOC@.
12395 If you receive a linker error message that saying you have overflowed
12396 the available TOC space, you can reduce the amount of TOC space used
12397 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
12398 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
12399 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
12400 generate code to calculate the sum of an address and a constant at
12401 run-time instead of putting that sum into the TOC@. You may specify one
12402 or both of these options. Each causes GCC to produce very slightly
12403 slower and larger code at the expense of conserving TOC space.
12405 If you still run out of space in the TOC even when you specify both of
12406 these options, specify @option{-mminimal-toc} instead. This option causes
12407 GCC to make only one TOC entry for every file. When you specify this
12408 option, GCC will produce code that is slower and larger but which
12409 uses extremely little TOC space. You may wish to use this option
12410 only on files that contain less frequently executed code.
12416 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
12417 @code{long} type, and the infrastructure needed to support them.
12418 Specifying @option{-maix64} implies @option{-mpowerpc64} and
12419 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
12420 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
12423 @itemx -mno-xl-compat
12424 @opindex mxl-compat
12425 @opindex mno-xl-compat
12426 Produce code that conforms more closely to IBM XL compiler semantics
12427 when using AIX-compatible ABI. Pass floating-point arguments to
12428 prototyped functions beyond the register save area (RSA) on the stack
12429 in addition to argument FPRs. Do not assume that most significant
12430 double in 128-bit long double value is properly rounded when comparing
12431 values and converting to double. Use XL symbol names for long double
12434 The AIX calling convention was extended but not initially documented to
12435 handle an obscure K&R C case of calling a function that takes the
12436 address of its arguments with fewer arguments than declared. IBM XL
12437 compilers access floating point arguments which do not fit in the
12438 RSA from the stack when a subroutine is compiled without
12439 optimization. Because always storing floating-point arguments on the
12440 stack is inefficient and rarely needed, this option is not enabled by
12441 default and only is necessary when calling subroutines compiled by IBM
12442 XL compilers without optimization.
12446 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
12447 application written to use message passing with special startup code to
12448 enable the application to run. The system must have PE installed in the
12449 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
12450 must be overridden with the @option{-specs=} option to specify the
12451 appropriate directory location. The Parallel Environment does not
12452 support threads, so the @option{-mpe} option and the @option{-pthread}
12453 option are incompatible.
12455 @item -malign-natural
12456 @itemx -malign-power
12457 @opindex malign-natural
12458 @opindex malign-power
12459 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12460 @option{-malign-natural} overrides the ABI-defined alignment of larger
12461 types, such as floating-point doubles, on their natural size-based boundary.
12462 The option @option{-malign-power} instructs GCC to follow the ABI-specified
12463 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
12465 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
12469 @itemx -mhard-float
12470 @opindex msoft-float
12471 @opindex mhard-float
12472 Generate code that does not use (uses) the floating-point register set.
12473 Software floating point emulation is provided if you use the
12474 @option{-msoft-float} option, and pass the option to GCC when linking.
12477 @itemx -mno-multiple
12479 @opindex mno-multiple
12480 Generate code that uses (does not use) the load multiple word
12481 instructions and the store multiple word instructions. These
12482 instructions are generated by default on POWER systems, and not
12483 generated on PowerPC systems. Do not use @option{-mmultiple} on little
12484 endian PowerPC systems, since those instructions do not work when the
12485 processor is in little endian mode. The exceptions are PPC740 and
12486 PPC750 which permit the instructions usage in little endian mode.
12491 @opindex mno-string
12492 Generate code that uses (does not use) the load string instructions
12493 and the store string word instructions to save multiple registers and
12494 do small block moves. These instructions are generated by default on
12495 POWER systems, and not generated on PowerPC systems. Do not use
12496 @option{-mstring} on little endian PowerPC systems, since those
12497 instructions do not work when the processor is in little endian mode.
12498 The exceptions are PPC740 and PPC750 which permit the instructions
12499 usage in little endian mode.
12504 @opindex mno-update
12505 Generate code that uses (does not use) the load or store instructions
12506 that update the base register to the address of the calculated memory
12507 location. These instructions are generated by default. If you use
12508 @option{-mno-update}, there is a small window between the time that the
12509 stack pointer is updated and the address of the previous frame is
12510 stored, which means code that walks the stack frame across interrupts or
12511 signals may get corrupted data.
12514 @itemx -mno-fused-madd
12515 @opindex mfused-madd
12516 @opindex mno-fused-madd
12517 Generate code that uses (does not use) the floating point multiply and
12518 accumulate instructions. These instructions are generated by default if
12519 hardware floating is used.
12525 Generate code that uses (does not use) the half-word multiply and
12526 multiply-accumulate instructions on the IBM 405 and 440 processors.
12527 These instructions are generated by default when targetting those
12534 Generate code that uses (does not use) the string-search @samp{dlmzb}
12535 instruction on the IBM 405 and 440 processors. This instruction is
12536 generated by default when targetting those processors.
12538 @item -mno-bit-align
12540 @opindex mno-bit-align
12541 @opindex mbit-align
12542 On System V.4 and embedded PowerPC systems do not (do) force structures
12543 and unions that contain bit-fields to be aligned to the base type of the
12546 For example, by default a structure containing nothing but 8
12547 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
12548 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
12549 the structure would be aligned to a 1 byte boundary and be one byte in
12552 @item -mno-strict-align
12553 @itemx -mstrict-align
12554 @opindex mno-strict-align
12555 @opindex mstrict-align
12556 On System V.4 and embedded PowerPC systems do not (do) assume that
12557 unaligned memory references will be handled by the system.
12559 @item -mrelocatable
12560 @itemx -mno-relocatable
12561 @opindex mrelocatable
12562 @opindex mno-relocatable
12563 On embedded PowerPC systems generate code that allows (does not allow)
12564 the program to be relocated to a different address at runtime. If you
12565 use @option{-mrelocatable} on any module, all objects linked together must
12566 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
12568 @item -mrelocatable-lib
12569 @itemx -mno-relocatable-lib
12570 @opindex mrelocatable-lib
12571 @opindex mno-relocatable-lib
12572 On embedded PowerPC systems generate code that allows (does not allow)
12573 the program to be relocated to a different address at runtime. Modules
12574 compiled with @option{-mrelocatable-lib} can be linked with either modules
12575 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
12576 with modules compiled with the @option{-mrelocatable} options.
12582 On System V.4 and embedded PowerPC systems do not (do) assume that
12583 register 2 contains a pointer to a global area pointing to the addresses
12584 used in the program.
12587 @itemx -mlittle-endian
12589 @opindex mlittle-endian
12590 On System V.4 and embedded PowerPC systems compile code for the
12591 processor in little endian mode. The @option{-mlittle-endian} option is
12592 the same as @option{-mlittle}.
12595 @itemx -mbig-endian
12597 @opindex mbig-endian
12598 On System V.4 and embedded PowerPC systems compile code for the
12599 processor in big endian mode. The @option{-mbig-endian} option is
12600 the same as @option{-mbig}.
12602 @item -mdynamic-no-pic
12603 @opindex mdynamic-no-pic
12604 On Darwin and Mac OS X systems, compile code so that it is not
12605 relocatable, but that its external references are relocatable. The
12606 resulting code is suitable for applications, but not shared
12609 @item -mprioritize-restricted-insns=@var{priority}
12610 @opindex mprioritize-restricted-insns
12611 This option controls the priority that is assigned to
12612 dispatch-slot restricted instructions during the second scheduling
12613 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
12614 @var{no/highest/second-highest} priority to dispatch slot restricted
12617 @item -msched-costly-dep=@var{dependence_type}
12618 @opindex msched-costly-dep
12619 This option controls which dependences are considered costly
12620 by the target during instruction scheduling. The argument
12621 @var{dependence_type} takes one of the following values:
12622 @var{no}: no dependence is costly,
12623 @var{all}: all dependences are costly,
12624 @var{true_store_to_load}: a true dependence from store to load is costly,
12625 @var{store_to_load}: any dependence from store to load is costly,
12626 @var{number}: any dependence which latency >= @var{number} is costly.
12628 @item -minsert-sched-nops=@var{scheme}
12629 @opindex minsert-sched-nops
12630 This option controls which nop insertion scheme will be used during
12631 the second scheduling pass. The argument @var{scheme} takes one of the
12633 @var{no}: Don't insert nops.
12634 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
12635 according to the scheduler's grouping.
12636 @var{regroup_exact}: Insert nops to force costly dependent insns into
12637 separate groups. Insert exactly as many nops as needed to force an insn
12638 to a new group, according to the estimated processor grouping.
12639 @var{number}: Insert nops to force costly dependent insns into
12640 separate groups. Insert @var{number} nops to force an insn to a new group.
12643 @opindex mcall-sysv
12644 On System V.4 and embedded PowerPC systems compile code using calling
12645 conventions that adheres to the March 1995 draft of the System V
12646 Application Binary Interface, PowerPC processor supplement. This is the
12647 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
12649 @item -mcall-sysv-eabi
12650 @opindex mcall-sysv-eabi
12651 Specify both @option{-mcall-sysv} and @option{-meabi} options.
12653 @item -mcall-sysv-noeabi
12654 @opindex mcall-sysv-noeabi
12655 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
12657 @item -mcall-solaris
12658 @opindex mcall-solaris
12659 On System V.4 and embedded PowerPC systems compile code for the Solaris
12663 @opindex mcall-linux
12664 On System V.4 and embedded PowerPC systems compile code for the
12665 Linux-based GNU system.
12669 On System V.4 and embedded PowerPC systems compile code for the
12670 Hurd-based GNU system.
12672 @item -mcall-netbsd
12673 @opindex mcall-netbsd
12674 On System V.4 and embedded PowerPC systems compile code for the
12675 NetBSD operating system.
12677 @item -maix-struct-return
12678 @opindex maix-struct-return
12679 Return all structures in memory (as specified by the AIX ABI)@.
12681 @item -msvr4-struct-return
12682 @opindex msvr4-struct-return
12683 Return structures smaller than 8 bytes in registers (as specified by the
12686 @item -mabi=@var{abi-type}
12688 Extend the current ABI with a particular extension, or remove such extension.
12689 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
12690 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
12694 Extend the current ABI with SPE ABI extensions. This does not change
12695 the default ABI, instead it adds the SPE ABI extensions to the current
12699 @opindex mabi=no-spe
12700 Disable Booke SPE ABI extensions for the current ABI@.
12702 @item -mabi=ibmlongdouble
12703 @opindex mabi=ibmlongdouble
12704 Change the current ABI to use IBM extended precision long double.
12705 This is a PowerPC 32-bit SYSV ABI option.
12707 @item -mabi=ieeelongdouble
12708 @opindex mabi=ieeelongdouble
12709 Change the current ABI to use IEEE extended precision long double.
12710 This is a PowerPC 32-bit Linux ABI option.
12713 @itemx -mno-prototype
12714 @opindex mprototype
12715 @opindex mno-prototype
12716 On System V.4 and embedded PowerPC systems assume that all calls to
12717 variable argument functions are properly prototyped. Otherwise, the
12718 compiler must insert an instruction before every non prototyped call to
12719 set or clear bit 6 of the condition code register (@var{CR}) to
12720 indicate whether floating point values were passed in the floating point
12721 registers in case the function takes a variable arguments. With
12722 @option{-mprototype}, only calls to prototyped variable argument functions
12723 will set or clear the bit.
