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 @gol
240 -Wno-invalid-offsetof -Winvalid-pch @gol
241 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -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 -Wvariadic-macros @gol
258 -Wvolatile-register-var -Wwrite-strings}
260 @item C-only Warning Options
261 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
262 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
263 -Wold-style-declaration -Wold-style-definition @gol
264 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
265 -Wdeclaration-after-statement -Wpointer-sign}
267 @item Debugging Options
268 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
269 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
270 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
271 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
272 -fdump-ipa-all -fdump-ipa-cgraph @gol
274 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
277 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
279 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
283 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-nrv -fdump-tree-vect @gol
288 -fdump-tree-sink @gol
289 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-salias @gol
291 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
292 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
293 -ftree-vectorizer-verbose=@var{n} @gol
294 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
295 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
296 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
297 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
298 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
299 -ftest-coverage -ftime-report -fvar-tracking @gol
300 -g -g@var{level} -gcoff -gdwarf-2 @gol
301 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
302 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
303 -print-multi-directory -print-multi-lib @gol
304 -print-prog-name=@var{program} -print-search-dirs -Q @gol
307 @item Optimization Options
308 @xref{Optimize Options,,Options that Control Optimization}.
309 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
310 -falign-labels=@var{n} -falign-loops=@var{n} @gol
311 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
312 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
313 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
314 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
315 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
316 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
317 -fexpensive-optimizations -ffast-math -ffloat-store @gol
318 -fforce-addr -fforward-propagate -ffunction-sections @gol
319 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
320 -fcrossjumping -fif-conversion -fif-conversion2 @gol
321 -finline-functions -finline-functions-called-once @gol
322 -finline-limit=@var{n} -fkeep-inline-functions @gol
323 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
324 -fmodulo-sched -fno-branch-count-reg @gol
325 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
326 -fno-function-cse -fno-guess-branch-probability @gol
327 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
328 -funsafe-math-optimizations -funsafe-loop-optimizations @gol
329 -ffinite-math-only -fno-signed-zeros @gol
330 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
331 -fomit-frame-pointer -foptimize-register-move @gol
332 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
333 -fprofile-generate -fprofile-use @gol
334 -fregmove -frename-registers @gol
335 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
336 -frerun-cse-after-loop @gol
337 -frounding-math -frtl-abstract-sequences @gol
338 -fschedule-insns -fschedule-insns2 @gol
339 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
340 -fsched-spec-load-dangerous @gol
341 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
342 -fsched2-use-superblocks @gol
343 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
344 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
345 -fno-split-wide-types -fstack-protector -fstack-protector-all @gol
346 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
347 -funroll-all-loops -funroll-loops -fpeel-loops @gol
348 -fsplit-ivs-in-unroller -funswitch-loops @gol
349 -fvariable-expansion-in-unroller @gol
350 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
351 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
352 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
353 -ftree-ch -ftree-sra -ftree-ter -ftree-fre -ftree-vectorize @gol
354 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
355 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
356 --param @var{name}=@var{value}
357 -O -O0 -O1 -O2 -O3 -Os}
359 @item Preprocessor Options
360 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
361 @gccoptlist{-A@var{question}=@var{answer} @gol
362 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
363 -C -dD -dI -dM -dN @gol
364 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
365 -idirafter @var{dir} @gol
366 -include @var{file} -imacros @var{file} @gol
367 -iprefix @var{file} -iwithprefix @var{dir} @gol
368 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
369 -imultilib @var{dir} -isysroot @var{dir} @gol
370 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
371 -P -fworking-directory -remap @gol
372 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
373 -Xpreprocessor @var{option}}
375 @item Assembler Option
376 @xref{Assembler Options,,Passing Options to the Assembler}.
377 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
380 @xref{Link Options,,Options for Linking}.
381 @gccoptlist{@var{object-file-name} -l@var{library} @gol
382 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
383 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
384 -Wl,@var{option} -Xlinker @var{option} @gol
387 @item Directory Options
388 @xref{Directory Options,,Options for Directory Search}.
389 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
390 -specs=@var{file} -I- --sysroot=@var{dir}}
393 @c I wrote this xref this way to avoid overfull hbox. -- rms
394 @xref{Target Options}.
395 @gccoptlist{-V @var{version} -b @var{machine}}
397 @item Machine Dependent Options
398 @xref{Submodel Options,,Hardware Models and Configurations}.
399 @c This list is ordered alphanumerically by subsection name.
400 @c Try and put the significant identifier (CPU or system) first,
401 @c so users have a clue at guessing where the ones they want will be.