12727 On embedded PowerPC systems, assume that the startup module is called
12728 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
12729 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
12734 On embedded PowerPC systems, assume that the startup module is called
12735 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
12740 On embedded PowerPC systems, assume that the startup module is called
12741 @file{crt0.o} and the standard C libraries are @file{libads.a} and
12744 @item -myellowknife
12745 @opindex myellowknife
12746 On embedded PowerPC systems, assume that the startup module is called
12747 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
12752 On System V.4 and embedded PowerPC systems, specify that you are
12753 compiling for a VxWorks system.
12757 Specify that you are compiling for the WindISS simulation environment.
12761 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
12762 header to indicate that @samp{eabi} extended relocations are used.
12768 On System V.4 and embedded PowerPC systems do (do not) adhere to the
12769 Embedded Applications Binary Interface (eabi) which is a set of
12770 modifications to the System V.4 specifications. Selecting @option{-meabi}
12771 means that the stack is aligned to an 8 byte boundary, a function
12772 @code{__eabi} is called to from @code{main} to set up the eabi
12773 environment, and the @option{-msdata} option can use both @code{r2} and
12774 @code{r13} to point to two separate small data areas. Selecting
12775 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
12776 do not call an initialization function from @code{main}, and the
12777 @option{-msdata} option will only use @code{r13} to point to a single
12778 small data area. The @option{-meabi} option is on by default if you
12779 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
12782 @opindex msdata=eabi
12783 On System V.4 and embedded PowerPC systems, put small initialized
12784 @code{const} global and static data in the @samp{.sdata2} section, which
12785 is pointed to by register @code{r2}. Put small initialized
12786 non-@code{const} global and static data in the @samp{.sdata} section,
12787 which is pointed to by register @code{r13}. Put small uninitialized
12788 global and static data in the @samp{.sbss} section, which is adjacent to
12789 the @samp{.sdata} section. The @option{-msdata=eabi} option is
12790 incompatible with the @option{-mrelocatable} option. The
12791 @option{-msdata=eabi} option also sets the @option{-memb} option.
12794 @opindex msdata=sysv
12795 On System V.4 and embedded PowerPC systems, put small global and static
12796 data in the @samp{.sdata} section, which is pointed to by register
12797 @code{r13}. Put small uninitialized global and static data in the
12798 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
12799 The @option{-msdata=sysv} option is incompatible with the
12800 @option{-mrelocatable} option.
12802 @item -msdata=default
12804 @opindex msdata=default
12806 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12807 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12808 same as @option{-msdata=sysv}.
12811 @opindex msdata-data
12812 On System V.4 and embedded PowerPC systems, put small global
12813 data in the @samp{.sdata} section. Put small uninitialized global
12814 data in the @samp{.sbss} section. Do not use register @code{r13}
12815 to address small data however. This is the default behavior unless
12816 other @option{-msdata} options are used.
12820 @opindex msdata=none
12822 On embedded PowerPC systems, put all initialized global and static data
12823 in the @samp{.data} section, and all uninitialized data in the
12824 @samp{.bss} section.
12828 @cindex smaller data references (PowerPC)
12829 @cindex .sdata/.sdata2 references (PowerPC)
12830 On embedded PowerPC systems, put global and static items less than or
12831 equal to @var{num} bytes into the small data or bss sections instead of
12832 the normal data or bss section. By default, @var{num} is 8. The
12833 @option{-G @var{num}} switch is also passed to the linker.
12834 All modules should be compiled with the same @option{-G @var{num}} value.
12837 @itemx -mno-regnames
12839 @opindex mno-regnames
12840 On System V.4 and embedded PowerPC systems do (do not) emit register
12841 names in the assembly language output using symbolic forms.
12844 @itemx -mno-longcall
12846 @opindex mno-longcall
12847 By default assume that all calls are far away so that a longer more
12848 expensive calling sequence is required. This is required for calls
12849 further than 32 megabytes (33,554,432 bytes) from the current location.
12850 A short call will be generated if the compiler knows
12851 the call cannot be that far away. This setting can be overridden by
12852 the @code{shortcall} function attribute, or by @code{#pragma
12855 Some linkers are capable of detecting out-of-range calls and generating
12856 glue code on the fly. On these systems, long calls are unnecessary and
12857 generate slower code. As of this writing, the AIX linker can do this,
12858 as can the GNU linker for PowerPC/64. It is planned to add this feature
12859 to the GNU linker for 32-bit PowerPC systems as well.
12861 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
12862 callee, L42'', plus a ``branch island'' (glue code). The two target
12863 addresses represent the callee and the ``branch island''. The
12864 Darwin/PPC linker will prefer the first address and generate a ``bl
12865 callee'' if the PPC ``bl'' instruction will reach the callee directly;
12866 otherwise, the linker will generate ``bl L42'' to call the ``branch
12867 island''. The ``branch island'' is appended to the body of the
12868 calling function; it computes the full 32-bit address of the callee
12871 On Mach-O (Darwin) systems, this option directs the compiler emit to
12872 the glue for every direct call, and the Darwin linker decides whether
12873 to use or discard it.
12875 In the future, we may cause GCC to ignore all longcall specifications
12876 when the linker is known to generate glue.
12880 Adds support for multithreading with the @dfn{pthreads} library.
12881 This option sets flags for both the preprocessor and linker.
12885 @node S/390 and zSeries Options
12886 @subsection S/390 and zSeries Options
12887 @cindex S/390 and zSeries Options
12889 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
12893 @itemx -msoft-float
12894 @opindex mhard-float
12895 @opindex msoft-float
12896 Use (do not use) the hardware floating-point instructions and registers
12897 for floating-point operations. When @option{-msoft-float} is specified,
12898 functions in @file{libgcc.a} will be used to perform floating-point
12899 operations. When @option{-mhard-float} is specified, the compiler
12900 generates IEEE floating-point instructions. This is the default.
12902 @item -mlong-double-64
12903 @itemx -mlong-double-128
12904 @opindex mlong-double-64
12905 @opindex mlong-double-128
12906 These switches control the size of @code{long double} type. A size
12907 of 64bit makes the @code{long double} type equivalent to the @code{double}
12908 type. This is the default.
12911 @itemx -mno-backchain
12912 @opindex mbackchain
12913 @opindex mno-backchain
12914 Store (do not store) the address of the caller's frame as backchain pointer
12915 into the callee's stack frame.
12916 A backchain may be needed to allow debugging using tools that do not understand
12917 DWARF-2 call frame information.
12918 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12919 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12920 the backchain is placed into the topmost word of the 96/160 byte register
12923 In general, code compiled with @option{-mbackchain} is call-compatible with
12924 code compiled with @option{-mmo-backchain}; however, use of the backchain
12925 for debugging purposes usually requires that the whole binary is built with
12926 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12927 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12928 to build a linux kernel use @option{-msoft-float}.
12930 The default is to not maintain the backchain.
12932 @item -mpacked-stack
12933 @item -mno-packed-stack
12934 @opindex mpacked-stack
12935 @opindex mno-packed-stack
12936 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12937 specified, the compiler uses the all fields of the 96/160 byte register save
12938 area only for their default purpose; unused fields still take up stack space.
12939 When @option{-mpacked-stack} is specified, register save slots are densely
12940 packed at the top of the register save area; unused space is reused for other
12941 purposes, allowing for more efficient use of the available stack space.
12942 However, when @option{-mbackchain} is also in effect, the topmost word of
12943 the save area is always used to store the backchain, and the return address
12944 register is always saved two words below the backchain.
12946 As long as the stack frame backchain is not used, code generated with
12947 @option{-mpacked-stack} is call-compatible with code generated with
12948 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12949 S/390 or zSeries generated code that uses the stack frame backchain at run
12950 time, not just for debugging purposes. Such code is not call-compatible
12951 with code compiled with @option{-mpacked-stack}. Also, note that the
12952 combination of @option{-mbackchain},
12953 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12954 to build a linux kernel use @option{-msoft-float}.
12956 The default is to not use the packed stack layout.
12959 @itemx -mno-small-exec
12960 @opindex msmall-exec
12961 @opindex mno-small-exec
12962 Generate (or do not generate) code using the @code{bras} instruction
12963 to do subroutine calls.
12964 This only works reliably if the total executable size does not
12965 exceed 64k. The default is to use the @code{basr} instruction instead,
12966 which does not have this limitation.
12972 When @option{-m31} is specified, generate code compliant to the
12973 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12974 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12975 particular to generate 64-bit instructions. For the @samp{s390}
12976 targets, the default is @option{-m31}, while the @samp{s390x}
12977 targets default to @option{-m64}.
12983 When @option{-mzarch} is specified, generate code using the
12984 instructions available on z/Architecture.
12985 When @option{-mesa} is specified, generate code using the
12986 instructions available on ESA/390. Note that @option{-mesa} is
12987 not possible with @option{-m64}.
12988 When generating code compliant to the GNU/Linux for S/390 ABI,
12989 the default is @option{-mesa}. When generating code compliant
12990 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12996 Generate (or do not generate) code using the @code{mvcle} instruction
12997 to perform block moves. When @option{-mno-mvcle} is specified,
12998 use a @code{mvc} loop instead. This is the default unless optimizing for
13005 Print (or do not print) additional debug information when compiling.
13006 The default is to not print debug information.
13008 @item -march=@var{cpu-type}
13010 Generate code that will run on @var{cpu-type}, which is the name of a system
13011 representing a certain processor type. Possible values for
13012 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13013 When generating code using the instructions available on z/Architecture,
13014 the default is @option{-march=z900}. Otherwise, the default is
13015 @option{-march=g5}.
13017 @item -mtune=@var{cpu-type}
13019 Tune to @var{cpu-type} everything applicable about the generated code,
13020 except for the ABI and the set of available instructions.
13021 The list of @var{cpu-type} values is the same as for @option{-march}.
13022 The default is the value used for @option{-march}.
13025 @itemx -mno-tpf-trace
13026 @opindex mtpf-trace
13027 @opindex mno-tpf-trace
13028 Generate code that adds (does not add) in TPF OS specific branches to trace
13029 routines in the operating system. This option is off by default, even
13030 when compiling for the TPF OS@.
13033 @itemx -mno-fused-madd
13034 @opindex mfused-madd
13035 @opindex mno-fused-madd
13036 Generate code that uses (does not use) the floating point multiply and
13037 accumulate instructions. These instructions are generated by default if
13038 hardware floating point is used.
13040 @item -mwarn-framesize=@var{framesize}
13041 @opindex mwarn-framesize
13042 Emit a warning if the current function exceeds the given frame size. Because
13043 this is a compile time check it doesn't need to be a real problem when the program
13044 runs. It is intended to identify functions which most probably cause
13045 a stack overflow. It is useful to be used in an environment with limited stack
13046 size e.g.@: the linux kernel.
13048 @item -mwarn-dynamicstack
13049 @opindex mwarn-dynamicstack
13050 Emit a warning if the function calls alloca or uses dynamically
13051 sized arrays. This is generally a bad idea with a limited stack size.
13053 @item -mstack-guard=@var{stack-guard}
13054 @item -mstack-size=@var{stack-size}
13055 @opindex mstack-guard
13056 @opindex mstack-size
13057 If these options are provided the s390 back end emits additional instructions in
13058 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13059 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13060 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13061 the frame size of the compiled function is chosen.
13062 These options are intended to be used to help debugging stack overflow problems.
13063 The additionally emitted code causes only little overhead and hence can also be
13064 used in production like systems without greater performance degradation. The given
13065 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13066 @var{stack-guard} without exceeding 64k.
13067 In order to be efficient the extra code makes the assumption that the stack starts
13068 at an address aligned to the value given by @var{stack-size}.
13069 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13072 @node Score Options
13073 @subsection Score Options
13074 @cindex Score Options
13076 These options are defined for Score implementations:
13081 Compile code for big endian mode. This is the default.
13085 Compile code for little endian mode.
13089 Disable generate bcnz instruction.
13093 Enable generate unaligned load and store instruction.
13097 Enable the use of multiply-accumulate instructions. Disabled by default.
13101 Specify the SCORE5 as the target architecture.
13105 Specify the SCORE5U of the target architecture.
13109 Specify the SCORE7 as the target architecture. This is the default.
13113 Specify the SCORE7D as the target architecture.
13117 @subsection SH Options
13119 These @samp{-m} options are defined for the SH implementations:
13124 Generate code for the SH1.
13128 Generate code for the SH2.
13131 Generate code for the SH2e.
13135 Generate code for the SH3.
13139 Generate code for the SH3e.
13143 Generate code for the SH4 without a floating-point unit.
13145 @item -m4-single-only
13146 @opindex m4-single-only
13147 Generate code for the SH4 with a floating-point unit that only
13148 supports single-precision arithmetic.
13152 Generate code for the SH4 assuming the floating-point unit is in
13153 single-precision mode by default.
13157 Generate code for the SH4.
13161 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13162 floating-point unit is not used.
13164 @item -m4a-single-only
13165 @opindex m4a-single-only
13166 Generate code for the SH4a, in such a way that no double-precision
13167 floating point operations are used.
13170 @opindex m4a-single
13171 Generate code for the SH4a assuming the floating-point unit is in
13172 single-precision mode by default.
13176 Generate code for the SH4a.
13180 Same as @option{-m4a-nofpu}, except that it implicitly passes
13181 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13182 instructions at the moment.
13186 Compile code for the processor in big endian mode.
13190 Compile code for the processor in little endian mode.
13194 Align doubles at 64-bit boundaries. Note that this changes the calling
13195 conventions, and thus some functions from the standard C library will
13196 not work unless you recompile it first with @option{-mdalign}.
13200 Shorten some address references at link time, when possible; uses the
13201 linker option @option{-relax}.
13205 Use 32-bit offsets in @code{switch} tables. The default is to use
13210 Enable the use of the instruction @code{fmovd}.
13214 Comply with the calling conventions defined by Renesas.
13218 Comply with the calling conventions defined by Renesas.
13222 Comply with the calling conventions defined for GCC before the Renesas
13223 conventions were available. This option is the default for all
13224 targets of the SH toolchain except for @samp{sh-symbianelf}.
13227 @opindex mnomacsave
13228 Mark the @code{MAC} register as call-clobbered, even if
13229 @option{-mhitachi} is given.
13233 Increase IEEE-compliance of floating-point code.
13234 At the moment, this is equivalent to @option{-fno-finite-math-only}.
13235 When generating 16 bit SH opcodes, getting IEEE-conforming results for
13236 comparisons of NANs / infinities incurs extra overhead in every
13237 floating point comparison, therefore the default is set to
13238 @option{-ffinite-math-only}.
13240 @item -minline-ic_invalidate
13241 @opindex minline-ic_invalidate
13242 Inline code to invalidate instruction cache entries after setting up
13243 nested function trampolines.
13244 This option has no effect if -musermode is in effect and the selected
13245 code generation option (e.g. -m4) does not allow the use of the icbi
13247 If the selected code generation option does not allow the use of the icbi
13248 instruction, and -musermode is not in effect, the inlined code will
13249 manipulate the instruction cache address array directly with an associative
13250 write. This not only requires privileged mode, but it will also
13251 fail if the cache line had been mapped via the TLB and has become unmapped.
13255 Dump instruction size and location in the assembly code.
13258 @opindex mpadstruct
13259 This option is deprecated. It pads structures to multiple of 4 bytes,
13260 which is incompatible with the SH ABI@.
13264 Optimize for space instead of speed. Implied by @option{-Os}.
13267 @opindex mprefergot
13268 When generating position-independent code, emit function calls using
13269 the Global Offset Table instead of the Procedure Linkage Table.
13273 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
13274 if the inlined code would not work in user mode.
13275 This is the default when the target is @code{sh-*-linux*}.
13277 @item -multcost=@var{number}
13278 @opindex multcost=@var{number}
13279 Set the cost to assume for a multiply insn.
13281 @item -mdiv=@var{strategy}
13282 @opindex mdiv=@var{strategy}
13283 Set the division strategy to use for SHmedia code. @var{strategy} must be
13284 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
13285 inv:call2, inv:fp .
13286 "fp" performs the operation in floating point. This has a very high latency,
13287 but needs only a few instructions, so it might be a good choice if
13288 your code has enough easily exploitable ILP to allow the compiler to
13289 schedule the floating point instructions together with other instructions.
13290 Division by zero causes a floating point exception.
13291 "inv" uses integer operations to calculate the inverse of the divisor,
13292 and then multiplies the dividend with the inverse. This strategy allows
13293 cse and hoisting of the inverse calculation. Division by zero calculates
13294 an unspecified result, but does not trap.
13295 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
13296 have been found, or if the entire operation has been hoisted to the same
13297 place, the last stages of the inverse calculation are intertwined with the
13298 final multiply to reduce the overall latency, at the expense of using a few
13299 more instructions, and thus offering fewer scheduling opportunities with
13301 "call" calls a library function that usually implements the inv:minlat
13303 This gives high code density for m5-*media-nofpu compilations.
13304 "call2" uses a different entry point of the same library function, where it
13305 assumes that a pointer to a lookup table has already been set up, which
13306 exposes the pointer load to cse / code hoisting optimizations.
13307 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
13308 code generation, but if the code stays unoptimized, revert to the "call",
13309 "call2", or "fp" strategies, respectively. Note that the
13310 potentially-trapping side effect of division by zero is carried by a
13311 separate instruction, so it is possible that all the integer instructions
13312 are hoisted out, but the marker for the side effect stays where it is.
13313 A recombination to fp operations or a call is not possible in that case.
13314 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
13315 that the inverse calculation was nor separated from the multiply, they speed
13316 up division where the dividend fits into 20 bits (plus sign where applicable),
13317 by inserting a test to skip a number of operations in this case; this test
13318 slows down the case of larger dividends. inv20u assumes the case of a such
13319 a small dividend to be unlikely, and inv20l assumes it to be likely.
13321 @item -mdivsi3_libfunc=@var{name}
13322 @opindex mdivsi3_libfunc=@var{name}
13323 Set the name of the library function used for 32 bit signed division to
13324 @var{name}. This only affect the name used in the call and inv:call
13325 division strategies, and the compiler will still expect the same
13326 sets of input/output/clobbered registers as if this option was not present.
13328 @item -madjust-unroll
13329 @opindex madjust-unroll
13330 Throttle unrolling to avoid thrashing target registers.
13331 This option only has an effect if the gcc code base supports the
13332 TARGET_ADJUST_UNROLL_MAX target hook.
13334 @item -mindexed-addressing
13335 @opindex mindexed-addressing
13336 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
13337 This is only safe if the hardware and/or OS implement 32 bit wrap-around
13338 semantics for the indexed addressing mode. The architecture allows the
13339 implementation of processors with 64 bit MMU, which the OS could use to
13340 get 32 bit addressing, but since no current hardware implementation supports
13341 this or any other way to make the indexed addressing mode safe to use in
13342 the 32 bit ABI, the default is -mno-indexed-addressing.
13344 @item -mgettrcost=@var{number}
13345 @opindex mgettrcost=@var{number}
13346 Set the cost assumed for the gettr instruction to @var{number}.
13347 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
13351 Assume pt* instructions won't trap. This will generally generate better
13352 scheduled code, but is unsafe on current hardware. The current architecture
13353 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
13354 This has the unintentional effect of making it unsafe to schedule ptabs /
13355 ptrel before a branch, or hoist it out of a loop. For example,
13356 __do_global_ctors, a part of libgcc that runs constructors at program
13357 startup, calls functions in a list which is delimited by @minus{}1. With the
13358 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
13359 That means that all the constructors will be run a bit quicker, but when
13360 the loop comes to the end of the list, the program crashes because ptabs
13361 loads @minus{}1 into a target register. Since this option is unsafe for any
13362 hardware implementing the current architecture specification, the default
13363 is -mno-pt-fixed. Unless the user specifies a specific cost with
13364 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
13365 this deters register allocation using target registers for storing
13368 @item -minvalid-symbols
13369 @opindex minvalid-symbols
13370 Assume symbols might be invalid. Ordinary function symbols generated by
13371 the compiler will always be valid to load with movi/shori/ptabs or
13372 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
13373 to generate symbols that will cause ptabs / ptrel to trap.
13374 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
13375 It will then prevent cross-basic-block cse, hoisting and most scheduling
13376 of symbol loads. The default is @option{-mno-invalid-symbols}.
13379 @node SPARC Options
13380 @subsection SPARC Options
13381 @cindex SPARC options
13383 These @samp{-m} options are supported on the SPARC:
13386 @item -mno-app-regs
13388 @opindex mno-app-regs
13390 Specify @option{-mapp-regs} to generate output using the global registers
13391 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
13394 To be fully SVR4 ABI compliant at the cost of some performance loss,
13395 specify @option{-mno-app-regs}. You should compile libraries and system
13396 software with this option.
13399 @itemx -mhard-float
13401 @opindex mhard-float
13402 Generate output containing floating point instructions. This is the
13406 @itemx -msoft-float
13408 @opindex msoft-float
13409 Generate output containing library calls for floating point.
13410 @strong{Warning:} the requisite libraries are not available for all SPARC
13411 targets. Normally the facilities of the machine's usual C compiler are
13412 used, but this cannot be done directly in cross-compilation. You must make
13413 your own arrangements to provide suitable library functions for
13414 cross-compilation. The embedded targets @samp{sparc-*-aout} and
13415 @samp{sparclite-*-*} do provide software floating point support.
13417 @option{-msoft-float} changes the calling convention in the output file;
13418 therefore, it is only useful if you compile @emph{all} of a program with
13419 this option. In particular, you need to compile @file{libgcc.a}, the
13420 library that comes with GCC, with @option{-msoft-float} in order for
13423 @item -mhard-quad-float
13424 @opindex mhard-quad-float
13425 Generate output containing quad-word (long double) floating point
13428 @item -msoft-quad-float
13429 @opindex msoft-quad-float
13430 Generate output containing library calls for quad-word (long double)
13431 floating point instructions. The functions called are those specified
13432 in the SPARC ABI@. This is the default.