404 @gccoptlist{-EB -EL @gol
405 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
406 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
409 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
410 -mabi=@var{name} @gol
411 -mapcs-stack-check -mno-apcs-stack-check @gol
412 -mapcs-float -mno-apcs-float @gol
413 -mapcs-reentrant -mno-apcs-reentrant @gol
414 -msched-prolog -mno-sched-prolog @gol
415 -mlittle-endian -mbig-endian -mwords-little-endian @gol
416 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
417 -mthumb-interwork -mno-thumb-interwork @gol
418 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
419 -mstructure-size-boundary=@var{n} @gol
420 -mabort-on-noreturn @gol
421 -mlong-calls -mno-long-calls @gol
422 -msingle-pic-base -mno-single-pic-base @gol
423 -mpic-register=@var{reg} @gol
424 -mnop-fun-dllimport @gol
425 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
426 -mpoke-function-name @gol
428 -mtpcs-frame -mtpcs-leaf-frame @gol
429 -mcaller-super-interworking -mcallee-super-interworking @gol
433 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
434 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
436 @emph{Blackfin Options}
437 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
438 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
439 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
440 -mno-id-shared-library -mshared-library-id=@var{n} @gol
441 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
442 -msep-data -mno-sep-data -mlong-calls -mno-long-calls}
445 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
446 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
447 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
448 -mstack-align -mdata-align -mconst-align @gol
449 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
450 -melf -maout -melinux -mlinux -sim -sim2 @gol
451 -mmul-bug-workaround -mno-mul-bug-workaround}
454 @gccoptlist{-mmac -mpush-args}
456 @emph{Darwin Options}
457 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
458 -arch_only -bind_at_load -bundle -bundle_loader @gol
459 -client_name -compatibility_version -current_version @gol
461 -dependency-file -dylib_file -dylinker_install_name @gol
462 -dynamic -dynamiclib -exported_symbols_list @gol
463 -filelist -flat_namespace -force_cpusubtype_ALL @gol
464 -force_flat_namespace -headerpad_max_install_names @gol
466 -image_base -init -install_name -keep_private_externs @gol
467 -multi_module -multiply_defined -multiply_defined_unused @gol
468 -noall_load -no_dead_strip_inits_and_terms @gol
469 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
470 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
471 -private_bundle -read_only_relocs -sectalign @gol
472 -sectobjectsymbols -whyload -seg1addr @gol
473 -sectcreate -sectobjectsymbols -sectorder @gol
474 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
475 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
476 -segprot -segs_read_only_addr -segs_read_write_addr @gol
477 -single_module -static -sub_library -sub_umbrella @gol
478 -twolevel_namespace -umbrella -undefined @gol
479 -unexported_symbols_list -weak_reference_mismatches @gol
480 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
481 -mkernel -mone-byte-bool}
483 @emph{DEC Alpha Options}
484 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
485 -mieee -mieee-with-inexact -mieee-conformant @gol
486 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
487 -mtrap-precision=@var{mode} -mbuild-constants @gol
488 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
489 -mbwx -mmax -mfix -mcix @gol
490 -mfloat-vax -mfloat-ieee @gol
491 -mexplicit-relocs -msmall-data -mlarge-data @gol
492 -msmall-text -mlarge-text @gol
493 -mmemory-latency=@var{time}}
495 @emph{DEC Alpha/VMS Options}
496 @gccoptlist{-mvms-return-codes}
499 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
500 -mhard-float -msoft-float @gol
501 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
502 -mdouble -mno-double @gol
503 -mmedia -mno-media -mmuladd -mno-muladd @gol
504 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
505 -mlinked-fp -mlong-calls -malign-labels @gol
506 -mlibrary-pic -macc-4 -macc-8 @gol
507 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
508 -moptimize-membar -mno-optimize-membar @gol
509 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
510 -mvliw-branch -mno-vliw-branch @gol
511 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
512 -mno-nested-cond-exec -mtomcat-stats @gol
516 @emph{GNU/Linux Options}
517 @gccoptlist{-muclibc}
519 @emph{H8/300 Options}
520 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
523 @gccoptlist{-march=@var{architecture-type} @gol
524 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
525 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
526 -mfixed-range=@var{register-range} @gol
527 -mjump-in-delay -mlinker-opt -mlong-calls @gol
528 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
529 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
530 -mno-jump-in-delay -mno-long-load-store @gol
531 -mno-portable-runtime -mno-soft-float @gol
532 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
533 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
534 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
535 -munix=@var{unix-std} -nolibdld -static -threads}
537 @emph{i386 and x86-64 Options}
538 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
539 -mfpmath=@var{unit} @gol
540 -masm=@var{dialect} -mno-fancy-math-387 @gol
541 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
542 -mno-wide-multiply -mrtd -malign-double @gol
543 -mpreferred-stack-boundary=@var{num} @gol
544 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm @gol
545 -mthreads -mno-align-stringops -minline-all-stringops @gol
546 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
547 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
549 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
550 -mcmodel=@var{code-model} @gol
551 -m32 -m64 -mlarge-data-threshold=@var{num}}
554 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
555 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
556 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
557 -minline-float-divide-max-throughput @gol
558 -minline-int-divide-min-latency @gol
559 -minline-int-divide-max-throughput @gol
560 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
561 -mno-dwarf2-asm -mearly-stop-bits @gol
562 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
563 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
564 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
565 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
566 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
567 -mno-sched-prefer-non-data-spec-insns @gol
568 -mno-sched-prefer-non-control-spec-insns @gol
569 -mno-sched-count-spec-in-critical-path}
571 @emph{M32R/D Options}
572 @gccoptlist{-m32r2 -m32rx -m32r @gol
574 -malign-loops -mno-align-loops @gol
575 -missue-rate=@var{number} @gol
576 -mbranch-cost=@var{number} @gol
577 -mmodel=@var{code-size-model-type} @gol
578 -msdata=@var{sdata-type} @gol
579 -mno-flush-func -mflush-func=@var{name} @gol
580 -mno-flush-trap -mflush-trap=@var{number} @gol
584 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
586 @emph{M680x0 Options}
587 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
588 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
589 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
590 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
591 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
592 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
593 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
594 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
596 @emph{M68hc1x Options}
597 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
598 -mauto-incdec -minmax -mlong-calls -mshort @gol
599 -msoft-reg-count=@var{count}}
602 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
603 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
604 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
605 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
606 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
609 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
610 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
611 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
612 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
613 -mfp32 -mfp64 -mhard-float -msoft-float @gol
614 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
615 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
616 -G@var{num} -membedded-data -mno-embedded-data @gol
617 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
618 -msplit-addresses -mno-split-addresses @gol
619 -mexplicit-relocs -mno-explicit-relocs @gol
620 -mcheck-zero-division -mno-check-zero-division @gol
621 -mdivide-traps -mdivide-breaks @gol
622 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
623 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
624 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
625 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
626 -mfix-sb1 -mno-fix-sb1 @gol
627 -mflush-func=@var{func} -mno-flush-func @gol
628 -mbranch-likely -mno-branch-likely @gol
629 -mfp-exceptions -mno-fp-exceptions @gol
630 -mvr4130-align -mno-vr4130-align}
633 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
634 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
635 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
636 -mno-base-addresses -msingle-exit -mno-single-exit}
638 @emph{MN10300 Options}
639 @gccoptlist{-mmult-bug -mno-mult-bug @gol
640 -mam33 -mno-am33 @gol
641 -mam33-2 -mno-am33-2 @gol
642 -mreturn-pointer-on-d0 @gol
646 @gccoptlist{-mno-crt0 -mbacc -msim @gol
647 -march=@var{cpu-type} }
649 @emph{PDP-11 Options}
650 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
651 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
652 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
653 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
654 -mbranch-expensive -mbranch-cheap @gol
655 -msplit -mno-split -munix-asm -mdec-asm}
657 @emph{PowerPC Options}
658 See RS/6000 and PowerPC Options.
660 @emph{RS/6000 and PowerPC Options}
661 @gccoptlist{-mcpu=@var{cpu-type} @gol
662 -mtune=@var{cpu-type} @gol
663 -mpower -mno-power -mpower2 -mno-power2 @gol
664 -mpowerpc -mpowerpc64 -mno-powerpc @gol
665 -maltivec -mno-altivec @gol
666 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
667 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
668 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
669 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mdfp -mno-dfp @gol
670 -mnew-mnemonics -mold-mnemonics @gol
671 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
672 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
673 -malign-power -malign-natural @gol
674 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
675 -mstring -mno-string -mupdate -mno-update @gol
676 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
677 -mstrict-align -mno-strict-align -mrelocatable @gol
678 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
679 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
680 -mdynamic-no-pic -maltivec -mswdiv @gol
681 -mprioritize-restricted-insns=@var{priority} @gol
682 -msched-costly-dep=@var{dependence_type} @gol
683 -minsert-sched-nops=@var{scheme} @gol
684 -mcall-sysv -mcall-netbsd @gol
685 -maix-struct-return -msvr4-struct-return @gol
686 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
687 -misel -mno-isel @gol
688 -misel=yes -misel=no @gol
690 -mspe=yes -mspe=no @gol
691 -mvrsave -mno-vrsave @gol
692 -mmulhw -mno-mulhw @gol
693 -mdlmzb -mno-dlmzb @gol
694 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
695 -mprototype -mno-prototype @gol
696 -msim -mmvme -mads -myellowknife -memb -msdata @gol
697 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
699 @emph{S/390 and zSeries Options}
700 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
701 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
702 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
703 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
704 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
705 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
706 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
709 @gccoptlist{-mel -mel @gol
714 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
715 -m4-nofpu -m4-single-only -m4-single -m4 @gol
716 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
717 -m5-64media -m5-64media-nofpu @gol
718 -m5-32media -m5-32media-nofpu @gol
719 -m5-compact -m5-compact-nofpu @gol
720 -mb -ml -mdalign -mrelax @gol
721 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
722 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
723 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
724 -mdivsi3_libfunc=@var{name} @gol
725 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
729 @gccoptlist{-mcpu=@var{cpu-type} @gol
730 -mtune=@var{cpu-type} @gol
731 -mcmodel=@var{code-model} @gol
732 -m32 -m64 -mapp-regs -mno-app-regs @gol
733 -mfaster-structs -mno-faster-structs @gol
734 -mfpu -mno-fpu -mhard-float -msoft-float @gol
735 -mhard-quad-float -msoft-quad-float @gol
736 -mimpure-text -mno-impure-text -mlittle-endian @gol
737 -mstack-bias -mno-stack-bias @gol
738 -munaligned-doubles -mno-unaligned-doubles @gol
739 -mv8plus -mno-v8plus -mvis -mno-vis
740 -threads -pthreads -pthread}
743 @gccoptlist{-mwarn-reloc -merror-reloc @gol
744 -msafe-dma -munsafe-dma @gol
746 -msmall-mem -mlarge-mem -mstdmain @gol
747 -mfixed-range=@var{register-range}}
749 @emph{System V Options}
750 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
752 @emph{TMS320C3x/C4x Options}
753 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
754 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
755 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
756 -mparallel-insns -mparallel-mpy -mpreserve-float}
759 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
760 -mprolog-function -mno-prolog-function -mspace @gol
761 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
762 -mapp-regs -mno-app-regs @gol
763 -mdisable-callt -mno-disable-callt @gol
769 @gccoptlist{-mg -mgnu -munix}
771 @emph{x86-64 Options}
772 See i386 and x86-64 Options.
774 @emph{Xstormy16 Options}
777 @emph{Xtensa Options}
778 @gccoptlist{-mconst16 -mno-const16 @gol
779 -mfused-madd -mno-fused-madd @gol
780 -mtext-section-literals -mno-text-section-literals @gol
781 -mtarget-align -mno-target-align @gol
782 -mlongcalls -mno-longcalls}
784 @emph{zSeries Options}
785 See S/390 and zSeries Options.
787 @item Code Generation Options
788 @xref{Code Gen Options,,Options for Code Generation Conventions}.
789 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
790 -ffixed-@var{reg} -fexceptions @gol
791 -fnon-call-exceptions -funwind-tables @gol
792 -fasynchronous-unwind-tables @gol
793 -finhibit-size-directive -finstrument-functions @gol
794 -fno-common -fno-ident @gol
795 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
796 -fno-jump-tables @gol
797 -frecord-gcc-switches @gol
798 -freg-struct-return -fshort-enums @gol
799 -fshort-double -fshort-wchar @gol
800 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
801 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
802 -fargument-alias -fargument-noalias @gol
803 -fargument-noalias-global -fargument-noalias-anything
804 -fleading-underscore -ftls-model=@var{model} @gol
805 -ftrapv -fwrapv -fbounds-check @gol
810 * Overall Options:: Controlling the kind of output:
811 an executable, object files, assembler files,
812 or preprocessed source.
813 * C Dialect Options:: Controlling the variant of C language compiled.
814 * C++ Dialect Options:: Variations on C++.
815 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
817 * Language Independent Options:: Controlling how diagnostics should be
819 * Warning Options:: How picky should the compiler be?
820 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
821 * Optimize Options:: How much optimization?
822 * Preprocessor Options:: Controlling header files and macro definitions.
823 Also, getting dependency information for Make.
824 * Assembler Options:: Passing options to the assembler.
825 * Link Options:: Specifying libraries and so on.
826 * Directory Options:: Where to find header files and libraries.
827 Where to find the compiler executable files.
828 * Spec Files:: How to pass switches to sub-processes.
829 * Target Options:: Running a cross-compiler, or an old version of GCC.
832 @node Overall Options
833 @section Options Controlling the Kind of Output
835 Compilation can involve up to four stages: preprocessing, compilation
836 proper, assembly and linking, always in that order. GCC is capable of
837 preprocessing and compiling several files either into several
838 assembler input files, or into one assembler input file; then each
839 assembler input file produces an object file, and linking combines all
840 the object files (those newly compiled, and those specified as input)
841 into an executable file.
843 @cindex file name suffix
844 For any given input file, the file name suffix determines what kind of
849 C source code which must be preprocessed.
852 C source code which should not be preprocessed.
855 C++ source code which should not be preprocessed.
858 Objective-C source code. Note that you must link with the @file{libobjc}
859 library to make an Objective-C program work.
862 Objective-C source code which should not be preprocessed.
866 Objective-C++ source code. Note that you must link with the @file{libobjc}
867 library to make an Objective-C++ program work. Note that @samp{.M} refers
868 to a literal capital M@.
871 Objective-C++ source code which should not be preprocessed.
874 C, C++, Objective-C or Objective-C++ header file to be turned into a
879 @itemx @var{file}.cxx
880 @itemx @var{file}.cpp
881 @itemx @var{file}.CPP
882 @itemx @var{file}.c++
884 C++ source code which must be preprocessed. Note that in @samp{.cxx},
885 the last two letters must both be literally @samp{x}. Likewise,
886 @samp{.C} refers to a literal capital C@.
890 Objective-C++ source code which must be preprocessed.
893 Objective-C++ source code which should not be preprocessed.
897 C++ header file to be turned into a precompiled header.
900 @itemx @var{file}.for
901 @itemx @var{file}.FOR
902 Fixed form Fortran source code which should not be preprocessed.
905 @itemx @var{file}.fpp
906 @itemx @var{file}.FPP
907 Fixed form Fortran source code which must be preprocessed (with the traditional
911 @itemx @var{file}.f95
912 Free form Fortran source code which should not be preprocessed.
915 @itemx @var{file}.F95
916 Free form Fortran source code which must be preprocessed (with the
917 traditional preprocessor).
919 @c FIXME: Descriptions of Java file types.
926 Ada source code file which contains a library unit declaration (a
927 declaration of a package, subprogram, or generic, or a generic
928 instantiation), or a library unit renaming declaration (a package,
929 generic, or subprogram renaming declaration). Such files are also
932 @itemx @var{file}.adb
933 Ada source code file containing a library unit body (a subprogram or
934 package body). Such files are also called @dfn{bodies}.
936 @c GCC also knows about some suffixes for languages not yet included:
947 Assembler code which must be preprocessed.
950 An object file to be fed straight into linking.
951 Any file name with no recognized suffix is treated this way.
955 You can specify the input language explicitly with the @option{-x} option:
958 @item -x @var{language}
959 Specify explicitly the @var{language} for the following input files
960 (rather than letting the compiler choose a default based on the file
961 name suffix). This option applies to all following input files until
962 the next @option{-x} option. Possible values for @var{language} are:
964 c c-header c-cpp-output
965 c++ c++-header c++-cpp-output
966 objective-c objective-c-header objective-c-cpp-output
967 objective-c++ objective-c++-header objective-c++-cpp-output
968 assembler assembler-with-cpp
976 Turn off any specification of a language, so that subsequent files are
977 handled according to their file name suffixes (as they are if @option{-x}
978 has not been used at all).
980 @item -pass-exit-codes
981 @opindex pass-exit-codes
982 Normally the @command{gcc} program will exit with the code of 1 if any
983 phase of the compiler returns a non-success return code. If you specify
984 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
985 numerically highest error produced by any phase that returned an error
986 indication. The C, C++, and Fortran frontends return 4, if an internal
987 compiler error is encountered.
990 If you only want some of the stages of compilation, you can use
991 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
992 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
993 @command{gcc} is to stop. Note that some combinations (for example,
994 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
999 Compile or assemble the source files, but do not link. The linking
1000 stage simply is not done. The ultimate output is in the form of an
1001 object file for each source file.
1003 By default, the object file name for a source file is made by replacing
1004 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1006 Unrecognized input files, not requiring compilation or assembly, are
1011 Stop after the stage of compilation proper; do not assemble. The output
1012 is in the form of an assembler code file for each non-assembler input
1015 By default, the assembler file name for a source file is made by
1016 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1018 Input files that don't require compilation are ignored.
1022 Stop after the preprocessing stage; do not run the compiler proper. The
1023 output is in the form of preprocessed source code, which is sent to the
1026 Input files which don't require preprocessing are ignored.
1028 @cindex output file option
1031 Place output in file @var{file}. This applies regardless to whatever
1032 sort of output is being produced, whether it be an executable file,
1033 an object file, an assembler file or preprocessed C code.
1035 If @option{-o} is not specified, the default is to put an executable
1036 file in @file{a.out}, the object file for
1037 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1038 assembler file in @file{@var{source}.s}, a precompiled header file in
1039 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1044 Print (on standard error output) the commands executed to run the stages
1045 of compilation. Also print the version number of the compiler driver
1046 program and of the preprocessor and the compiler proper.
1050 Like @option{-v} except the commands are not executed and all command
1051 arguments are quoted. This is useful for shell scripts to capture the
1052 driver-generated command lines.
1056 Use pipes rather than temporary files for communication between the
1057 various stages of compilation. This fails to work on some systems where
1058 the assembler is unable to read from a pipe; but the GNU assembler has
1063 If you are compiling multiple source files, this option tells the driver
1064 to pass all the source files to the compiler at once (for those
1065 languages for which the compiler can handle this). This will allow
1066 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1067 language for which this is supported is C@. If you pass source files for
1068 multiple languages to the driver, using this option, the driver will invoke
1069 the compiler(s) that support IMA once each, passing each compiler all the
1070 source files appropriate for it. For those languages that do not support
1071 IMA this option will be ignored, and the compiler will be invoked once for
1072 each source file in that language. If you use this option in conjunction
1073 with @option{-save-temps}, the compiler will generate multiple
1075 (one for each source file), but only one (combined) @file{.o} or
1080 Print (on the standard output) a description of the command line options
1081 understood by @command{gcc}. If the @option{-v} option is also specified
1082 then @option{--help} will also be passed on to the various processes
1083 invoked by @command{gcc}, so that they can display the command line options
1084 they accept. If the @option{-Wextra} option has also been specified
1085 (prior to the @option{--help} option), then command line options which
1086 have no documentation associated with them will also be displayed.
1089 @opindex target-help
1090 Print (on the standard output) a description of target-specific command
1091 line options for each tool.
1093 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1094 Print (on the standard output) a description of the command line
1095 options understood by the compiler that fit into a specific class.
1096 The class can be one of @var{optimizers}, @var{warnings}, @var{target}
1100 @item @var{optimizers}
1101 This will display all of the optimization options supported by the
1104 @item @var{warnings}
1105 This will display all of the options controlling warning messages
1106 produced by the compiler.
1109 This will display target-specific options. Unlike the
1110 @option{--target-help} option however, target-specific options of the
1111 linker and assembler will not be displayed. This is because those
1112 tools do not currently support the extended @option{--help=} syntax.