13434 As of this writing, there are no SPARC implementations that have hardware
13435 support for the quad-word floating point instructions. They all invoke
13436 a trap handler for one of these instructions, and then the trap handler
13437 emulates the effect of the instruction. Because of the trap handler overhead,
13438 this is much slower than calling the ABI library routines. Thus the
13439 @option{-msoft-quad-float} option is the default.
13441 @item -mno-unaligned-doubles
13442 @itemx -munaligned-doubles
13443 @opindex mno-unaligned-doubles
13444 @opindex munaligned-doubles
13445 Assume that doubles have 8 byte alignment. This is the default.
13447 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
13448 alignment only if they are contained in another type, or if they have an
13449 absolute address. Otherwise, it assumes they have 4 byte alignment.
13450 Specifying this option avoids some rare compatibility problems with code
13451 generated by other compilers. It is not the default because it results
13452 in a performance loss, especially for floating point code.
13454 @item -mno-faster-structs
13455 @itemx -mfaster-structs
13456 @opindex mno-faster-structs
13457 @opindex mfaster-structs
13458 With @option{-mfaster-structs}, the compiler assumes that structures
13459 should have 8 byte alignment. This enables the use of pairs of
13460 @code{ldd} and @code{std} instructions for copies in structure
13461 assignment, in place of twice as many @code{ld} and @code{st} pairs.
13462 However, the use of this changed alignment directly violates the SPARC
13463 ABI@. Thus, it's intended only for use on targets where the developer
13464 acknowledges that their resulting code will not be directly in line with
13465 the rules of the ABI@.
13467 @item -mimpure-text
13468 @opindex mimpure-text
13469 @option{-mimpure-text}, used in addition to @option{-shared}, tells
13470 the compiler to not pass @option{-z text} to the linker when linking a
13471 shared object. Using this option, you can link position-dependent
13472 code into a shared object.
13474 @option{-mimpure-text} suppresses the ``relocations remain against
13475 allocatable but non-writable sections'' linker error message.
13476 However, the necessary relocations will trigger copy-on-write, and the
13477 shared object is not actually shared across processes. Instead of
13478 using @option{-mimpure-text}, you should compile all source code with
13479 @option{-fpic} or @option{-fPIC}.
13481 This option is only available on SunOS and Solaris.
13483 @item -mcpu=@var{cpu_type}
13485 Set the instruction set, register set, and instruction scheduling parameters
13486 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
13487 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
13488 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
13489 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
13490 @samp{ultrasparc3}, and @samp{niagara}.
13492 Default instruction scheduling parameters are used for values that select
13493 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
13494 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
13496 Here is a list of each supported architecture and their supported
13501 v8: supersparc, hypersparc
13502 sparclite: f930, f934, sparclite86x
13504 v9: ultrasparc, ultrasparc3, niagara
13507 By default (unless configured otherwise), GCC generates code for the V7
13508 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
13509 additionally optimizes it for the Cypress CY7C602 chip, as used in the
13510 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
13511 SPARCStation 1, 2, IPX etc.
13513 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
13514 architecture. The only difference from V7 code is that the compiler emits
13515 the integer multiply and integer divide instructions which exist in SPARC-V8
13516 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
13517 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
13520 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
13521 the SPARC architecture. This adds the integer multiply, integer divide step
13522 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
13523 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
13524 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
13525 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
13526 MB86934 chip, which is the more recent SPARClite with FPU@.
13528 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
13529 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
13530 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
13531 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
13532 optimizes it for the TEMIC SPARClet chip.
13534 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
13535 architecture. This adds 64-bit integer and floating-point move instructions,
13536 3 additional floating-point condition code registers and conditional move
13537 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
13538 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
13539 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
13540 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
13541 @option{-mcpu=niagara}, the compiler additionally optimizes it for
13542 Sun UltraSPARC T1 chips.
13544 @item -mtune=@var{cpu_type}
13546 Set the instruction scheduling parameters for machine type
13547 @var{cpu_type}, but do not set the instruction set or register set that the
13548 option @option{-mcpu=@var{cpu_type}} would.
13550 The same values for @option{-mcpu=@var{cpu_type}} can be used for
13551 @option{-mtune=@var{cpu_type}}, but the only useful values are those
13552 that select a particular cpu implementation. Those are @samp{cypress},
13553 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
13554 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
13555 @samp{ultrasparc3}, and @samp{niagara}.
13560 @opindex mno-v8plus
13561 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
13562 difference from the V8 ABI is that the global and out registers are
13563 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
13564 mode for all SPARC-V9 processors.
13570 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
13571 Visual Instruction Set extensions. The default is @option{-mno-vis}.
13574 These @samp{-m} options are supported in addition to the above
13575 on SPARC-V9 processors in 64-bit environments:
13578 @item -mlittle-endian
13579 @opindex mlittle-endian
13580 Generate code for a processor running in little-endian mode. It is only
13581 available for a few configurations and most notably not on Solaris and Linux.
13587 Generate code for a 32-bit or 64-bit environment.
13588 The 32-bit environment sets int, long and pointer to 32 bits.
13589 The 64-bit environment sets int to 32 bits and long and pointer
13592 @item -mcmodel=medlow
13593 @opindex mcmodel=medlow
13594 Generate code for the Medium/Low code model: 64-bit addresses, programs
13595 must be linked in the low 32 bits of memory. Programs can be statically
13596 or dynamically linked.
13598 @item -mcmodel=medmid
13599 @opindex mcmodel=medmid
13600 Generate code for the Medium/Middle code model: 64-bit addresses, programs
13601 must be linked in the low 44 bits of memory, the text and data segments must
13602 be less than 2GB in size and the data segment must be located within 2GB of
13605 @item -mcmodel=medany
13606 @opindex mcmodel=medany
13607 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
13608 may be linked anywhere in memory, the text and data segments must be less
13609 than 2GB in size and the data segment must be located within 2GB of the
13612 @item -mcmodel=embmedany
13613 @opindex mcmodel=embmedany
13614 Generate code for the Medium/Anywhere code model for embedded systems:
13615 64-bit addresses, the text and data segments must be less than 2GB in
13616 size, both starting anywhere in memory (determined at link time). The
13617 global register %g4 points to the base of the data segment. Programs
13618 are statically linked and PIC is not supported.
13621 @itemx -mno-stack-bias
13622 @opindex mstack-bias
13623 @opindex mno-stack-bias
13624 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
13625 frame pointer if present, are offset by @minus{}2047 which must be added back
13626 when making stack frame references. This is the default in 64-bit mode.
13627 Otherwise, assume no such offset is present.
13630 These switches are supported in addition to the above on Solaris:
13635 Add support for multithreading using the Solaris threads library. This
13636 option sets flags for both the preprocessor and linker. This option does
13637 not affect the thread safety of object code produced by the compiler or
13638 that of libraries supplied with it.
13642 Add support for multithreading using the POSIX threads library. This
13643 option sets flags for both the preprocessor and linker. This option does
13644 not affect the thread safety of object code produced by the compiler or
13645 that of libraries supplied with it.
13649 This is a synonym for @option{-pthreads}.
13653 @subsection SPU Options
13654 @cindex SPU options
13656 These @samp{-m} options are supported on the SPU:
13660 @itemx -merror-reloc
13661 @opindex mwarn-reloc
13662 @opindex merror-reloc
13664 The loader for SPU does not handle dynamic relocations. By default, GCC
13665 will give an error when it generates code that requires a dynamic
13666 relocation. @option{-mno-error-reloc} disables the error,
13667 @option{-mwarn-reloc} will generate a warning instead.
13670 @itemx -munsafe-dma
13672 @opindex munsafe-dma
13674 Instructions which initiate or test completion of DMA must not be
13675 reordered with respect to loads and stores of the memory which is being
13676 accessed. Users typically address this problem using the volatile
13677 keyword, but that can lead to inefficient code in places where the
13678 memory is known to not change. Rather than mark the memory as volatile
13679 we treat the DMA instructions as potentially effecting all memory. With
13680 @option{-munsafe-dma} users must use the volatile keyword to protect
13683 @item -mbranch-hints
13684 @opindex mbranch-hints
13686 By default, GCC will generate a branch hint instruction to avoid
13687 pipeline stalls for always taken or probably taken branches. A hint
13688 will not be generated closer than 8 instructions away from its branch.
13689 There is little reason to disable them, except for debugging purposes,
13690 or to make an object a little bit smaller.
13694 @opindex msmall-mem
13695 @opindex mlarge-mem
13697 By default, GCC generates code assuming that addresses are never larger
13698 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
13699 a full 32 bit address.
13704 By default, GCC links against startup code that assumes the SPU-style
13705 main function interface (which has an unconventional parameter list).
13706 With @option{-mstdmain}, GCC will link your program against startup
13707 code that assumes a C99-style interface to @code{main}, including a
13708 local copy of @code{argv} strings.
13710 @item -mfixed-range=@var{register-range}
13711 @opindex mfixed-range
13712 Generate code treating the given register range as fixed registers.
13713 A fixed register is one that the register allocator can not use. This is
13714 useful when compiling kernel code. A register range is specified as
13715 two registers separated by a dash. Multiple register ranges can be
13716 specified separated by a comma.
13720 @node System V Options
13721 @subsection Options for System V
13723 These additional options are available on System V Release 4 for
13724 compatibility with other compilers on those systems:
13729 Create a shared object.
13730 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
13734 Identify the versions of each tool used by the compiler, in a
13735 @code{.ident} assembler directive in the output.
13739 Refrain from adding @code{.ident} directives to the output file (this is
13742 @item -YP,@var{dirs}
13744 Search the directories @var{dirs}, and no others, for libraries
13745 specified with @option{-l}.
13747 @item -Ym,@var{dir}
13749 Look in the directory @var{dir} to find the M4 preprocessor.
13750 The assembler uses this option.
13751 @c This is supposed to go with a -Yd for predefined M4 macro files, but
13752 @c the generic assembler that comes with Solaris takes just -Ym.
13755 @node TMS320C3x/C4x Options
13756 @subsection TMS320C3x/C4x Options
13757 @cindex TMS320C3x/C4x Options
13759 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
13763 @item -mcpu=@var{cpu_type}
13765 Set the instruction set, register set, and instruction scheduling
13766 parameters for machine type @var{cpu_type}. Supported values for
13767 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
13768 @samp{c44}. The default is @samp{c40} to generate code for the
13773 @itemx -msmall-memory
13775 @opindex mbig-memory
13777 @opindex msmall-memory
13779 Generates code for the big or small memory model. The small memory
13780 model assumed that all data fits into one 64K word page. At run-time
13781 the data page (DP) register must be set to point to the 64K page
13782 containing the .bss and .data program sections. The big memory model is
13783 the default and requires reloading of the DP register for every direct
13790 Allow (disallow) allocation of general integer operands into the block
13791 count register BK@.
13797 Enable (disable) generation of code using decrement and branch,
13798 DBcond(D), instructions. This is enabled by default for the C4x. To be
13799 on the safe side, this is disabled for the C3x, since the maximum
13800 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
13801 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
13802 that it can utilize the decrement and branch instruction, but will give
13803 up if there is more than one memory reference in the loop. Thus a loop
13804 where the loop counter is decremented can generate slightly more
13805 efficient code, in cases where the RPTB instruction cannot be utilized.
13807 @item -mdp-isr-reload
13809 @opindex mdp-isr-reload
13811 Force the DP register to be saved on entry to an interrupt service
13812 routine (ISR), reloaded to point to the data section, and restored on
13813 exit from the ISR@. This should not be required unless someone has
13814 violated the small memory model by modifying the DP register, say within
13821 For the C3x use the 24-bit MPYI instruction for integer multiplies
13822 instead of a library call to guarantee 32-bit results. Note that if one
13823 of the operands is a constant, then the multiplication will be performed
13824 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
13825 then squaring operations are performed inline instead of a library call.
13828 @itemx -mno-fast-fix
13830 @opindex mno-fast-fix
13831 The C3x/C4x FIX instruction to convert a floating point value to an
13832 integer value chooses the nearest integer less than or equal to the
13833 floating point value rather than to the nearest integer. Thus if the
13834 floating point number is negative, the result will be incorrectly
13835 truncated an additional code is necessary to detect and correct this
13836 case. This option can be used to disable generation of the additional
13837 code required to correct the result.
13843 Enable (disable) generation of repeat block sequences using the RPTB
13844 instruction for zero overhead looping. The RPTB construct is only used
13845 for innermost loops that do not call functions or jump across the loop
13846 boundaries. There is no advantage having nested RPTB loops due to the
13847 overhead required to save and restore the RC, RS, and RE registers.
13848 This is enabled by default with @option{-O2}.
13850 @item -mrpts=@var{count}
13854 Enable (disable) the use of the single instruction repeat instruction
13855 RPTS@. If a repeat block contains a single instruction, and the loop
13856 count can be guaranteed to be less than the value @var{count}, GCC will
13857 emit a RPTS instruction instead of a RPTB@. If no value is specified,
13858 then a RPTS will be emitted even if the loop count cannot be determined
13859 at compile time. Note that the repeated instruction following RPTS does
13860 not have to be reloaded from memory each iteration, thus freeing up the
13861 CPU buses for operands. However, since interrupts are blocked by this
13862 instruction, it is disabled by default.
13864 @item -mloop-unsigned
13865 @itemx -mno-loop-unsigned
13866 @opindex mloop-unsigned
13867 @opindex mno-loop-unsigned
13868 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
13869 is @math{2^{31} + 1} since these instructions test if the iteration count is
13870 negative to terminate the loop. If the iteration count is unsigned
13871 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
13872 exceeded. This switch allows an unsigned iteration count.
13876 Try to emit an assembler syntax that the TI assembler (asm30) is happy
13877 with. This also enforces compatibility with the API employed by the TI
13878 C3x C compiler. For example, long doubles are passed as structures
13879 rather than in floating point registers.
13885 Generate code that uses registers (stack) for passing arguments to functions.
13886 By default, arguments are passed in registers where possible rather
13887 than by pushing arguments on to the stack.
13889 @item -mparallel-insns
13890 @itemx -mno-parallel-insns
13891 @opindex mparallel-insns
13892 @opindex mno-parallel-insns
13893 Allow the generation of parallel instructions. This is enabled by
13894 default with @option{-O2}.
13896 @item -mparallel-mpy
13897 @itemx -mno-parallel-mpy
13898 @opindex mparallel-mpy
13899 @opindex mno-parallel-mpy
13900 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
13901 provided @option{-mparallel-insns} is also specified. These instructions have
13902 tight register constraints which can pessimize the code generation
13903 of large functions.
13908 @subsection V850 Options
13909 @cindex V850 Options
13911 These @samp{-m} options are defined for V850 implementations:
13915 @itemx -mno-long-calls
13916 @opindex mlong-calls
13917 @opindex mno-long-calls
13918 Treat all calls as being far away (near). If calls are assumed to be
13919 far away, the compiler will always load the functions address up into a
13920 register, and call indirect through the pointer.
13926 Do not optimize (do optimize) basic blocks that use the same index
13927 pointer 4 or more times to copy pointer into the @code{ep} register, and
13928 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
13929 option is on by default if you optimize.
13931 @item -mno-prolog-function
13932 @itemx -mprolog-function
13933 @opindex mno-prolog-function
13934 @opindex mprolog-function
13935 Do not use (do use) external functions to save and restore registers
13936 at the prologue and epilogue of a function. The external functions
13937 are slower, but use less code space if more than one function saves
13938 the same number of registers. The @option{-mprolog-function} option
13939 is on by default if you optimize.
13943 Try to make the code as small as possible. At present, this just turns
13944 on the @option{-mep} and @option{-mprolog-function} options.
13946 @item -mtda=@var{n}
13948 Put static or global variables whose size is @var{n} bytes or less into
13949 the tiny data area that register @code{ep} points to. The tiny data
13950 area can hold up to 256 bytes in total (128 bytes for byte references).
13952 @item -msda=@var{n}
13954 Put static or global variables whose size is @var{n} bytes or less into
13955 the small data area that register @code{gp} points to. The small data
13956 area can hold up to 64 kilobytes.
13958 @item -mzda=@var{n}
13960 Put static or global variables whose size is @var{n} bytes or less into
13961 the first 32 kilobytes of memory.
13965 Specify that the target processor is the V850.
13968 @opindex mbig-switch
13969 Generate code suitable for big switch tables. Use this option only if
13970 the assembler/linker complain about out of range branches within a switch
13975 This option will cause r2 and r5 to be used in the code generated by
13976 the compiler. This setting is the default.
13978 @item -mno-app-regs
13979 @opindex mno-app-regs
13980 This option will cause r2 and r5 to be treated as fixed registers.
13984 Specify that the target processor is the V850E1. The preprocessor
13985 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13986 this option is used.
13990 Specify that the target processor is the V850E@. The preprocessor
13991 constant @samp{__v850e__} will be defined if this option is used.
13993 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13994 are defined then a default target processor will be chosen and the
13995 relevant @samp{__v850*__} preprocessor constant will be defined.
13997 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13998 defined, regardless of which processor variant is the target.
14000 @item -mdisable-callt
14001 @opindex mdisable-callt
14002 This option will suppress generation of the CALLT instruction for the
14003 v850e and v850e1 flavors of the v850 architecture. The default is
14004 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14009 @subsection VAX Options
14010 @cindex VAX options
14012 These @samp{-m} options are defined for the VAX:
14017 Do not output certain jump instructions (@code{aobleq} and so on)
14018 that the Unix assembler for the VAX cannot handle across long
14023 Do output those jump instructions, on the assumption that you
14024 will assemble with the GNU assembler.
14028 Output code for g-format floating point numbers instead of d-format.
14031 @node VxWorks Options
14032 @subsection VxWorks Options
14033 @cindex VxWorks Options
14035 The options in this section are defined for all VxWorks targets.
14036 Options specific to the target hardware are listed with the other
14037 options for that target.
14042 GCC can generate code for both VxWorks kernels and real time processes
14043 (RTPs). This option switches from the former to the latter. It also
14044 defines the preprocessor macro @code{__RTP__}.
14047 @opindex non-static
14048 Link an RTP executable against shared libraries rather than static
14049 libraries. The options @option{-static} and @option{-shared} can
14050 also be used for RTPs (@pxref{Link Options}); @option{-static}
14057 These options are passed down to the linker. They are defined for
14058 compatibility with Diab.
14061 @opindex Xbind-lazy
14062 Enable lazy binding of function calls. This option is equivalent to
14063 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14067 Disable lazy binding of function calls. This option is the default and
14068 is defined for compatibility with Diab.
14071 @node x86-64 Options
14072 @subsection x86-64 Options
14073 @cindex x86-64 options
14075 These are listed under @xref{i386 and x86-64 Options}.
14077 @node Xstormy16 Options
14078 @subsection Xstormy16 Options
14079 @cindex Xstormy16 Options
14081 These options are defined for Xstormy16:
14086 Choose startup files and linker script suitable for the simulator.
14089 @node Xtensa Options
14090 @subsection Xtensa Options
14091 @cindex Xtensa Options
14093 These options are supported for Xtensa targets:
14097 @itemx -mno-const16
14099 @opindex mno-const16
14100 Enable or disable use of @code{CONST16} instructions for loading
14101 constant values. The @code{CONST16} instruction is currently not a
14102 standard option from Tensilica. When enabled, @code{CONST16}
14103 instructions are always used in place of the standard @code{L32R}
14104 instructions. The use of @code{CONST16} is enabled by default only if
14105 the @code{L32R} instruction is not available.
14108 @itemx -mno-fused-madd
14109 @opindex mfused-madd
14110 @opindex mno-fused-madd
14111 Enable or disable use of fused multiply/add and multiply/subtract
14112 instructions in the floating-point option. This has no effect if the
14113 floating-point option is not also enabled. Disabling fused multiply/add
14114 and multiply/subtract instructions forces the compiler to use separate
14115 instructions for the multiply and add/subtract operations. This may be
14116 desirable in some cases where strict IEEE 754-compliant results are
14117 required: the fused multiply add/subtract instructions do not round the
14118 intermediate result, thereby producing results with @emph{more} bits of
14119 precision than specified by the IEEE standard. Disabling fused multiply
14120 add/subtract instructions also ensures that the program output is not
14121 sensitive to the compiler's ability to combine multiply and add/subtract
14124 @item -mtext-section-literals
14125 @itemx -mno-text-section-literals
14126 @opindex mtext-section-literals
14127 @opindex mno-text-section-literals
14128 Control the treatment of literal pools. The default is
14129 @option{-mno-text-section-literals}, which places literals in a separate
14130 section in the output file. This allows the literal pool to be placed
14131 in a data RAM/ROM, and it also allows the linker to combine literal
14132 pools from separate object files to remove redundant literals and
14133 improve code size. With @option{-mtext-section-literals}, the literals
14134 are interspersed in the text section in order to keep them as close as
14135 possible to their references. This may be necessary for large assembly
14138 @item -mtarget-align
14139 @itemx -mno-target-align
14140 @opindex mtarget-align
14141 @opindex mno-target-align
14142 When this option is enabled, GCC instructs the assembler to
14143 automatically align instructions to reduce branch penalties at the
14144 expense of some code density. The assembler attempts to widen density
14145 instructions to align branch targets and the instructions following call
14146 instructions. If there are not enough preceding safe density
14147 instructions to align a target, no widening will be performed. The
14148 default is @option{-mtarget-align}. These options do not affect the
14149 treatment of auto-aligned instructions like @code{LOOP}, which the
14150 assembler will always align, either by widening density instructions or
14151 by inserting no-op instructions.