1115 This will display the values recognized by the @option{--param}
1119 It is possible to further refine the output of the @option{--help=}
1120 option by adding a comma separated list of qualifiers after the
1121 class. These can be any from the following list:
1125 Display only those options which are undocumented.
1128 Display options which take an argument that appears after an equal
1129 sign in the same continuous piece of text, such as:
1130 @samp{--help=target}.
1133 Display options which take an argument that appears as a separate word
1134 following the original option, such as: @samp{-o output-file}.
1137 Thus for example to display all the undocumented target-specific
1138 switches supported by the compiler the following can be used:
1141 --help=target,undocumented
1144 The sense of a qualifier can be inverted by prefixing it with the
1145 @var{^} character, so for example to display all binary warning
1146 options (i.e. ones that are either on or off and that do not take an
1147 argument), which have a description the following can be used:
1150 --help=warnings,^joined,^undocumented
1153 A class can also be used as a qualifier, although this usually
1154 restricts the output by so much that there is nothing to display. One
1155 case where it does work however is when one of the classes is
1156 @var{target}. So for example to display all the target-specific
1157 optimization options the following can be used:
1160 --help=target,optimizers
1163 The @option{--help=} option can be repeated on the command line. Each
1164 successive use will display its requested class of options, skipping
1165 those that have already been displayed.
1167 If the @option{-Q} option appears on the command line before the
1168 @option{--help=} option, then the descriptive text displayed by
1169 @option{--help=} is changed. Instead of describing the displayed
1170 options, an indication is given as to whether the option is enabled,
1171 disabled or set to a specific value (assuming that the compiler
1172 knows this at the point where the @option{--help=} option is used).
1174 Here is a truncated example from the ARM port of @command{gcc}:
1177 % gcc -Q -mabi=2 --help=target -c
1178 The following options are target specific:
1180 -mabort-on-noreturn [disabled]
1184 The output is sensitive to the effects of previous command line
1185 options, so for example it is possible to find out which optimizations
1186 are enabled at @option{-O2} by using:
1189 -O2 --help=optimizers
1192 Alternatively you can discover which binary optimizations are enabled
1193 by @option{-O3} by using:
1196 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1197 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1198 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1203 Display the version number and copyrights of the invoked GCC@.
1205 @include @value{srcdir}/../libiberty/at-file.texi
1209 @section Compiling C++ Programs
1211 @cindex suffixes for C++ source
1212 @cindex C++ source file suffixes
1213 C++ source files conventionally use one of the suffixes @samp{.C},
1214 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1215 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1216 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1217 files with these names and compiles them as C++ programs even if you
1218 call the compiler the same way as for compiling C programs (usually
1219 with the name @command{gcc}).
1223 However, the use of @command{gcc} does not add the C++ library.
1224 @command{g++} is a program that calls GCC and treats @samp{.c},
1225 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1226 files unless @option{-x} is used, and automatically specifies linking
1227 against the C++ library. This program is also useful when
1228 precompiling a C header file with a @samp{.h} extension for use in C++
1229 compilations. On many systems, @command{g++} is also installed with
1230 the name @command{c++}.
1232 @cindex invoking @command{g++}
1233 When you compile C++ programs, you may specify many of the same
1234 command-line options that you use for compiling programs in any
1235 language; or command-line options meaningful for C and related
1236 languages; or options that are meaningful only for C++ programs.
1237 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1238 explanations of options for languages related to C@.
1239 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1240 explanations of options that are meaningful only for C++ programs.
1242 @node C Dialect Options
1243 @section Options Controlling C Dialect
1244 @cindex dialect options
1245 @cindex language dialect options
1246 @cindex options, dialect
1248 The following options control the dialect of C (or languages derived
1249 from C, such as C++, Objective-C and Objective-C++) that the compiler
1253 @cindex ANSI support
1257 In C mode, support all ISO C90 programs. In C++ mode,
1258 remove GNU extensions that conflict with ISO C++.
1260 This turns off certain features of GCC that are incompatible with ISO
1261 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1262 such as the @code{asm} and @code{typeof} keywords, and
1263 predefined macros such as @code{unix} and @code{vax} that identify the
1264 type of system you are using. It also enables the undesirable and
1265 rarely used ISO trigraph feature. For the C compiler,
1266 it disables recognition of C++ style @samp{//} comments as well as
1267 the @code{inline} keyword.
1269 The alternate keywords @code{__asm__}, @code{__extension__},
1270 @code{__inline__} and @code{__typeof__} continue to work despite
1271 @option{-ansi}. You would not want to use them in an ISO C program, of
1272 course, but it is useful to put them in header files that might be included
1273 in compilations done with @option{-ansi}. Alternate predefined macros
1274 such as @code{__unix__} and @code{__vax__} are also available, with or
1275 without @option{-ansi}.
1277 The @option{-ansi} option does not cause non-ISO programs to be
1278 rejected gratuitously. For that, @option{-pedantic} is required in
1279 addition to @option{-ansi}. @xref{Warning Options}.
1281 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1282 option is used. Some header files may notice this macro and refrain
1283 from declaring certain functions or defining certain macros that the
1284 ISO standard doesn't call for; this is to avoid interfering with any
1285 programs that might use these names for other things.
1287 Functions which would normally be built in but do not have semantics
1288 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1289 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1290 built-in functions provided by GCC}, for details of the functions
1295 Determine the language standard. This option is currently only
1296 supported when compiling C or C++. A value for this option must be
1297 provided; possible values are
1302 ISO C90 (same as @option{-ansi}).
1304 @item iso9899:199409
1305 ISO C90 as modified in amendment 1.
1311 ISO C99. Note that this standard is not yet fully supported; see
1312 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1313 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1316 Default, ISO C90 plus GNU extensions (including some C99 features).
1320 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1321 this will become the default. The name @samp{gnu9x} is deprecated.
1324 The 1998 ISO C++ standard plus amendments.
1327 The same as @option{-std=c++98} plus GNU extensions. This is the
1328 default for C++ code.
1331 The working draft of the upcoming ISO C++0x standard. This option
1332 enables experimental features that are likely to be included in
1333 C++0x. The working draft is constantly changing, and any feature that is
1334 enabled by this flag may be removed from future versions of GCC if it is
1335 not part of the C++0x standard.
1338 The same as @option{-std=c++0x} plus GNU extensions. As with
1339 @option{-std=c++0x}, this option enables experimental features that may
1340 be removed in future versions of GCC.
1343 Even when this option is not specified, you can still use some of the
1344 features of newer standards in so far as they do not conflict with
1345 previous C standards. For example, you may use @code{__restrict__} even
1346 when @option{-std=c99} is not specified.
1348 The @option{-std} options specifying some version of ISO C have the same
1349 effects as @option{-ansi}, except that features that were not in ISO C90
1350 but are in the specified version (for example, @samp{//} comments and
1351 the @code{inline} keyword in ISO C99) are not disabled.
1353 @xref{Standards,,Language Standards Supported by GCC}, for details of
1354 these standard versions.
1356 @item -fgnu89-inline
1357 @opindex fgnu89-inline
1358 The option @option{-fgnu89-inline} tells GCC to use the traditional
1359 GNU semantics for @code{inline} functions when in C99 mode.
1360 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1361 is accepted and ignored by GCC versions 4.1.3 up to but not including
1362 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1363 C99 mode. Using this option is roughly equivalent to adding the
1364 @code{gnu_inline} function attribute to all inline functions
1365 (@pxref{Function Attributes}).
1367 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1368 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1369 specifies the default behavior). This option was first supported in
1370 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1372 The preprocesor macros @code{__GNUC_GNU_INLINE__} and
1373 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1374 in effect for @code{inline} functions. @xref{Common Predefined
1375 Macros,,,cpp.info,The C Preprocessor}.
1377 @item -aux-info @var{filename}
1379 Output to the given filename prototyped declarations for all functions
1380 declared and/or defined in a translation unit, including those in header
1381 files. This option is silently ignored in any language other than C@.
1383 Besides declarations, the file indicates, in comments, the origin of
1384 each declaration (source file and line), whether the declaration was
1385 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1386 @samp{O} for old, respectively, in the first character after the line
1387 number and the colon), and whether it came from a declaration or a
1388 definition (@samp{C} or @samp{F}, respectively, in the following
1389 character). In the case of function definitions, a K&R-style list of
1390 arguments followed by their declarations is also provided, inside
1391 comments, after the declaration.
1395 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1396 keyword, so that code can use these words as identifiers. You can use
1397 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1398 instead. @option{-ansi} implies @option{-fno-asm}.
1400 In C++, this switch only affects the @code{typeof} keyword, since
1401 @code{asm} and @code{inline} are standard keywords. You may want to
1402 use the @option{-fno-gnu-keywords} flag instead, which has the same
1403 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1404 switch only affects the @code{asm} and @code{typeof} keywords, since
1405 @code{inline} is a standard keyword in ISO C99.
1408 @itemx -fno-builtin-@var{function}
1409 @opindex fno-builtin
1410 @cindex built-in functions
1411 Don't recognize built-in functions that do not begin with
1412 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1413 functions provided by GCC}, for details of the functions affected,
1414 including those which are not built-in functions when @option{-ansi} or
1415 @option{-std} options for strict ISO C conformance are used because they
1416 do not have an ISO standard meaning.
1418 GCC normally generates special code to handle certain built-in functions
1419 more efficiently; for instance, calls to @code{alloca} may become single
1420 instructions that adjust the stack directly, and calls to @code{memcpy}
1421 may become inline copy loops. The resulting code is often both smaller
1422 and faster, but since the function calls no longer appear as such, you
1423 cannot set a breakpoint on those calls, nor can you change the behavior
1424 of the functions by linking with a different library. In addition,
1425 when a function is recognized as a built-in function, GCC may use
1426 information about that function to warn about problems with calls to
1427 that function, or to generate more efficient code, even if the
1428 resulting code still contains calls to that function. For example,
1429 warnings are given with @option{-Wformat} for bad calls to
1430 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1431 known not to modify global memory.
1433 With the @option{-fno-builtin-@var{function}} option
1434 only the built-in function @var{function} is
1435 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1436 function is named this is not built-in in this version of GCC, this
1437 option is ignored. There is no corresponding
1438 @option{-fbuiltin-@var{function}} option; if you wish to enable
1439 built-in functions selectively when using @option{-fno-builtin} or
1440 @option{-ffreestanding}, you may define macros such as:
1443 #define abs(n) __builtin_abs ((n))
1444 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1449 @cindex hosted environment
1451 Assert that compilation takes place in a hosted environment. This implies
1452 @option{-fbuiltin}. A hosted environment is one in which the
1453 entire standard library is available, and in which @code{main} has a return
1454 type of @code{int}. Examples are nearly everything except a kernel.
1455 This is equivalent to @option{-fno-freestanding}.
1457 @item -ffreestanding
1458 @opindex ffreestanding
1459 @cindex hosted environment
1461 Assert that compilation takes place in a freestanding environment. This
1462 implies @option{-fno-builtin}. A freestanding environment
1463 is one in which the standard library may not exist, and program startup may
1464 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1465 This is equivalent to @option{-fno-hosted}.
1467 @xref{Standards,,Language Standards Supported by GCC}, for details of
1468 freestanding and hosted environments.
1472 @cindex openmp parallel
1473 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1474 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1475 compiler generates parallel code according to the OpenMP Application
1476 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1478 @item -fms-extensions
1479 @opindex fms-extensions
1480 Accept some non-standard constructs used in Microsoft header files.
1482 Some cases of unnamed fields in structures and unions are only
1483 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1484 fields within structs/unions}, for details.
1488 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1489 options for strict ISO C conformance) implies @option{-trigraphs}.
1491 @item -no-integrated-cpp
1492 @opindex no-integrated-cpp
1493 Performs a compilation in two passes: preprocessing and compiling. This
1494 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1495 @option{-B} option. The user supplied compilation step can then add in
1496 an additional preprocessing step after normal preprocessing but before
1497 compiling. The default is to use the integrated cpp (internal cpp)
1499 The semantics of this option will change if "cc1", "cc1plus", and
1500 "cc1obj" are merged.
1502 @cindex traditional C language
1503 @cindex C language, traditional
1505 @itemx -traditional-cpp
1506 @opindex traditional-cpp
1507 @opindex traditional
1508 Formerly, these options caused GCC to attempt to emulate a pre-standard
1509 C compiler. They are now only supported with the @option{-E} switch.
1510 The preprocessor continues to support a pre-standard mode. See the GNU
1511 CPP manual for details.
1513 @item -fcond-mismatch
1514 @opindex fcond-mismatch
1515 Allow conditional expressions with mismatched types in the second and
1516 third arguments. The value of such an expression is void. This option
1517 is not supported for C++.
1519 @item -flax-vector-conversions
1520 @opindex flax-vector-conversions
1521 Allow implicit conversions between vectors with differing numbers of
1522 elements and/or incompatible element types. This option should not be
1525 @item -funsigned-char
1526 @opindex funsigned-char
1527 Let the type @code{char} be unsigned, like @code{unsigned char}.
1529 Each kind of machine has a default for what @code{char} should
1530 be. It is either like @code{unsigned char} by default or like
1531 @code{signed char} by default.
1533 Ideally, a portable program should always use @code{signed char} or
1534 @code{unsigned char} when it depends on the signedness of an object.
1535 But many programs have been written to use plain @code{char} and
1536 expect it to be signed, or expect it to be unsigned, depending on the
1537 machines they were written for. This option, and its inverse, let you
1538 make such a program work with the opposite default.
1540 The type @code{char} is always a distinct type from each of
1541 @code{signed char} or @code{unsigned char}, even though its behavior
1542 is always just like one of those two.
1545 @opindex fsigned-char
1546 Let the type @code{char} be signed, like @code{signed char}.
1548 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1549 the negative form of @option{-funsigned-char}. Likewise, the option
1550 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1552 @item -fsigned-bitfields
1553 @itemx -funsigned-bitfields
1554 @itemx -fno-signed-bitfields
1555 @itemx -fno-unsigned-bitfields
1556 @opindex fsigned-bitfields
1557 @opindex funsigned-bitfields
1558 @opindex fno-signed-bitfields
1559 @opindex fno-unsigned-bitfields
1560 These options control whether a bit-field is signed or unsigned, when the
1561 declaration does not use either @code{signed} or @code{unsigned}. By
1562 default, such a bit-field is signed, because this is consistent: the
1563 basic integer types such as @code{int} are signed types.
1566 @node C++ Dialect Options
1567 @section Options Controlling C++ Dialect
1569 @cindex compiler options, C++
1570 @cindex C++ options, command line
1571 @cindex options, C++
1572 This section describes the command-line options that are only meaningful
1573 for C++ programs; but you can also use most of the GNU compiler options
1574 regardless of what language your program is in. For example, you
1575 might compile a file @code{firstClass.C} like this:
1578 g++ -g -frepo -O -c firstClass.C
1582 In this example, only @option{-frepo} is an option meant
1583 only for C++ programs; you can use the other options with any
1584 language supported by GCC@.
1586 Here is a list of options that are @emph{only} for compiling C++ programs:
1590 @item -fabi-version=@var{n}
1591 @opindex fabi-version
1592 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1593 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1594 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1595 the version that conforms most closely to the C++ ABI specification.
1596 Therefore, the ABI obtained using version 0 will change as ABI bugs
1599 The default is version 2.
1601 @item -fno-access-control
1602 @opindex fno-access-control
1603 Turn off all access checking. This switch is mainly useful for working
1604 around bugs in the access control code.
1608 Check that the pointer returned by @code{operator new} is non-null
1609 before attempting to modify the storage allocated. This check is
1610 normally unnecessary because the C++ standard specifies that
1611 @code{operator new} will only return @code{0} if it is declared
1612 @samp{throw()}, in which case the compiler will always check the
1613 return value even without this option. In all other cases, when
1614 @code{operator new} has a non-empty exception specification, memory
1615 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1616 @samp{new (nothrow)}.
1618 @item -fconserve-space
1619 @opindex fconserve-space
1620 Put uninitialized or runtime-initialized global variables into the
1621 common segment, as C does. This saves space in the executable at the
1622 cost of not diagnosing duplicate definitions. If you compile with this
1623 flag and your program mysteriously crashes after @code{main()} has
1624 completed, you may have an object that is being destroyed twice because
1625 two definitions were merged.
1627 This option is no longer useful on most targets, now that support has
1628 been added for putting variables into BSS without making them common.
1630 @item -ffriend-injection
1631 @opindex ffriend-injection
1632 Inject friend functions into the enclosing namespace, so that they are
1633 visible outside the scope of the class in which they are declared.
1634 Friend functions were documented to work this way in the old Annotated
1635 C++ Reference Manual, and versions of G++ before 4.1 always worked
1636 that way. However, in ISO C++ a friend function which is not declared
1637 in an enclosing scope can only be found using argument dependent
1638 lookup. This option causes friends to be injected as they were in
1641 This option is for compatibility, and may be removed in a future
1644 @item -fno-elide-constructors
1645 @opindex fno-elide-constructors
1646 The C++ standard allows an implementation to omit creating a temporary
1647 which is only used to initialize another object of the same type.
1648 Specifying this option disables that optimization, and forces G++ to
1649 call the copy constructor in all cases.
1651 @item -fno-enforce-eh-specs
1652 @opindex fno-enforce-eh-specs
1653 Don't generate code to check for violation of exception specifications
1654 at runtime. This option violates the C++ standard, but may be useful
1655 for reducing code size in production builds, much like defining
1656 @samp{NDEBUG}. This does not give user code permission to throw
1657 exceptions in violation of the exception specifications; the compiler
1658 will still optimize based on the specifications, so throwing an
1659 unexpected exception will result in undefined behavior.
1662 @itemx -fno-for-scope
1664 @opindex fno-for-scope
1665 If @option{-ffor-scope} is specified, the scope of variables declared in
1666 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1667 as specified by the C++ standard.
1668 If @option{-fno-for-scope} is specified, the scope of variables declared in
1669 a @i{for-init-statement} extends to the end of the enclosing scope,
1670 as was the case in old versions of G++, and other (traditional)
1671 implementations of C++.
1673 The default if neither flag is given to follow the standard,
1674 but to allow and give a warning for old-style code that would
1675 otherwise be invalid, or have different behavior.
1677 @item -fno-gnu-keywords
1678 @opindex fno-gnu-keywords
1679 Do not recognize @code{typeof} as a keyword, so that code can use this
1680 word as an identifier. You can use the keyword @code{__typeof__} instead.
1681 @option{-ansi} implies @option{-fno-gnu-keywords}.
1683 @item -fno-implicit-templates
1684 @opindex fno-implicit-templates
1685 Never emit code for non-inline templates which are instantiated
1686 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1687 @xref{Template Instantiation}, for more information.
1689 @item -fno-implicit-inline-templates
1690 @opindex fno-implicit-inline-templates
1691 Don't emit code for implicit instantiations of inline templates, either.
1692 The default is to handle inlines differently so that compiles with and
1693 without optimization will need the same set of explicit instantiations.
1695 @item -fno-implement-inlines
1696 @opindex fno-implement-inlines
1697 To save space, do not emit out-of-line copies of inline functions
1698 controlled by @samp{#pragma implementation}. This will cause linker
1699 errors if these functions are not inlined everywhere they are called.
1701 @item -fms-extensions
1702 @opindex fms-extensions
1703 Disable pedantic warnings about constructs used in MFC, such as implicit
1704 int and getting a pointer to member function via non-standard syntax.
1706 @item -fno-nonansi-builtins
1707 @opindex fno-nonansi-builtins
1708 Disable built-in declarations of functions that are not mandated by
1709 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1710 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1712 @item -fno-operator-names
1713 @opindex fno-operator-names
1714 Do not treat the operator name keywords @code{and}, @code{bitand},
1715 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1716 synonyms as keywords.
1718 @item -fno-optional-diags
1719 @opindex fno-optional-diags
1720 Disable diagnostics that the standard says a compiler does not need to
1721 issue. Currently, the only such diagnostic issued by G++ is the one for
1722 a name having multiple meanings within a class.
1725 @opindex fpermissive
1726 Downgrade some diagnostics about nonconformant code from errors to
1727 warnings. Thus, using @option{-fpermissive} will allow some
1728 nonconforming code to compile.
1732 Enable automatic template instantiation at link time. This option also
1733 implies @option{-fno-implicit-templates}. @xref{Template
1734 Instantiation}, for more information.
1738 Disable generation of information about every class with virtual
1739 functions for use by the C++ runtime type identification features
1740 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1741 of the language, you can save some space by using this flag. Note that
1742 exception handling uses the same information, but it will generate it as
1743 needed. The @samp{dynamic_cast} operator can still be used for casts that
1744 do not require runtime type information, i.e. casts to @code{void *} or to
1745 unambiguous base classes.
1749 Emit statistics about front-end processing at the end of the compilation.
1750 This information is generally only useful to the G++ development team.
1752 @item -ftemplate-depth-@var{n}
1753 @opindex ftemplate-depth
1754 Set the maximum instantiation depth for template classes to @var{n}.
1755 A limit on the template instantiation depth is needed to detect
1756 endless recursions during template class instantiation. ANSI/ISO C++
1757 conforming programs must not rely on a maximum depth greater than 17.
1759 @item -fno-threadsafe-statics
1760 @opindex fno-threadsafe-statics
1761 Do not emit the extra code to use the routines specified in the C++
1762 ABI for thread-safe initialization of local statics. You can use this
1763 option to reduce code size slightly in code that doesn't need to be
1766 @item -fuse-cxa-atexit
1767 @opindex fuse-cxa-atexit
1768 Register destructors for objects with static storage duration with the
1769 @code{__cxa_atexit} function rather than the @code{atexit} function.
1770 This option is required for fully standards-compliant handling of static
1771 destructors, but will only work if your C library supports
1772 @code{__cxa_atexit}.
1774 @item -fno-use-cxa-get-exception-ptr
1775 @opindex fno-use-cxa-get-exception-ptr
1776 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1777 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1778 if the runtime routine is not available.
1780 @item -fvisibility-inlines-hidden
1781 @opindex fvisibility-inlines-hidden
1782 This switch declares that the user does not attempt to compare
1783 pointers to inline methods where the addresses of the two functions
1784 were taken in different shared objects.
1786 The effect of this is that GCC may, effectively, mark inline methods with
1787 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1788 appear in the export table of a DSO and do not require a PLT indirection
1789 when used within the DSO@. Enabling this option can have a dramatic effect
1790 on load and link times of a DSO as it massively reduces the size of the
1791 dynamic export table when the library makes heavy use of templates.
1793 The behavior of this switch is not quite the same as marking the
1794 methods as hidden directly, because it does not affect static variables
1795 local to the function or cause the compiler to deduce that
1796 the function is defined in only one shared object.
1798 You may mark a method as having a visibility explicitly to negate the
1799 effect of the switch for that method. For example, if you do want to
1800 compare pointers to a particular inline method, you might mark it as
1801 having default visibility. Marking the enclosing class with explicit
1802 visibility will have no effect.
1804 Explicitly instantiated inline methods are unaffected by this option
1805 as their linkage might otherwise cross a shared library boundary.
1806 @xref{Template Instantiation}.
1810 Do not use weak symbol support, even if it is provided by the linker.
1811 By default, G++ will use weak symbols if they are available. This
1812 option exists only for testing, and should not be used by end-users;
1813 it will result in inferior code and has no benefits. This option may
1814 be removed in a future release of G++.
1818 Do not search for header files in the standard directories specific to
1819 C++, but do still search the other standard directories. (This option
1820 is used when building the C++ library.)
1823 In addition, these optimization, warning, and code generation options
1824 have meanings only for C++ programs:
1827 @item -fno-default-inline
1828 @opindex fno-default-inline
1829 Do not assume @samp{inline} for functions defined inside a class scope.
1830 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1831 functions will have linkage like inline functions; they just won't be
1834 @item -Wabi @r{(C++ only)}
1836 Warn when G++ generates code that is probably not compatible with the
1837 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1838 all such cases, there are probably some cases that are not warned about,
1839 even though G++ is generating incompatible code. There may also be
1840 cases where warnings are emitted even though the code that is generated
1843 You should rewrite your code to avoid these warnings if you are
1844 concerned about the fact that code generated by G++ may not be binary
1845 compatible with code generated by other compilers.
1847 The known incompatibilities at this point include:
1852 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1853 pack data into the same byte as a base class. For example:
1856 struct A @{ virtual void f(); int f1 : 1; @};
1857 struct B : public A @{ int f2 : 1; @};
1861 In this case, G++ will place @code{B::f2} into the same byte
1862 as@code{A::f1}; other compilers will not. You can avoid this problem
1863 by explicitly padding @code{A} so that its size is a multiple of the
1864 byte size on your platform; that will cause G++ and other compilers to
1865 layout @code{B} identically.