14154 @itemx -mno-longcalls
14155 @opindex mlongcalls
14156 @opindex mno-longcalls
14157 When this option is enabled, GCC instructs the assembler to translate
14158 direct calls to indirect calls unless it can determine that the target
14159 of a direct call is in the range allowed by the call instruction. This
14160 translation typically occurs for calls to functions in other source
14161 files. Specifically, the assembler translates a direct @code{CALL}
14162 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14163 The default is @option{-mno-longcalls}. This option should be used in
14164 programs where the call target can potentially be out of range. This
14165 option is implemented in the assembler, not the compiler, so the
14166 assembly code generated by GCC will still show direct call
14167 instructions---look at the disassembled object code to see the actual
14168 instructions. Note that the assembler will use an indirect call for
14169 every cross-file call, not just those that really will be out of range.
14172 @node zSeries Options
14173 @subsection zSeries Options
14174 @cindex zSeries options
14176 These are listed under @xref{S/390 and zSeries Options}.
14178 @node Code Gen Options
14179 @section Options for Code Generation Conventions
14180 @cindex code generation conventions
14181 @cindex options, code generation
14182 @cindex run-time options
14184 These machine-independent options control the interface conventions
14185 used in code generation.
14187 Most of them have both positive and negative forms; the negative form
14188 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14189 one of the forms is listed---the one which is not the default. You
14190 can figure out the other form by either removing @samp{no-} or adding
14194 @item -fbounds-check
14195 @opindex fbounds-check
14196 For front-ends that support it, generate additional code to check that
14197 indices used to access arrays are within the declared range. This is
14198 currently only supported by the Java and Fortran front-ends, where
14199 this option defaults to true and false respectively.
14203 This option generates traps for signed overflow on addition, subtraction,
14204 multiplication operations.
14208 This option instructs the compiler to assume that signed arithmetic
14209 overflow of addition, subtraction and multiplication wraps around
14210 using twos-complement representation. This flag enables some optimizations
14211 and disables others. This option is enabled by default for the Java
14212 front-end, as required by the Java language specification.
14215 @opindex fexceptions
14216 Enable exception handling. Generates extra code needed to propagate
14217 exceptions. For some targets, this implies GCC will generate frame
14218 unwind information for all functions, which can produce significant data
14219 size overhead, although it does not affect execution. If you do not
14220 specify this option, GCC will enable it by default for languages like
14221 C++ which normally require exception handling, and disable it for
14222 languages like C that do not normally require it. However, you may need
14223 to enable this option when compiling C code that needs to interoperate
14224 properly with exception handlers written in C++. You may also wish to
14225 disable this option if you are compiling older C++ programs that don't
14226 use exception handling.
14228 @item -fnon-call-exceptions
14229 @opindex fnon-call-exceptions
14230 Generate code that allows trapping instructions to throw exceptions.
14231 Note that this requires platform-specific runtime support that does
14232 not exist everywhere. Moreover, it only allows @emph{trapping}
14233 instructions to throw exceptions, i.e.@: memory references or floating
14234 point instructions. It does not allow exceptions to be thrown from
14235 arbitrary signal handlers such as @code{SIGALRM}.
14237 @item -funwind-tables
14238 @opindex funwind-tables
14239 Similar to @option{-fexceptions}, except that it will just generate any needed
14240 static data, but will not affect the generated code in any other way.
14241 You will normally not enable this option; instead, a language processor
14242 that needs this handling would enable it on your behalf.
14244 @item -fasynchronous-unwind-tables
14245 @opindex fasynchronous-unwind-tables
14246 Generate unwind table in dwarf2 format, if supported by target machine. The
14247 table is exact at each instruction boundary, so it can be used for stack
14248 unwinding from asynchronous events (such as debugger or garbage collector).
14250 @item -fpcc-struct-return
14251 @opindex fpcc-struct-return
14252 Return ``short'' @code{struct} and @code{union} values in memory like
14253 longer ones, rather than in registers. This convention is less
14254 efficient, but it has the advantage of allowing intercallability between
14255 GCC-compiled files and files compiled with other compilers, particularly
14256 the Portable C Compiler (pcc).
14258 The precise convention for returning structures in memory depends
14259 on the target configuration macros.
14261 Short structures and unions are those whose size and alignment match
14262 that of some integer type.
14264 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14265 switch is not binary compatible with code compiled with the
14266 @option{-freg-struct-return} switch.
14267 Use it to conform to a non-default application binary interface.
14269 @item -freg-struct-return
14270 @opindex freg-struct-return
14271 Return @code{struct} and @code{union} values in registers when possible.
14272 This is more efficient for small structures than
14273 @option{-fpcc-struct-return}.
14275 If you specify neither @option{-fpcc-struct-return} nor
14276 @option{-freg-struct-return}, GCC defaults to whichever convention is
14277 standard for the target. If there is no standard convention, GCC
14278 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14279 the principal compiler. In those cases, we can choose the standard, and
14280 we chose the more efficient register return alternative.
14282 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14283 switch is not binary compatible with code compiled with the
14284 @option{-fpcc-struct-return} switch.
14285 Use it to conform to a non-default application binary interface.
14287 @item -fshort-enums
14288 @opindex fshort-enums
14289 Allocate to an @code{enum} type only as many bytes as it needs for the
14290 declared range of possible values. Specifically, the @code{enum} type
14291 will be equivalent to the smallest integer type which has enough room.
14293 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14294 code that is not binary compatible with code generated without that switch.
14295 Use it to conform to a non-default application binary interface.
14297 @item -fshort-double
14298 @opindex fshort-double
14299 Use the same size for @code{double} as for @code{float}.
14301 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14302 code that is not binary compatible with code generated without that switch.
14303 Use it to conform to a non-default application binary interface.
14305 @item -fshort-wchar
14306 @opindex fshort-wchar
14307 Override the underlying type for @samp{wchar_t} to be @samp{short
14308 unsigned int} instead of the default for the target. This option is
14309 useful for building programs to run under WINE@.
14311 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14312 code that is not binary compatible with code generated without that switch.
14313 Use it to conform to a non-default application binary interface.
14316 @opindex fno-common
14317 In C, allocate even uninitialized global variables in the data section of the
14318 object file, rather than generating them as common blocks. This has the
14319 effect that if the same variable is declared (without @code{extern}) in
14320 two different compilations, you will get an error when you link them.
14321 The only reason this might be useful is if you wish to verify that the
14322 program will work on other systems which always work this way.
14326 Ignore the @samp{#ident} directive.
14328 @item -finhibit-size-directive
14329 @opindex finhibit-size-directive
14330 Don't output a @code{.size} assembler directive, or anything else that
14331 would cause trouble if the function is split in the middle, and the
14332 two halves are placed at locations far apart in memory. This option is
14333 used when compiling @file{crtstuff.c}; you should not need to use it
14336 @item -fverbose-asm
14337 @opindex fverbose-asm
14338 Put extra commentary information in the generated assembly code to
14339 make it more readable. This option is generally only of use to those
14340 who actually need to read the generated assembly code (perhaps while
14341 debugging the compiler itself).
14343 @option{-fno-verbose-asm}, the default, causes the
14344 extra information to be omitted and is useful when comparing two assembler
14347 @item -frecord-gcc-switches
14348 @opindex frecord-gcc-switches
14349 This switch causes the command line that was used to invoke the
14350 compiler to be recorded into the object file that is being created.
14351 This switch is only implemented on some targets and the exact format
14352 of the recording is target and binary file format dependent, but it
14353 usually takes the form of a section containing ASCII text. This
14354 switch is related to the @option{-fverbose-asm} switch, but that
14355 switch only records information in the assembler output file as
14356 comments, so it never reaches the object file.
14360 @cindex global offset table
14362 Generate position-independent code (PIC) suitable for use in a shared
14363 library, if supported for the target machine. Such code accesses all
14364 constant addresses through a global offset table (GOT)@. The dynamic
14365 loader resolves the GOT entries when the program starts (the dynamic
14366 loader is not part of GCC; it is part of the operating system). If
14367 the GOT size for the linked executable exceeds a machine-specific
14368 maximum size, you get an error message from the linker indicating that
14369 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14370 instead. (These maximums are 8k on the SPARC and 32k
14371 on the m68k and RS/6000. The 386 has no such limit.)
14373 Position-independent code requires special support, and therefore works
14374 only on certain machines. For the 386, GCC supports PIC for System V
14375 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14376 position-independent.
14378 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14383 If supported for the target machine, emit position-independent code,
14384 suitable for dynamic linking and avoiding any limit on the size of the
14385 global offset table. This option makes a difference on the m68k,
14386 PowerPC and SPARC@.
14388 Position-independent code requires special support, and therefore works
14389 only on certain machines.
14391 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14398 These options are similar to @option{-fpic} and @option{-fPIC}, but
14399 generated position independent code can be only linked into executables.
14400 Usually these options are used when @option{-pie} GCC option will be
14401 used during linking.
14403 @option{-fpie} and @option{-fPIE} both define the macros
14404 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14405 for @option{-fpie} and 2 for @option{-fPIE}.
14407 @item -fno-jump-tables
14408 @opindex fno-jump-tables
14409 Do not use jump tables for switch statements even where it would be
14410 more efficient than other code generation strategies. This option is
14411 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14412 building code which forms part of a dynamic linker and cannot
14413 reference the address of a jump table. On some targets, jump tables
14414 do not require a GOT and this option is not needed.
14416 @item -ffixed-@var{reg}
14418 Treat the register named @var{reg} as a fixed register; generated code
14419 should never refer to it (except perhaps as a stack pointer, frame
14420 pointer or in some other fixed role).
14422 @var{reg} must be the name of a register. The register names accepted
14423 are machine-specific and are defined in the @code{REGISTER_NAMES}
14424 macro in the machine description macro file.
14426 This flag does not have a negative form, because it specifies a
14429 @item -fcall-used-@var{reg}
14430 @opindex fcall-used
14431 Treat the register named @var{reg} as an allocable register that is
14432 clobbered by function calls. It may be allocated for temporaries or
14433 variables that do not live across a call. Functions compiled this way
14434 will not save and restore the register @var{reg}.
14436 It is an error to used this flag with the frame pointer or stack pointer.
14437 Use of this flag for other registers that have fixed pervasive roles in
14438 the machine's execution model will produce disastrous results.
14440 This flag does not have a negative form, because it specifies a
14443 @item -fcall-saved-@var{reg}
14444 @opindex fcall-saved
14445 Treat the register named @var{reg} as an allocable register saved by
14446 functions. It may be allocated even for temporaries or variables that
14447 live across a call. Functions compiled this way will save and restore
14448 the register @var{reg} if they use it.
14450 It is an error to used this flag with the frame pointer or stack pointer.
14451 Use of this flag for other registers that have fixed pervasive roles in
14452 the machine's execution model will produce disastrous results.
14454 A different sort of disaster will result from the use of this flag for
14455 a register in which function values may be returned.
14457 This flag does not have a negative form, because it specifies a
14460 @item -fpack-struct[=@var{n}]
14461 @opindex fpack-struct
14462 Without a value specified, pack all structure members together without
14463 holes. When a value is specified (which must be a small power of two), pack
14464 structure members according to this value, representing the maximum
14465 alignment (that is, objects with default alignment requirements larger than
14466 this will be output potentially unaligned at the next fitting location.