1868 Incorrect handling of tail-padding for virtual bases. G++ does not use
1869 tail padding when laying out virtual bases. For example:
1872 struct A @{ virtual void f(); char c1; @};
1873 struct B @{ B(); char c2; @};
1874 struct C : public A, public virtual B @{@};
1878 In this case, G++ will not place @code{B} into the tail-padding for
1879 @code{A}; other compilers will. You can avoid this problem by
1880 explicitly padding @code{A} so that its size is a multiple of its
1881 alignment (ignoring virtual base classes); that will cause G++ and other
1882 compilers to layout @code{C} identically.
1885 Incorrect handling of bit-fields with declared widths greater than that
1886 of their underlying types, when the bit-fields appear in a union. For
1890 union U @{ int i : 4096; @};
1894 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1895 union too small by the number of bits in an @code{int}.
1898 Empty classes can be placed at incorrect offsets. For example:
1908 struct C : public B, public A @{@};
1912 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1913 it should be placed at offset zero. G++ mistakenly believes that the
1914 @code{A} data member of @code{B} is already at offset zero.
1917 Names of template functions whose types involve @code{typename} or
1918 template template parameters can be mangled incorrectly.
1921 template <typename Q>
1922 void f(typename Q::X) @{@}
1924 template <template <typename> class Q>
1925 void f(typename Q<int>::X) @{@}
1929 Instantiations of these templates may be mangled incorrectly.
1933 @item -Wctor-dtor-privacy @r{(C++ only)}
1934 @opindex Wctor-dtor-privacy
1935 Warn when a class seems unusable because all the constructors or
1936 destructors in that class are private, and it has neither friends nor
1937 public static member functions.
1939 @item -Wnon-virtual-dtor @r{(C++ only)}
1940 @opindex Wnon-virtual-dtor
1941 Warn when a class appears to be polymorphic, thereby requiring a virtual
1942 destructor, yet it declares a non-virtual one. This warning is also
1943 enabled if -Weffc++ is specified.
1945 @item -Wreorder @r{(C++ only)}
1947 @cindex reordering, warning
1948 @cindex warning for reordering of member initializers
1949 Warn when the order of member initializers given in the code does not
1950 match the order in which they must be executed. For instance:
1956 A(): j (0), i (1) @{ @}
1960 The compiler will rearrange the member initializers for @samp{i}
1961 and @samp{j} to match the declaration order of the members, emitting
1962 a warning to that effect. This warning is enabled by @option{-Wall}.
1965 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1968 @item -Weffc++ @r{(C++ only)}
1970 Warn about violations of the following style guidelines from Scott Meyers'
1971 @cite{Effective C++} book:
1975 Item 11: Define a copy constructor and an assignment operator for classes
1976 with dynamically allocated memory.
1979 Item 12: Prefer initialization to assignment in constructors.
1982 Item 14: Make destructors virtual in base classes.
1985 Item 15: Have @code{operator=} return a reference to @code{*this}.
1988 Item 23: Don't try to return a reference when you must return an object.
1992 Also warn about violations of the following style guidelines from
1993 Scott Meyers' @cite{More Effective C++} book:
1997 Item 6: Distinguish between prefix and postfix forms of increment and
1998 decrement operators.
2001 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2005 When selecting this option, be aware that the standard library
2006 headers do not obey all of these guidelines; use @samp{grep -v}
2007 to filter out those warnings.
2009 @item -Wno-deprecated @r{(C++ only)}
2010 @opindex Wno-deprecated
2011 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
2013 @item -Wstrict-null-sentinel @r{(C++ only)}
2014 @opindex Wstrict-null-sentinel
2015 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2016 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2017 to @code{__null}. Although it is a null pointer constant not a null pointer,
2018 it is guaranteed to of the same size as a pointer. But this use is
2019 not portable across different compilers.
2021 @item -Wno-non-template-friend @r{(C++ only)}
2022 @opindex Wno-non-template-friend
2023 Disable warnings when non-templatized friend functions are declared
2024 within a template. Since the advent of explicit template specification
2025 support in G++, if the name of the friend is an unqualified-id (i.e.,
2026 @samp{friend foo(int)}), the C++ language specification demands that the
2027 friend declare or define an ordinary, nontemplate function. (Section
2028 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2029 could be interpreted as a particular specialization of a templatized
2030 function. Because this non-conforming behavior is no longer the default
2031 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2032 check existing code for potential trouble spots and is on by default.
2033 This new compiler behavior can be turned off with
2034 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2035 but disables the helpful warning.
2037 @item -Wold-style-cast @r{(C++ only)}
2038 @opindex Wold-style-cast
2039 Warn if an old-style (C-style) cast to a non-void type is used within
2040 a C++ program. The new-style casts (@samp{dynamic_cast},
2041 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2042 less vulnerable to unintended effects and much easier to search for.
2044 @item -Woverloaded-virtual @r{(C++ only)}
2045 @opindex Woverloaded-virtual
2046 @cindex overloaded virtual fn, warning
2047 @cindex warning for overloaded virtual fn
2048 Warn when a function declaration hides virtual functions from a
2049 base class. For example, in:
2056 struct B: public A @{
2061 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2069 will fail to compile.
2071 @item -Wno-pmf-conversions @r{(C++ only)}
2072 @opindex Wno-pmf-conversions
2073 Disable the diagnostic for converting a bound pointer to member function
2076 @item -Wsign-promo @r{(C++ only)}
2077 @opindex Wsign-promo
2078 Warn when overload resolution chooses a promotion from unsigned or
2079 enumerated type to a signed type, over a conversion to an unsigned type of
2080 the same size. Previous versions of G++ would try to preserve
2081 unsignedness, but the standard mandates the current behavior.
2086 A& operator = (int);
2096 In this example, G++ will synthesize a default @samp{A& operator =
2097 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2100 @node Objective-C and Objective-C++ Dialect Options
2101 @section Options Controlling Objective-C and Objective-C++ Dialects
2103 @cindex compiler options, Objective-C and Objective-C++
2104 @cindex Objective-C and Objective-C++ options, command line
2105 @cindex options, Objective-C and Objective-C++
2106 (NOTE: This manual does not describe the Objective-C and Objective-C++
2107 languages themselves. See @xref{Standards,,Language Standards
2108 Supported by GCC}, for references.)
2110 This section describes the command-line options that are only meaningful
2111 for Objective-C and Objective-C++ programs, but you can also use most of
2112 the language-independent GNU compiler options.
2113 For example, you might compile a file @code{some_class.m} like this:
2116 gcc -g -fgnu-runtime -O -c some_class.m
2120 In this example, @option{-fgnu-runtime} is an option meant only for
2121 Objective-C and Objective-C++ programs; you can use the other options with
2122 any language supported by GCC@.
2124 Note that since Objective-C is an extension of the C language, Objective-C
2125 compilations may also use options specific to the C front-end (e.g.,
2126 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2127 C++-specific options (e.g., @option{-Wabi}).
2129 Here is a list of options that are @emph{only} for compiling Objective-C
2130 and Objective-C++ programs:
2133 @item -fconstant-string-class=@var{class-name}
2134 @opindex fconstant-string-class
2135 Use @var{class-name} as the name of the class to instantiate for each
2136 literal string specified with the syntax @code{@@"@dots{}"}. The default
2137 class name is @code{NXConstantString} if the GNU runtime is being used, and
2138 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2139 @option{-fconstant-cfstrings} option, if also present, will override the
2140 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2141 to be laid out as constant CoreFoundation strings.
2144 @opindex fgnu-runtime
2145 Generate object code compatible with the standard GNU Objective-C
2146 runtime. This is the default for most types of systems.
2148 @item -fnext-runtime
2149 @opindex fnext-runtime
2150 Generate output compatible with the NeXT runtime. This is the default
2151 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2152 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2155 @item -fno-nil-receivers
2156 @opindex fno-nil-receivers
2157 Assume that all Objective-C message dispatches (e.g.,
2158 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2159 is not @code{nil}. This allows for more efficient entry points in the runtime
2160 to be used. Currently, this option is only available in conjunction with
2161 the NeXT runtime on Mac OS X 10.3 and later.
2163 @item -fobjc-call-cxx-cdtors
2164 @opindex fobjc-call-cxx-cdtors
2165 For each Objective-C class, check if any of its instance variables is a
2166 C++ object with a non-trivial default constructor. If so, synthesize a
2167 special @code{- (id) .cxx_construct} instance method that will run
2168 non-trivial default constructors on any such instance variables, in order,
2169 and then return @code{self}. Similarly, check if any instance variable
2170 is a C++ object with a non-trivial destructor, and if so, synthesize a
2171 special @code{- (void) .cxx_destruct} method that will run
2172 all such default destructors, in reverse order.
2174 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2175 thusly generated will only operate on instance variables declared in the
2176 current Objective-C class, and not those inherited from superclasses. It
2177 is the responsibility of the Objective-C runtime to invoke all such methods
2178 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2179 will be invoked by the runtime immediately after a new object
2180 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2181 be invoked immediately before the runtime deallocates an object instance.
2183 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2184 support for invoking the @code{- (id) .cxx_construct} and
2185 @code{- (void) .cxx_destruct} methods.
2187 @item -fobjc-direct-dispatch
2188 @opindex fobjc-direct-dispatch
2189 Allow fast jumps to the message dispatcher. On Darwin this is
2190 accomplished via the comm page.
2192 @item -fobjc-exceptions
2193 @opindex fobjc-exceptions
2194 Enable syntactic support for structured exception handling in Objective-C,
2195 similar to what is offered by C++ and Java. This option is
2196 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2205 @@catch (AnObjCClass *exc) @{
2212 @@catch (AnotherClass *exc) @{
2215 @@catch (id allOthers) @{
2225 The @code{@@throw} statement may appear anywhere in an Objective-C or
2226 Objective-C++ program; when used inside of a @code{@@catch} block, the
2227 @code{@@throw} may appear without an argument (as shown above), in which case
2228 the object caught by the @code{@@catch} will be rethrown.
2230 Note that only (pointers to) Objective-C objects may be thrown and
2231 caught using this scheme. When an object is thrown, it will be caught
2232 by the nearest @code{@@catch} clause capable of handling objects of that type,
2233 analogously to how @code{catch} blocks work in C++ and Java. A
2234 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2235 any and all Objective-C exceptions not caught by previous @code{@@catch}
2238 The @code{@@finally} clause, if present, will be executed upon exit from the
2239 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2240 regardless of whether any exceptions are thrown, caught or rethrown
2241 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2242 of the @code{finally} clause in Java.
2244 There are several caveats to using the new exception mechanism:
2248 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2249 idioms provided by the @code{NSException} class, the new
2250 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2251 systems, due to additional functionality needed in the (NeXT) Objective-C
2255 As mentioned above, the new exceptions do not support handling
2256 types other than Objective-C objects. Furthermore, when used from
2257 Objective-C++, the Objective-C exception model does not interoperate with C++
2258 exceptions at this time. This means you cannot @code{@@throw} an exception
2259 from Objective-C and @code{catch} it in C++, or vice versa
2260 (i.e., @code{throw @dots{} @@catch}).
2263 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2264 blocks for thread-safe execution:
2267 @@synchronized (ObjCClass *guard) @{
2272 Upon entering the @code{@@synchronized} block, a thread of execution shall
2273 first check whether a lock has been placed on the corresponding @code{guard}
2274 object by another thread. If it has, the current thread shall wait until
2275 the other thread relinquishes its lock. Once @code{guard} becomes available,
2276 the current thread will place its own lock on it, execute the code contained in
2277 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2278 making @code{guard} available to other threads).
2280 Unlike Java, Objective-C does not allow for entire methods to be marked
2281 @code{@@synchronized}. Note that throwing exceptions out of
2282 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2283 to be unlocked properly.
2287 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2289 @item -freplace-objc-classes
2290 @opindex freplace-objc-classes
2291 Emit a special marker instructing @command{ld(1)} not to statically link in
2292 the resulting object file, and allow @command{dyld(1)} to load it in at
2293 run time instead. This is used in conjunction with the Fix-and-Continue
2294 debugging mode, where the object file in question may be recompiled and
2295 dynamically reloaded in the course of program execution, without the need
2296 to restart the program itself. Currently, Fix-and-Continue functionality
2297 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2302 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2303 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2304 compile time) with static class references that get initialized at load time,
2305 which improves run-time performance. Specifying the @option{-fzero-link} flag
2306 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2307 to be retained. This is useful in Zero-Link debugging mode, since it allows
2308 for individual class implementations to be modified during program execution.
2312 Dump interface declarations for all classes seen in the source file to a
2313 file named @file{@var{sourcename}.decl}.
2315 @item -Wassign-intercept
2316 @opindex Wassign-intercept
2317 Warn whenever an Objective-C assignment is being intercepted by the
2321 @opindex Wno-protocol
2322 If a class is declared to implement a protocol, a warning is issued for
2323 every method in the protocol that is not implemented by the class. The
2324 default behavior is to issue a warning for every method not explicitly
2325 implemented in the class, even if a method implementation is inherited
2326 from the superclass. If you use the @option{-Wno-protocol} option, then
2327 methods inherited from the superclass are considered to be implemented,
2328 and no warning is issued for them.
2332 Warn if multiple methods of different types for the same selector are
2333 found during compilation. The check is performed on the list of methods
2334 in the final stage of compilation. Additionally, a check is performed
2335 for each selector appearing in a @code{@@selector(@dots{})}
2336 expression, and a corresponding method for that selector has been found
2337 during compilation. Because these checks scan the method table only at
2338 the end of compilation, these warnings are not produced if the final
2339 stage of compilation is not reached, for example because an error is
2340 found during compilation, or because the @option{-fsyntax-only} option is
2343 @item -Wstrict-selector-match
2344 @opindex Wstrict-selector-match
2345 Warn if multiple methods with differing argument and/or return types are
2346 found for a given selector when attempting to send a message using this
2347 selector to a receiver of type @code{id} or @code{Class}. When this flag
2348 is off (which is the default behavior), the compiler will omit such warnings
2349 if any differences found are confined to types which share the same size
2352 @item -Wundeclared-selector
2353 @opindex Wundeclared-selector
2354 Warn if a @code{@@selector(@dots{})} expression referring to an
2355 undeclared selector is found. A selector is considered undeclared if no
2356 method with that name has been declared before the
2357 @code{@@selector(@dots{})} expression, either explicitly in an
2358 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2359 an @code{@@implementation} section. This option always performs its
2360 checks as soon as a @code{@@selector(@dots{})} expression is found,
2361 while @option{-Wselector} only performs its checks in the final stage of
2362 compilation. This also enforces the coding style convention
2363 that methods and selectors must be declared before being used.
2365 @item -print-objc-runtime-info
2366 @opindex print-objc-runtime-info
2367 Generate C header describing the largest structure that is passed by
2372 @node Language Independent Options
2373 @section Options to Control Diagnostic Messages Formatting
2374 @cindex options to control diagnostics formatting
2375 @cindex diagnostic messages
2376 @cindex message formatting
2378 Traditionally, diagnostic messages have been formatted irrespective of
2379 the output device's aspect (e.g.@: its width, @dots{}). The options described
2380 below can be used to control the diagnostic messages formatting
2381 algorithm, e.g.@: how many characters per line, how often source location
2382 information should be reported. Right now, only the C++ front end can
2383 honor these options. However it is expected, in the near future, that
2384 the remaining front ends would be able to digest them correctly.
2387 @item -fmessage-length=@var{n}
2388 @opindex fmessage-length
2389 Try to format error messages so that they fit on lines of about @var{n}
2390 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2391 the front ends supported by GCC@. If @var{n} is zero, then no
2392 line-wrapping will be done; each error message will appear on a single
2395 @opindex fdiagnostics-show-location
2396 @item -fdiagnostics-show-location=once
2397 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2398 reporter to emit @emph{once} source location information; that is, in
2399 case the message is too long to fit on a single physical line and has to
2400 be wrapped, the source location won't be emitted (as prefix) again,
2401 over and over, in subsequent continuation lines. This is the default
2404 @item -fdiagnostics-show-location=every-line
2405 Only meaningful in line-wrapping mode. Instructs the diagnostic
2406 messages reporter to emit the same source location information (as
2407 prefix) for physical lines that result from the process of breaking
2408 a message which is too long to fit on a single line.
2410 @item -fdiagnostics-show-option
2411 @opindex fdiagnostics-show-option
2412 This option instructs the diagnostic machinery to add text to each
2413 diagnostic emitted, which indicates which command line option directly
2414 controls that diagnostic, when such an option is known to the
2415 diagnostic machinery.
2417 @item -Wcoverage-mismatch
2418 @opindex Wcoverage-mismatch
2419 Warn if feedback profiles do not match when using the
2420 @option{-fprofile-use} option.
2421 If a source file was changed between @option{-fprofile-gen} and
2422 @option{-fprofile-use}, the files with the profile feedback can fail
2423 to match the source file and GCC can not use the profile feedback
2424 information. By default, GCC emits an error message in this case.
2425 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2426 error. GCC does not use appropriate feedback profiles, so using this
2427 option can result in poorly optimized code. This option is useful
2428 only in the case of very minor changes such as bug fixes to an
2433 @node Warning Options
2434 @section Options to Request or Suppress Warnings
2435 @cindex options to control warnings
2436 @cindex warning messages
2437 @cindex messages, warning
2438 @cindex suppressing warnings
2440 Warnings are diagnostic messages that report constructions which
2441 are not inherently erroneous but which are risky or suggest there
2442 may have been an error.
2444 You can request many specific warnings with options beginning @samp{-W},
2445 for example @option{-Wimplicit} to request warnings on implicit
2446 declarations. Each of these specific warning options also has a
2447 negative form beginning @samp{-Wno-} to turn off warnings;
2448 for example, @option{-Wno-implicit}. This manual lists only one of the
2449 two forms, whichever is not the default.
2451 The following options control the amount and kinds of warnings produced
2452 by GCC; for further, language-specific options also refer to
2453 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2457 @cindex syntax checking
2459 @opindex fsyntax-only
2460 Check the code for syntax errors, but don't do anything beyond that.
2464 Issue all the warnings demanded by strict ISO C and ISO C++;
2465 reject all programs that use forbidden extensions, and some other
2466 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2467 version of the ISO C standard specified by any @option{-std} option used.
2469 Valid ISO C and ISO C++ programs should compile properly with or without
2470 this option (though a rare few will require @option{-ansi} or a
2471 @option{-std} option specifying the required version of ISO C)@. However,
2472 without this option, certain GNU extensions and traditional C and C++
2473 features are supported as well. With this option, they are rejected.
2475 @option{-pedantic} does not cause warning messages for use of the
2476 alternate keywords whose names begin and end with @samp{__}. Pedantic
2477 warnings are also disabled in the expression that follows
2478 @code{__extension__}. However, only system header files should use
2479 these escape routes; application programs should avoid them.
2480 @xref{Alternate Keywords}.
2482 Some users try to use @option{-pedantic} to check programs for strict ISO
2483 C conformance. They soon find that it does not do quite what they want:
2484 it finds some non-ISO practices, but not all---only those for which
2485 ISO C @emph{requires} a diagnostic, and some others for which
2486 diagnostics have been added.
2488 A feature to report any failure to conform to ISO C might be useful in
2489 some instances, but would require considerable additional work and would
2490 be quite different from @option{-pedantic}. We don't have plans to
2491 support such a feature in the near future.
2493 Where the standard specified with @option{-std} represents a GNU
2494 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2495 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2496 extended dialect is based. Warnings from @option{-pedantic} are given
2497 where they are required by the base standard. (It would not make sense
2498 for such warnings to be given only for features not in the specified GNU
2499 C dialect, since by definition the GNU dialects of C include all
2500 features the compiler supports with the given option, and there would be
2501 nothing to warn about.)
2503 @item -pedantic-errors
2504 @opindex pedantic-errors
2505 Like @option{-pedantic}, except that errors are produced rather than
2510 Inhibit all warning messages.
2514 Inhibit warning messages about the use of @samp{#import}.
2516 @item -Wchar-subscripts
2517 @opindex Wchar-subscripts
2518 Warn if an array subscript has type @code{char}. This is a common cause
2519 of error, as programmers often forget that this type is signed on some
2521 This warning is enabled by @option{-Wall}.
2525 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2526 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2527 This warning is enabled by @option{-Wall}.
2529 @item -Wfatal-errors
2530 @opindex Wfatal-errors
2531 This option causes the compiler to abort compilation on the first error
2532 occurred rather than trying to keep going and printing further error
2537 @opindex ffreestanding
2538 @opindex fno-builtin
2539 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2540 the arguments supplied have types appropriate to the format string
2541 specified, and that the conversions specified in the format string make
2542 sense. This includes standard functions, and others specified by format
2543 attributes (@pxref{Function Attributes}), in the @code{printf},
2544 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2545 not in the C standard) families (or other target-specific families).
2546 Which functions are checked without format attributes having been
2547 specified depends on the standard version selected, and such checks of
2548 functions without the attribute specified are disabled by
2549 @option{-ffreestanding} or @option{-fno-builtin}.
2551 The formats are checked against the format features supported by GNU
2552 libc version 2.2. These include all ISO C90 and C99 features, as well
2553 as features from the Single Unix Specification and some BSD and GNU
2554 extensions. Other library implementations may not support all these
2555 features; GCC does not support warning about features that go beyond a
2556 particular library's limitations. However, if @option{-pedantic} is used
2557 with @option{-Wformat}, warnings will be given about format features not
2558 in the selected standard version (but not for @code{strfmon} formats,
2559 since those are not in any version of the C standard). @xref{C Dialect
2560 Options,,Options Controlling C Dialect}.
2562 Since @option{-Wformat} also checks for null format arguments for
2563 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2565 @option{-Wformat} is included in @option{-Wall}. For more control over some
2566 aspects of format checking, the options @option{-Wformat-y2k},
2567 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2568 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2569 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2572 @opindex Wformat-y2k
2573 If @option{-Wformat} is specified, also warn about @code{strftime}
2574 formats which may yield only a two-digit year.
2576 @item -Wno-format-extra-args
2577 @opindex Wno-format-extra-args
2578 If @option{-Wformat} is specified, do not warn about excess arguments to a
2579 @code{printf} or @code{scanf} format function. The C standard specifies
2580 that such arguments are ignored.
2582 Where the unused arguments lie between used arguments that are
2583 specified with @samp{$} operand number specifications, normally
2584 warnings are still given, since the implementation could not know what
2585 type to pass to @code{va_arg} to skip the unused arguments. However,
2586 in the case of @code{scanf} formats, this option will suppress the
2587 warning if the unused arguments are all pointers, since the Single
2588 Unix Specification says that such unused arguments are allowed.
2590 @item -Wno-format-zero-length
2591 @opindex Wno-format-zero-length
2592 If @option{-Wformat} is specified, do not warn about zero-length formats.
2593 The C standard specifies that zero-length formats are allowed.
2595 @item -Wformat-nonliteral
2596 @opindex Wformat-nonliteral
2597 If @option{-Wformat} is specified, also warn if the format string is not a
2598 string literal and so cannot be checked, unless the format function
2599 takes its format arguments as a @code{va_list}.
2601 @item -Wformat-security
2602 @opindex Wformat-security
2603 If @option{-Wformat} is specified, also warn about uses of format
2604 functions that represent possible security problems. At present, this
2605 warns about calls to @code{printf} and @code{scanf} functions where the
2606 format string is not a string literal and there are no format arguments,
2607 as in @code{printf (foo);}. This may be a security hole if the format
2608 string came from untrusted input and contains @samp{%n}. (This is
2609 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2610 in future warnings may be added to @option{-Wformat-security} that are not
2611 included in @option{-Wformat-nonliteral}.)
2615 Enable @option{-Wformat} plus format checks not included in
2616 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2617 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2621 Warn about passing a null pointer for arguments marked as
2622 requiring a non-null value by the @code{nonnull} function attribute.
2624 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2625 can be disabled with the @option{-Wno-nonnull} option.
2627 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2629 Warn about uninitialized variables which are initialized with themselves.
2630 Note this option can only be used with the @option{-Wuninitialized} option,
2631 which in turn only works with @option{-O1} and above.
2633 For example, GCC will warn about @code{i} being uninitialized in the
2634 following snippet only when @option{-Winit-self} has been specified:
2645 @item -Wimplicit-int
2646 @opindex Wimplicit-int
2647 Warn when a declaration does not specify a type.
2648 This warning is enabled by @option{-Wall}.
2650 @item -Wimplicit-function-declaration
2651 @opindex Wimplicit-function-declaration
2652 @opindex Wno-implicit-function-declaration
2653 Give a warning whenever a function is used before being declared. In
2654 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2655 enabled by default and it is made into an error by
2656 @option{-pedantic-errors}. This warning is also enabled by
2661 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2662 This warning is enabled by @option{-Wall}.
2666 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2667 function with external linkage, returning int, taking either zero
2668 arguments, two, or three arguments of appropriate types.
2669 This warning is enabled by @option{-Wall}.
2671 @item -Wmissing-braces
2672 @opindex Wmissing-braces
2673 Warn if an aggregate or union initializer is not fully bracketed. In
2674 the following example, the initializer for @samp{a} is not fully
2675 bracketed, but that for @samp{b} is fully bracketed.