14468 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14469 code that is not binary compatible with code generated without that switch.
14470 Additionally, it makes the code suboptimal.
14471 Use it to conform to a non-default application binary interface.
14473 @item -finstrument-functions
14474 @opindex finstrument-functions
14475 Generate instrumentation calls for entry and exit to functions. Just
14476 after function entry and just before function exit, the following
14477 profiling functions will be called with the address of the current
14478 function and its call site. (On some platforms,
14479 @code{__builtin_return_address} does not work beyond the current
14480 function, so the call site information may not be available to the
14481 profiling functions otherwise.)
14484 void __cyg_profile_func_enter (void *this_fn,
14486 void __cyg_profile_func_exit (void *this_fn,
14490 The first argument is the address of the start of the current function,
14491 which may be looked up exactly in the symbol table.
14493 This instrumentation is also done for functions expanded inline in other
14494 functions. The profiling calls will indicate where, conceptually, the
14495 inline function is entered and exited. This means that addressable
14496 versions of such functions must be available. If all your uses of a
14497 function are expanded inline, this may mean an additional expansion of
14498 code size. If you use @samp{extern inline} in your C code, an
14499 addressable version of such functions must be provided. (This is
14500 normally the case anyways, but if you get lucky and the optimizer always
14501 expands the functions inline, you might have gotten away without
14502 providing static copies.)
14504 A function may be given the attribute @code{no_instrument_function}, in
14505 which case this instrumentation will not be done. This can be used, for
14506 example, for the profiling functions listed above, high-priority
14507 interrupt routines, and any functions from which the profiling functions
14508 cannot safely be called (perhaps signal handlers, if the profiling
14509 routines generate output or allocate memory).
14511 @item -fstack-check
14512 @opindex fstack-check
14513 Generate code to verify that you do not go beyond the boundary of the
14514 stack. You should specify this flag if you are running in an
14515 environment with multiple threads, but only rarely need to specify it in
14516 a single-threaded environment since stack overflow is automatically
14517 detected on nearly all systems if there is only one stack.
14519 Note that this switch does not actually cause checking to be done; the
14520 operating system must do that. The switch causes generation of code
14521 to ensure that the operating system sees the stack being extended.
14523 @item -fstack-limit-register=@var{reg}
14524 @itemx -fstack-limit-symbol=@var{sym}
14525 @itemx -fno-stack-limit
14526 @opindex fstack-limit-register
14527 @opindex fstack-limit-symbol
14528 @opindex fno-stack-limit
14529 Generate code to ensure that the stack does not grow beyond a certain value,
14530 either the value of a register or the address of a symbol. If the stack
14531 would grow beyond the value, a signal is raised. For most targets,
14532 the signal is raised before the stack overruns the boundary, so
14533 it is possible to catch the signal without taking special precautions.
14535 For instance, if the stack starts at absolute address @samp{0x80000000}
14536 and grows downwards, you can use the flags
14537 @option{-fstack-limit-symbol=__stack_limit} and
14538 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
14539 of 128KB@. Note that this may only work with the GNU linker.
14541 @cindex aliasing of parameters
14542 @cindex parameters, aliased
14543 @item -fargument-alias
14544 @itemx -fargument-noalias
14545 @itemx -fargument-noalias-global
14546 @itemx -fargument-noalias-anything
14547 @opindex fargument-alias
14548 @opindex fargument-noalias
14549 @opindex fargument-noalias-global
14550 @opindex fargument-noalias-anything
14551 Specify the possible relationships among parameters and between
14552 parameters and global data.
14554 @option{-fargument-alias} specifies that arguments (parameters) may
14555 alias each other and may alias global storage.@*
14556 @option{-fargument-noalias} specifies that arguments do not alias
14557 each other, but may alias global storage.@*
14558 @option{-fargument-noalias-global} specifies that arguments do not
14559 alias each other and do not alias global storage.
14560 @option{-fargument-noalias-anything} specifies that arguments do not
14561 alias any other storage.
14563 Each language will automatically use whatever option is required by
14564 the language standard. You should not need to use these options yourself.
14566 @item -fleading-underscore
14567 @opindex fleading-underscore
14568 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14569 change the way C symbols are represented in the object file. One use
14570 is to help link with legacy assembly code.
14572 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14573 generate code that is not binary compatible with code generated without that
14574 switch. Use it to conform to a non-default application binary interface.
14575 Not all targets provide complete support for this switch.
14577 @item -ftls-model=@var{model}
14578 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14579 The @var{model} argument should be one of @code{global-dynamic},
14580 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
14582 The default without @option{-fpic} is @code{initial-exec}; with
14583 @option{-fpic} the default is @code{global-dynamic}.
14585 @item -fvisibility=@var{default|internal|hidden|protected}
14586 @opindex fvisibility
14587 Set the default ELF image symbol visibility to the specified option---all
14588 symbols will be marked with this unless overridden within the code.
14589 Using this feature can very substantially improve linking and
14590 load times of shared object libraries, produce more optimized
14591 code, provide near-perfect API export and prevent symbol clashes.
14592 It is @strong{strongly} recommended that you use this in any shared objects
14595 Despite the nomenclature, @code{default} always means public ie;
14596 available to be linked against from outside the shared object.
14597 @code{protected} and @code{internal} are pretty useless in real-world
14598 usage so the only other commonly used option will be @code{hidden}.
14599 The default if @option{-fvisibility} isn't specified is
14600 @code{default}, i.e., make every
14601 symbol public---this causes the same behavior as previous versions of
14604 A good explanation of the benefits offered by ensuring ELF
14605 symbols have the correct visibility is given by ``How To Write
14606 Shared Libraries'' by Ulrich Drepper (which can be found at
14607 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
14608 solution made possible by this option to marking things hidden when
14609 the default is public is to make the default hidden and mark things
14610 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
14611 and @code{__attribute__ ((visibility("default")))} instead of
14612 @code{__declspec(dllexport)} you get almost identical semantics with
14613 identical syntax. This is a great boon to those working with
14614 cross-platform projects.
14616 For those adding visibility support to existing code, you may find
14617 @samp{#pragma GCC visibility} of use. This works by you enclosing
14618 the declarations you wish to set visibility for with (for example)
14619 @samp{#pragma GCC visibility push(hidden)} and
14620 @samp{#pragma GCC visibility pop}.
14621 Bear in mind that symbol visibility should be viewed @strong{as
14622 part of the API interface contract} and thus all new code should
14623 always specify visibility when it is not the default ie; declarations
14624 only for use within the local DSO should @strong{always} be marked explicitly
14625 as hidden as so to avoid PLT indirection overheads---making this
14626 abundantly clear also aids readability and self-documentation of the code.
14627 Note that due to ISO C++ specification requirements, operator new and
14628 operator delete must always be of default visibility.
14630 Be aware that headers from outside your project, in particular system
14631 headers and headers from any other library you use, may not be
14632 expecting to be compiled with visibility other than the default. You
14633 may need to explicitly say @samp{#pragma GCC visibility push(default)}
14634 before including any such headers.
14636 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
14637 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
14638 no modifications. However, this means that calls to @samp{extern}
14639 functions with no explicit visibility will use the PLT, so it is more
14640 effective to use @samp{__attribute ((visibility))} and/or
14641 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
14642 declarations should be treated as hidden.
14644 Note that @samp{-fvisibility} does affect C++ vague linkage
14645 entities. This means that, for instance, an exception class that will
14646 be thrown between DSOs must be explicitly marked with default
14647 visibility so that the @samp{type_info} nodes will be unified between
14650 An overview of these techniques, their benefits and how to use them
14651 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
14657 @node Environment Variables
14658 @section Environment Variables Affecting GCC
14659 @cindex environment variables
14661 @c man begin ENVIRONMENT
14662 This section describes several environment variables that affect how GCC
14663 operates. Some of them work by specifying directories or prefixes to use
14664 when searching for various kinds of files. Some are used to specify other
14665 aspects of the compilation environment.
14667 Note that you can also specify places to search using options such as
14668 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
14669 take precedence over places specified using environment variables, which
14670 in turn take precedence over those specified by the configuration of GCC@.
14671 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
14672 GNU Compiler Collection (GCC) Internals}.
14677 @c @itemx LC_COLLATE
14679 @c @itemx LC_MONETARY
14680 @c @itemx LC_NUMERIC
14685 @c @findex LC_COLLATE
14686 @findex LC_MESSAGES
14687 @c @findex LC_MONETARY
14688 @c @findex LC_NUMERIC
14692 These environment variables control the way that GCC uses
14693 localization information that allow GCC to work with different
14694 national conventions. GCC inspects the locale categories
14695 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
14696 so. These locale categories can be set to any value supported by your
14697 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
14698 Kingdom encoded in UTF-8.
14700 The @env{LC_CTYPE} environment variable specifies character
14701 classification. GCC uses it to determine the character boundaries in
14702 a string; this is needed for some multibyte encodings that contain quote
14703 and escape characters that would otherwise be interpreted as a string
14706 The @env{LC_MESSAGES} environment variable specifies the language to
14707 use in diagnostic messages.
14709 If the @env{LC_ALL} environment variable is set, it overrides the value
14710 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
14711 and @env{LC_MESSAGES} default to the value of the @env{LANG}
14712 environment variable. If none of these variables are set, GCC
14713 defaults to traditional C English behavior.
14717 If @env{TMPDIR} is set, it specifies the directory to use for temporary
14718 files. GCC uses temporary files to hold the output of one stage of
14719 compilation which is to be used as input to the next stage: for example,
14720 the output of the preprocessor, which is the input to the compiler
14723 @item GCC_EXEC_PREFIX
14724 @findex GCC_EXEC_PREFIX
14725 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
14726 names of the subprograms executed by the compiler. No slash is added
14727 when this prefix is combined with the name of a subprogram, but you can
14728 specify a prefix that ends with a slash if you wish.
14730 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
14731 an appropriate prefix to use based on the pathname it was invoked with.
14733 If GCC cannot find the subprogram using the specified prefix, it
14734 tries looking in the usual places for the subprogram.
14736 The default value of @env{GCC_EXEC_PREFIX} is
14737 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
14738 the installed compiler. In many cases @var{prefix} is the value
14739 of @code{prefix} when you ran the @file{configure} script.
14741 Other prefixes specified with @option{-B} take precedence over this prefix.
14743 This prefix is also used for finding files such as @file{crt0.o} that are
14746 In addition, the prefix is used in an unusual way in finding the
14747 directories to search for header files. For each of the standard
14748 directories whose name normally begins with @samp{/usr/local/lib/gcc}
14749 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
14750 replacing that beginning with the specified prefix to produce an
14751 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
14752 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
14753 These alternate directories are searched first; the standard directories
14754 come next. If a standard directory begins with the configured
14755 @var{prefix} then the value of @var{prefix} is replaced by
14756 @env{GCC_EXEC_PREFIX} when looking for header files.
14758 @item COMPILER_PATH
14759 @findex COMPILER_PATH
14760 The value of @env{COMPILER_PATH} is a colon-separated list of
14761 directories, much like @env{PATH}. GCC tries the directories thus
14762 specified when searching for subprograms, if it can't find the
14763 subprograms using @env{GCC_EXEC_PREFIX}.