2678 int a[2][2] = @{ 0, 1, 2, 3 @};
2679 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2682 This warning is enabled by @option{-Wall}.
2684 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2685 @opindex Wmissing-include-dirs
2686 Warn if a user-supplied include directory does not exist.
2689 @opindex Wparentheses
2690 Warn if parentheses are omitted in certain contexts, such
2691 as when there is an assignment in a context where a truth value
2692 is expected, or when operators are nested whose precedence people
2693 often get confused about.
2695 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2696 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2697 interpretation from that of ordinary mathematical notation.
2699 Also warn about constructions where there may be confusion to which
2700 @code{if} statement an @code{else} branch belongs. Here is an example of
2715 In C/C++, every @code{else} branch belongs to the innermost possible
2716 @code{if} statement, which in this example is @code{if (b)}. This is
2717 often not what the programmer expected, as illustrated in the above
2718 example by indentation the programmer chose. When there is the
2719 potential for this confusion, GCC will issue a warning when this flag
2720 is specified. To eliminate the warning, add explicit braces around
2721 the innermost @code{if} statement so there is no way the @code{else}
2722 could belong to the enclosing @code{if}. The resulting code would
2739 This warning is enabled by @option{-Wall}.
2741 @item -Wsequence-point
2742 @opindex Wsequence-point
2743 Warn about code that may have undefined semantics because of violations
2744 of sequence point rules in the C and C++ standards.
2746 The C and C++ standards defines the order in which expressions in a C/C++
2747 program are evaluated in terms of @dfn{sequence points}, which represent
2748 a partial ordering between the execution of parts of the program: those
2749 executed before the sequence point, and those executed after it. These
2750 occur after the evaluation of a full expression (one which is not part
2751 of a larger expression), after the evaluation of the first operand of a
2752 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2753 function is called (but after the evaluation of its arguments and the
2754 expression denoting the called function), and in certain other places.
2755 Other than as expressed by the sequence point rules, the order of
2756 evaluation of subexpressions of an expression is not specified. All
2757 these rules describe only a partial order rather than a total order,
2758 since, for example, if two functions are called within one expression
2759 with no sequence point between them, the order in which the functions
2760 are called is not specified. However, the standards committee have
2761 ruled that function calls do not overlap.
2763 It is not specified when between sequence points modifications to the
2764 values of objects take effect. Programs whose behavior depends on this
2765 have undefined behavior; the C and C++ standards specify that ``Between
2766 the previous and next sequence point an object shall have its stored
2767 value modified at most once by the evaluation of an expression.
2768 Furthermore, the prior value shall be read only to determine the value
2769 to be stored.''. If a program breaks these rules, the results on any
2770 particular implementation are entirely unpredictable.
2772 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2773 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2774 diagnosed by this option, and it may give an occasional false positive
2775 result, but in general it has been found fairly effective at detecting
2776 this sort of problem in programs.
2778 The standard is worded confusingly, therefore there is some debate
2779 over the precise meaning of the sequence point rules in subtle cases.
2780 Links to discussions of the problem, including proposed formal
2781 definitions, may be found on the GCC readings page, at
2782 @w{@uref{http://gcc.gnu.org/readings.html}}.
2784 This warning is enabled by @option{-Wall} for C and C++.
2787 @opindex Wreturn-type
2788 Warn whenever a function is defined with a return-type that defaults to
2789 @code{int}. Also warn about any @code{return} statement with no
2790 return-value in a function whose return-type is not @code{void}.
2792 Also warn if the return type of a function has a type qualifier
2793 such as @code{const}. For ISO C such a type qualifier has no effect,
2794 since the value returned by a function is not an lvalue.
2795 For C++, the warning is only emitted for scalar types or @code{void}.
2796 ISO C prohibits qualified @code{void} return types on function
2797 definitions, so such return types always receive a warning
2798 even without this option.
2800 For C++, a function without return type always produces a diagnostic
2801 message, even when @option{-Wno-return-type} is specified. The only
2802 exceptions are @samp{main} and functions defined in system headers.
2804 This warning is enabled by @option{-Wall}.
2808 Warn whenever a @code{switch} statement has an index of enumerated type
2809 and lacks a @code{case} for one or more of the named codes of that
2810 enumeration. (The presence of a @code{default} label prevents this
2811 warning.) @code{case} labels outside the enumeration range also
2812 provoke warnings when this option is used.
2813 This warning is enabled by @option{-Wall}.
2815 @item -Wswitch-default
2816 @opindex Wswitch-switch
2817 Warn whenever a @code{switch} statement does not have a @code{default}
2821 @opindex Wswitch-enum
2822 Warn whenever a @code{switch} statement has an index of enumerated type
2823 and lacks a @code{case} for one or more of the named codes of that
2824 enumeration. @code{case} labels outside the enumeration range also
2825 provoke warnings when this option is used.
2829 Warn if any trigraphs are encountered that might change the meaning of
2830 the program (trigraphs within comments are not warned about).
2831 This warning is enabled by @option{-Wall}.
2833 @item -Wunused-function
2834 @opindex Wunused-function
2835 Warn whenever a static function is declared but not defined or a
2836 non-inline static function is unused.
2837 This warning is enabled by @option{-Wall}.
2839 @item -Wunused-label
2840 @opindex Wunused-label
2841 Warn whenever a label is declared but not used.
2842 This warning is enabled by @option{-Wall}.
2844 To suppress this warning use the @samp{unused} attribute
2845 (@pxref{Variable Attributes}).
2847 @item -Wunused-parameter
2848 @opindex Wunused-parameter
2849 Warn whenever a function parameter is unused aside from its declaration.
2851 To suppress this warning use the @samp{unused} attribute
2852 (@pxref{Variable Attributes}).
2854 @item -Wunused-variable
2855 @opindex Wunused-variable
2856 Warn whenever a local variable or non-constant static variable is unused
2857 aside from its declaration.
2858 This warning is enabled by @option{-Wall}.
2860 To suppress this warning use the @samp{unused} attribute
2861 (@pxref{Variable Attributes}).
2863 @item -Wunused-value
2864 @opindex Wunused-value
2865 Warn whenever a statement computes a result that is explicitly not
2866 used. To suppress this warning cast the unused expression to
2867 @samp{void}. This includes an expression-statement or the left-hand
2868 side of a comma expression that contains no side effects. For example,
2869 an expression such as @samp{x[i,j]} will cause a warning, while
2870 @samp{x[(void)i,j]} will not.
2872 This warning is enabled by @option{-Wall}.
2876 All the above @option{-Wunused} options combined.
2878 In order to get a warning about an unused function parameter, you must
2879 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2880 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2882 @item -Wuninitialized
2883 @opindex Wuninitialized
2884 Warn if an automatic variable is used without first being initialized or
2885 if a variable may be clobbered by a @code{setjmp} call.
2887 These warnings are possible only in optimizing compilation,
2888 because they require data flow information that is computed only
2889 when optimizing. If you do not specify @option{-O}, you will not get
2890 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2891 requiring @option{-O}.
2893 If you want to warn about code which uses the uninitialized value of the
2894 variable in its own initializer, use the @option{-Winit-self} option.
2896 These warnings occur for individual uninitialized or clobbered
2897 elements of structure, union or array variables as well as for
2898 variables which are uninitialized or clobbered as a whole. They do
2899 not occur for variables or elements declared @code{volatile}. Because
2900 these warnings depend on optimization, the exact variables or elements
2901 for which there are warnings will depend on the precise optimization
2902 options and version of GCC used.
2904 Note that there may be no warning about a variable that is used only
2905 to compute a value that itself is never used, because such
2906 computations may be deleted by data flow analysis before the warnings
2909 These warnings are made optional because GCC is not smart
2910 enough to see all the reasons why the code might be correct
2911 despite appearing to have an error. Here is one example of how
2932 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2933 always initialized, but GCC doesn't know this. Here is
2934 another common case:
2939 if (change_y) save_y = y, y = new_y;
2941 if (change_y) y = save_y;
2946 This has no bug because @code{save_y} is used only if it is set.
2948 @cindex @code{longjmp} warnings
2949 This option also warns when a non-volatile automatic variable might be
2950 changed by a call to @code{longjmp}. These warnings as well are possible
2951 only in optimizing compilation.
2953 The compiler sees only the calls to @code{setjmp}. It cannot know
2954 where @code{longjmp} will be called; in fact, a signal handler could
2955 call it at any point in the code. As a result, you may get a warning
2956 even when there is in fact no problem because @code{longjmp} cannot
2957 in fact be called at the place which would cause a problem.
2959 Some spurious warnings can be avoided if you declare all the functions
2960 you use that never return as @code{noreturn}. @xref{Function
2963 This warning is enabled by @option{-Wall}.
2965 @item -Wunknown-pragmas
2966 @opindex Wunknown-pragmas
2967 @cindex warning for unknown pragmas
2968 @cindex unknown pragmas, warning
2969 @cindex pragmas, warning of unknown
2970 Warn when a #pragma directive is encountered which is not understood by
2971 GCC@. If this command line option is used, warnings will even be issued
2972 for unknown pragmas in system header files. This is not the case if
2973 the warnings were only enabled by the @option{-Wall} command line option.
2976 @opindex Wno-pragmas
2978 Do not warn about misuses of pragmas, such as incorrect parameters,
2979 invalid syntax, or conflicts between pragmas. See also
2980 @samp{-Wunknown-pragmas}.
2982 @item -Wstrict-aliasing
2983 @opindex Wstrict-aliasing
2984 This option is only active when @option{-fstrict-aliasing} is active.
2985 It warns about code which might break the strict aliasing rules that the
2986 compiler is using for optimization. The warning does not catch all
2987 cases, but does attempt to catch the more common pitfalls. It is
2988 included in @option{-Wall}.
2990 @item -Wstrict-aliasing=2
2991 @opindex Wstrict-aliasing=2
2992 This option is only active when @option{-fstrict-aliasing} is active.
2993 It warns about code which might break the strict aliasing rules that the
2994 compiler is using for optimization. This warning catches more cases than
2995 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2996 cases that are safe.
2998 @item -Wstrict-overflow
2999 @item -Wstrict-overflow=@var{n}
3000 @opindex -Wstrict-overflow
3001 This option is only active when @option{-fstrict-overflow} is active.
3002 It warns about cases where the compiler optimizes based on the
3003 assumption that signed overflow does not occur. Note that it does not
3004 warn about all cases where the code might overflow: it only warns
3005 about cases where the compiler implements some optimization. Thus
3006 this warning depends on the optimization level.
3008 An optimization which assumes that signed overflow does not occur is
3009 perfectly safe if the values of the variables involved are such that
3010 overflow never does, in fact, occur. Therefore this warning can
3011 easily give a false positive: a warning about code which is not
3012 actually a problem. To help focus on important issues, several
3013 warning levels are defined.
3016 @item -Wstrict-overflow=1
3017 Warn about cases which are both questionable and easy to avoid. For
3018 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3019 compiler will simplify this to @code{1}. @option{-Wstrict-overflow}
3020 (with no level) is the same as @option{-Wstrict-overflow=1}. This
3021 level of @option{-Wstrict-overflow} is enabled by @option{-Wall};
3022 higher levels are not, and must be explicitly requested.
3024 @item -Wstrict-overflow=2
3025 Also warn about other cases where a comparison is simplified to a
3026 constant. For example: @code{abs (x) >= 0}. This can only be
3027 simplified when @option{-fstrict-overflow} is in effect, because
3028 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3031 @item -Wstrict-overflow=3
3032 Also warn about other cases where a comparison is simplified. For
3033 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3035 @item -Wstrict-overflow=4
3036 Also warn about other simplifications not covered by the above cases.
3037 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3039 @item -Wstrict-overflow=5
3040 Also warn about cases where the compiler reduces the magnitude of a
3041 constant involved in a comparison. For example: @code{x + 2 > y} will
3042 be simplified to @code{x + 1 >= y}. This is reported only at the
3043 highest warning level because this simplification applies to many
3044 comparisons, so this warning level will give a very large number of
3048 @item -Warray-bounds
3049 @opindex Wno-array-bounds
3050 @opindex Warray-bounds
3051 This option is only active when @option{-ftree-vrp} is active
3052 (default for -O2 and above). It warns about subscripts to arrays
3053 that are always out of bounds. This warning is enabled by @option{-Wall}.
3057 All of the above @samp{-W} options combined. This enables all the
3058 warnings about constructions that some users consider questionable, and
3059 that are easy to avoid (or modify to prevent the warning), even in
3060 conjunction with macros. This also enables some language-specific
3061 warnings described in @ref{C++ Dialect Options} and
3062 @ref{Objective-C and Objective-C++ Dialect Options}.