14766 @findex LIBRARY_PATH
14767 The value of @env{LIBRARY_PATH} is a colon-separated list of
14768 directories, much like @env{PATH}. When configured as a native compiler,
14769 GCC tries the directories thus specified when searching for special
14770 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
14771 using GCC also uses these directories when searching for ordinary
14772 libraries for the @option{-l} option (but directories specified with
14773 @option{-L} come first).
14777 @cindex locale definition
14778 This variable is used to pass locale information to the compiler. One way in
14779 which this information is used is to determine the character set to be used
14780 when character literals, string literals and comments are parsed in C and C++.
14781 When the compiler is configured to allow multibyte characters,
14782 the following values for @env{LANG} are recognized:
14786 Recognize JIS characters.
14788 Recognize SJIS characters.
14790 Recognize EUCJP characters.
14793 If @env{LANG} is not defined, or if it has some other value, then the
14794 compiler will use mblen and mbtowc as defined by the default locale to
14795 recognize and translate multibyte characters.
14799 Some additional environments variables affect the behavior of the
14802 @include cppenv.texi
14806 @node Precompiled Headers
14807 @section Using Precompiled Headers
14808 @cindex precompiled headers
14809 @cindex speed of compilation
14811 Often large projects have many header files that are included in every
14812 source file. The time the compiler takes to process these header files
14813 over and over again can account for nearly all of the time required to
14814 build the project. To make builds faster, GCC allows users to
14815 `precompile' a header file; then, if builds can use the precompiled
14816 header file they will be much faster.
14818 To create a precompiled header file, simply compile it as you would any
14819 other file, if necessary using the @option{-x} option to make the driver
14820 treat it as a C or C++ header file. You will probably want to use a
14821 tool like @command{make} to keep the precompiled header up-to-date when
14822 the headers it contains change.
14824 A precompiled header file will be searched for when @code{#include} is
14825 seen in the compilation. As it searches for the included file
14826 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
14827 compiler looks for a precompiled header in each directory just before it
14828 looks for the include file in that directory. The name searched for is
14829 the name specified in the @code{#include} with @samp{.gch} appended. If
14830 the precompiled header file can't be used, it is ignored.
14832 For instance, if you have @code{#include "all.h"}, and you have
14833 @file{all.h.gch} in the same directory as @file{all.h}, then the
14834 precompiled header file will be used if possible, and the original
14835 header will be used otherwise.
14837 Alternatively, you might decide to put the precompiled header file in a
14838 directory and use @option{-I} to ensure that directory is searched
14839 before (or instead of) the directory containing the original header.
14840 Then, if you want to check that the precompiled header file is always
14841 used, you can put a file of the same name as the original header in this
14842 directory containing an @code{#error} command.
14844 This also works with @option{-include}. So yet another way to use
14845 precompiled headers, good for projects not designed with precompiled
14846 header files in mind, is to simply take most of the header files used by
14847 a project, include them from another header file, precompile that header
14848 file, and @option{-include} the precompiled header. If the header files
14849 have guards against multiple inclusion, they will be skipped because
14850 they've already been included (in the precompiled header).
14852 If you need to precompile the same header file for different
14853 languages, targets, or compiler options, you can instead make a
14854 @emph{directory} named like @file{all.h.gch}, and put each precompiled
14855 header in the directory, perhaps using @option{-o}. It doesn't matter
14856 what you call the files in the directory, every precompiled header in
14857 the directory will be considered. The first precompiled header
14858 encountered in the directory that is valid for this compilation will
14859 be used; they're searched in no particular order.
14861 There are many other possibilities, limited only by your imagination,
14862 good sense, and the constraints of your build system.
14864 A precompiled header file can be used only when these conditions apply:
14868 Only one precompiled header can be used in a particular compilation.
14871 A precompiled header can't be used once the first C token is seen. You
14872 can have preprocessor directives before a precompiled header; you can
14873 even include a precompiled header from inside another header, so long as
14874 there are no C tokens before the @code{#include}.
14877 The precompiled header file must be produced for the same language as
14878 the current compilation. You can't use a C precompiled header for a C++
14882 The precompiled header file must have been produced by the same compiler
14883 binary as the current compilation is using.
14886 Any macros defined before the precompiled header is included must
14887 either be defined in the same way as when the precompiled header was
14888 generated, or must not affect the precompiled header, which usually
14889 means that they don't appear in the precompiled header at all.
14891 The @option{-D} option is one way to define a macro before a
14892 precompiled header is included; using a @code{#define} can also do it.
14893 There are also some options that define macros implicitly, like
14894 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
14897 @item If debugging information is output when using the precompiled
14898 header, using @option{-g} or similar, the same kind of debugging information
14899 must have been output when building the precompiled header. However,
14900 a precompiled header built using @option{-g} can be used in a compilation
14901 when no debugging information is being output.
14903 @item The same @option{-m} options must generally be used when building
14904 and using the precompiled header. @xref{Submodel Options},
14905 for any cases where this rule is relaxed.
14907 @item Each of the following options must be the same when building and using
14908 the precompiled header:
14910 @gccoptlist{-fexceptions -funit-at-a-time}
14913 Some other command-line options starting with @option{-f},
14914 @option{-p}, or @option{-O} must be defined in the same way as when
14915 the precompiled header was generated. At present, it's not clear
14916 which options are safe to change and which are not; the safest choice
14917 is to use exactly the same options when generating and using the
14918 precompiled header. The following are known to be safe:
14920 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
14921 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
14922 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
14927 For all of these except the last, the compiler will automatically
14928 ignore the precompiled header if the conditions aren't met. If you
14929 find an option combination that doesn't work and doesn't cause the
14930 precompiled header to be ignored, please consider filing a bug report,
14933 If you do use differing options when generating and using the
14934 precompiled header, the actual behavior will be a mixture of the
14935 behavior for the options. For instance, if you use @option{-g} to
14936 generate the precompiled header but not when using it, you may or may
14937 not get debugging information for routines in the precompiled header.
14939 @node Running Protoize
14940 @section Running Protoize
14942 The program @code{protoize} is an optional part of GCC@. You can use
14943 it to add prototypes to a program, thus converting the program to ISO
14944 C in one respect. The companion program @code{unprotoize} does the
14945 reverse: it removes argument types from any prototypes that are found.
14947 When you run these programs, you must specify a set of source files as
14948 command line arguments. The conversion programs start out by compiling
14949 these files to see what functions they define. The information gathered
14950 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
14952 After scanning comes actual conversion. The specified files are all
14953 eligible to be converted; any files they include (whether sources or
14954 just headers) are eligible as well.
14956 But not all the eligible files are converted. By default,
14957 @code{protoize} and @code{unprotoize} convert only source and header
14958 files in the current directory. You can specify additional directories
14959 whose files should be converted with the @option{-d @var{directory}}
14960 option. You can also specify particular files to exclude with the
14961 @option{-x @var{file}} option. A file is converted if it is eligible, its
14962 directory name matches one of the specified directory names, and its
14963 name within the directory has not been excluded.
14965 Basic conversion with @code{protoize} consists of rewriting most
14966 function definitions and function declarations to specify the types of
14967 the arguments. The only ones not rewritten are those for varargs
14970 @code{protoize} optionally inserts prototype declarations at the
14971 beginning of the source file, to make them available for any calls that
14972 precede the function's definition. Or it can insert prototype
14973 declarations with block scope in the blocks where undeclared functions
14976 Basic conversion with @code{unprotoize} consists of rewriting most
14977 function declarations to remove any argument types, and rewriting
14978 function definitions to the old-style pre-ISO form.
14980 Both conversion programs print a warning for any function declaration or
14981 definition that they can't convert. You can suppress these warnings
14984 The output from @code{protoize} or @code{unprotoize} replaces the
14985 original source file. The original file is renamed to a name ending
14986 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
14987 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
14988 for DOS) file already exists, then the source file is simply discarded.
14990 @code{protoize} and @code{unprotoize} both depend on GCC itself to
14991 scan the program and collect information about the functions it uses.
14992 So neither of these programs will work until GCC is installed.
14994 Here is a table of the options you can use with @code{protoize} and
14995 @code{unprotoize}. Each option works with both programs unless
14999 @item -B @var{directory}
15000 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15001 usual directory (normally @file{/usr/local/lib}). This file contains
15002 prototype information about standard system functions. This option
15003 applies only to @code{protoize}.
15005 @item -c @var{compilation-options}
15006 Use @var{compilation-options} as the options when running @command{gcc} to
15007 produce the @samp{.X} files. The special option @option{-aux-info} is
15008 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15010 Note that the compilation options must be given as a single argument to
15011 @code{protoize} or @code{unprotoize}. If you want to specify several
15012 @command{gcc} options, you must quote the entire set of compilation options
15013 to make them a single word in the shell.
15015 There are certain @command{gcc} arguments that you cannot use, because they
15016 would produce the wrong kind of output. These include @option{-g},
15017 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15018 the @var{compilation-options}, they are ignored.
15021 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15022 systems) instead of @samp{.c}. This is convenient if you are converting
15023 a C program to C++. This option applies only to @code{protoize}.
15026 Add explicit global declarations. This means inserting explicit
15027 declarations at the beginning of each source file for each function
15028 that is called in the file and was not declared. These declarations
15029 precede the first function definition that contains a call to an
15030 undeclared function. This option applies only to @code{protoize}.
15032 @item -i @var{string}
15033 Indent old-style parameter declarations with the string @var{string}.
15034 This option applies only to @code{protoize}.
15036 @code{unprotoize} converts prototyped function definitions to old-style
15037 function definitions, where the arguments are declared between the
15038 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15039 uses five spaces as the indentation. If you want to indent with just
15040 one space instead, use @option{-i " "}.
15043 Keep the @samp{.X} files. Normally, they are deleted after conversion
15047 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15048 a prototype declaration for each function in each block which calls the
15049 function without any declaration. This option applies only to
15053 Make no real changes. This mode just prints information about the conversions
15054 that would have been done without @option{-n}.
15057 Make no @samp{.save} files. The original files are simply deleted.
15058 Use this option with caution.
15060 @item -p @var{program}
15061 Use the program @var{program} as the compiler. Normally, the name
15062 @file{gcc} is used.
15065 Work quietly. Most warnings are suppressed.
15068 Print the version number, just like @option{-v} for @command{gcc}.
15071 If you need special compiler options to compile one of your program's
15072 source files, then you should generate that file's @samp{.X} file
15073 specially, by running @command{gcc} on that source file with the
15074 appropriate options and the option @option{-aux-info}. Then run
15075 @code{protoize} on the entire set of files. @code{protoize} will use
15076 the existing @samp{.X} file because it is newer than the source file.
15080 gcc -Dfoo=bar file1.c -aux-info file1.X
15085 You need to include the special files along with the rest in the
15086 @code{protoize} command, even though their @samp{.X} files already
15087 exist, because otherwise they won't get converted.
15089 @xref{Protoize Caveats}, for more information on how to use
15090 @code{protoize} successfully.