3065 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
3066 Some of them warn about constructions that users generally do not
3067 consider questionable, but which occasionally you might wish to check
3068 for; others warn about constructions that are necessary or hard to avoid
3069 in some cases, and there is no simple way to modify the code to suppress
3076 (This option used to be called @option{-W}. The older name is still
3077 supported, but the newer name is more descriptive.) Print extra warning
3078 messages for these events:
3082 A function can return either with or without a value. (Falling
3083 off the end of the function body is considered returning without
3084 a value.) For example, this function would evoke such a
3098 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
3101 Storage-class specifiers like @code{static} are not the first things
3102 in a declaration. According to the C Standard, this usage is
3103 obsolescent. This warning can be independently controlled by
3104 @option{-Wold-style-declaration}.
3107 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
3111 A comparison between signed and unsigned values could produce an
3112 incorrect result when the signed value is converted to unsigned.
3113 (But don't warn if @option{-Wno-sign-compare} is also specified.)
3116 An aggregate has an initializer which does not initialize all members.
3117 This warning can be independently controlled by
3118 @option{-Wmissing-field-initializers}.
3121 An initialized field without side effects is overridden when using
3122 designated initializers (@pxref{Designated Inits, , Designated
3123 Initializers}). This warning can be independently controlled by
3124 @option{-Woverride-init}.
3127 A function parameter is declared without a type specifier in K&R-style
3128 functions. This warning can be independently controlled by
3129 @option{-Wmissing-parameter-type}.
3132 An empty body occurs in an @samp{if} or @samp{else} statement. This
3133 warning can be independently controlled by @option{-Wempty-body}.
3136 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3137 @samp{>}, or @samp{>=}.
3140 A variable might be changed by @samp{longjmp} or @samp{vfork}.
3141 This warning can be independently controlled by @option{-Wclobbered}.
3143 @item @r{(C++ only)}
3144 An enumerator and a non-enumerator both appear in a conditional expression.
3146 @item @r{(C++ only)}
3147 A non-static reference or non-static @samp{const} member appears in a
3148 class without constructors.
3150 @item @r{(C++ only)}
3151 Ambiguous virtual bases.
3153 @item @r{(C++ only)}
3154 Subscripting an array which has been declared @samp{register}.
3156 @item @r{(C++ only)}
3157 Taking the address of a variable which has been declared @samp{register}.
3159 @item @r{(C++ only)}
3160 A base class is not initialized in a derived class' copy constructor.
3163 @item -Wno-div-by-zero
3164 @opindex Wno-div-by-zero
3165 @opindex Wdiv-by-zero
3166 Do not warn about compile-time integer division by zero. Floating point
3167 division by zero is not warned about, as it can be a legitimate way of
3168 obtaining infinities and NaNs.
3170 @item -Wsystem-headers
3171 @opindex Wsystem-headers
3172 @cindex warnings from system headers
3173 @cindex system headers, warnings from
3174 Print warning messages for constructs found in system header files.
3175 Warnings from system headers are normally suppressed, on the assumption
3176 that they usually do not indicate real problems and would only make the
3177 compiler output harder to read. Using this command line option tells
3178 GCC to emit warnings from system headers as if they occurred in user
3179 code. However, note that using @option{-Wall} in conjunction with this
3180 option will @emph{not} warn about unknown pragmas in system
3181 headers---for that, @option{-Wunknown-pragmas} must also be used.
3184 @opindex Wfloat-equal
3185 Warn if floating point values are used in equality comparisons.
3187 The idea behind this is that sometimes it is convenient (for the
3188 programmer) to consider floating-point values as approximations to
3189 infinitely precise real numbers. If you are doing this, then you need
3190 to compute (by analyzing the code, or in some other way) the maximum or
3191 likely maximum error that the computation introduces, and allow for it
3192 when performing comparisons (and when producing output, but that's a
3193 different problem). In particular, instead of testing for equality, you
3194 would check to see whether the two values have ranges that overlap; and
3195 this is done with the relational operators, so equality comparisons are
3198 @item -Wtraditional @r{(C only)}
3199 @opindex Wtraditional
3200 Warn about certain constructs that behave differently in traditional and
3201 ISO C@. Also warn about ISO C constructs that have no traditional C
3202 equivalent, and/or problematic constructs which should be avoided.
3206 Macro parameters that appear within string literals in the macro body.
3207 In traditional C macro replacement takes place within string literals,
3208 but does not in ISO C@.
3211 In traditional C, some preprocessor directives did not exist.
3212 Traditional preprocessors would only consider a line to be a directive
3213 if the @samp{#} appeared in column 1 on the line. Therefore
3214 @option{-Wtraditional} warns about directives that traditional C
3215 understands but would ignore because the @samp{#} does not appear as the
3216 first character on the line. It also suggests you hide directives like
3217 @samp{#pragma} not understood by traditional C by indenting them. Some
3218 traditional implementations would not recognize @samp{#elif}, so it
3219 suggests avoiding it altogether.
3222 A function-like macro that appears without arguments.
3225 The unary plus operator.
3228 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3229 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3230 constants.) Note, these suffixes appear in macros defined in the system
3231 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3232 Use of these macros in user code might normally lead to spurious
3233 warnings, however GCC's integrated preprocessor has enough context to
3234 avoid warning in these cases.
3237 A function declared external in one block and then used after the end of
3241 A @code{switch} statement has an operand of type @code{long}.
3244 A non-@code{static} function declaration follows a @code{static} one.
3245 This construct is not accepted by some traditional C compilers.
3248 The ISO type of an integer constant has a different width or
3249 signedness from its traditional type. This warning is only issued if
3250 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3251 typically represent bit patterns, are not warned about.
3254 Usage of ISO string concatenation is detected.
3257 Initialization of automatic aggregates.
3260 Identifier conflicts with labels. Traditional C lacks a separate
3261 namespace for labels.
3264 Initialization of unions. If the initializer is zero, the warning is
3265 omitted. This is done under the assumption that the zero initializer in
3266 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3267 initializer warnings and relies on default initialization to zero in the
3271 Conversions by prototypes between fixed/floating point values and vice
3272 versa. The absence of these prototypes when compiling with traditional
3273 C would cause serious problems. This is a subset of the possible
3274 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3277 Use of ISO C style function definitions. This warning intentionally is
3278 @emph{not} issued for prototype declarations or variadic functions
3279 because these ISO C features will appear in your code when using
3280 libiberty's traditional C compatibility macros, @code{PARAMS} and
3281 @code{VPARAMS}. This warning is also bypassed for nested functions
3282 because that feature is already a GCC extension and thus not relevant to
3283 traditional C compatibility.
3286 @item -Wtraditional-conversion @r{(C only)}
3287 @opindex Wtraditional-conversion
3288 Warn if a prototype causes a type conversion that is different from what
3289 would happen to the same argument in the absence of a prototype. This
3290 includes conversions of fixed point to floating and vice versa, and
3291 conversions changing the width or signedness of a fixed point argument
3292 except when the same as the default promotion.
3294 @item -Wdeclaration-after-statement @r{(C only)}
3295 @opindex Wdeclaration-after-statement
3296 Warn when a declaration is found after a statement in a block. This
3297 construct, known from C++, was introduced with ISO C99 and is by default
3298 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3299 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3303 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3305 @item -Wno-endif-labels
3306 @opindex Wno-endif-labels
3307 @opindex Wendif-labels
3308 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3312 Warn whenever a local variable shadows another local variable, parameter or
3313 global variable or whenever a built-in function is shadowed.
3315 @item -Wlarger-than-@var{len}
3316 @opindex Wlarger-than
3317 Warn whenever an object of larger than @var{len} bytes is defined.
3319 @item -Wunsafe-loop-optimizations
3320 @opindex Wunsafe-loop-optimizations
3321 Warn if the loop cannot be optimized because the compiler could not
3322 assume anything on the bounds of the loop indices. With
3323 @option{-funsafe-loop-optimizations} warn if the compiler made
3326 @item -Wpointer-arith
3327 @opindex Wpointer-arith
3328 Warn about anything that depends on the ``size of'' a function type or
3329 of @code{void}. GNU C assigns these types a size of 1, for
3330 convenience in calculations with @code{void *} pointers and pointers
3331 to functions. In C++, warn also when an arithmetic operation involves
3332 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3334 @item -Wbad-function-cast @r{(C only)}
3335 @opindex Wbad-function-cast
3336 Warn whenever a function call is cast to a non-matching type.
3337 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3340 Warn about ISO C constructs that are outside of the common subset of
3341 ISO C and ISO C++, e.g.@: request for implicit conversion from
3342 @code{void *} to a pointer to non-@code{void} type.
3344 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3345 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3346 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3347 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3351 Warn whenever a pointer is cast so as to remove a type qualifier from
3352 the target type. For example, warn if a @code{const char *} is cast
3353 to an ordinary @code{char *}.
3356 @opindex Wcast-align
3357 Warn whenever a pointer is cast such that the required alignment of the
3358 target is increased. For example, warn if a @code{char *} is cast to
3359 an @code{int *} on machines where integers can only be accessed at
3360 two- or four-byte boundaries.
3362 @item -Wwrite-strings
3363 @opindex Wwrite-strings
3364 When compiling C, give string constants the type @code{const
3365 char[@var{length}]} so that
3366 copying the address of one into a non-@code{const} @code{char *}
3367 pointer will get a warning; when compiling C++, warn about the
3368 deprecated conversion from string literals to @code{char *}. This
3369 warning, by default, is enabled for C++ programs.
3370 These warnings will help you find at
3371 compile time code that can try to write into a string constant, but
3372 only if you have been very careful about using @code{const} in
3373 declarations and prototypes. Otherwise, it will just be a nuisance;
3374 this is why we did not make @option{-Wall} request these warnings.
3378 Warn for variables that might be changed by @samp{longjmp} or
3379 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3382 @opindex Wconversion
3383 Warn for implicit conversions that may alter a value. This includes
3384 conversions between real and integer, like @code{abs (x)} when
3385 @code{x} is @code{double}; conversions between signed and unsigned,
3386 like @code{unsigned ui = -1}; and conversions to smaller types, like
3387 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3388 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3389 changed by the conversion like in @code{abs (2.0)}.
3392 @opindex Wempty-body
3393 An empty body occurs in an @samp{if} or @samp{else} statement.
3394 This warning is also enabled by @option{-Wextra}.
3396 @item -Wsign-compare
3397 @opindex Wsign-compare
3398 @cindex warning for comparison of signed and unsigned values
3399 @cindex comparison of signed and unsigned values, warning
3400 @cindex signed and unsigned values, comparison warning
3401 Warn when a comparison between signed and unsigned values could produce
3402 an incorrect result when the signed value is converted to unsigned.
3403 This warning is also enabled by @option{-Wextra}; to get the other warnings
3404 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3408 @opindex Wno-address
3409 Warn about suspicious uses of memory addresses. These include using
3410 the address of a function in a conditional expression, such as
3411 @code{void func(void); if (func)}, and comparisons against the memory
3412 address of a string literal, such as @code{if (x == "abc")}. Such
3413 uses typically indicate a programmer error: the address of a function
3414 always evaluates to true, so their use in a conditional usually
3415 indicate that the programmer forgot the parentheses in a function
3416 call; and comparisons against string literals result in unspecified
3417 behavior and are not portable in C, so they usually indicate that the
3418 programmer intended to use @code{strcmp}. This warning is enabled by
3421 @item -Waggregate-return
3422 @opindex Waggregate-return
3423 Warn if any functions that return structures or unions are defined or
3424 called. (In languages where you can return an array, this also elicits
3427 @item -Wno-attributes
3428 @opindex Wno-attributes
3429 @opindex Wattributes
3430 Do not warn if an unexpected @code{__attribute__} is used, such as
3431 unrecognized attributes, function attributes applied to variables,
3432 etc. This will not stop errors for incorrect use of supported
3435 @item -Wstrict-prototypes @r{(C only)}
3436 @opindex Wstrict-prototypes
3437 Warn if a function is declared or defined without specifying the
3438 argument types. (An old-style function definition is permitted without
3439 a warning if preceded by a declaration which specifies the argument
3442 @item -Wold-style-declaration @r{(C only)}
3443 @opindex Wold-style-declaration
3444 Warn for obsolescent usages, according to the C Standard, in a
3445 declaration. For example, warn if storage-class specifiers like
3446 @code{static} are not the first things in a declaration. This warning
3447 is also enabled by @option{-Wextra}.
3449 @item -Wold-style-definition @r{(C only)}
3450 @opindex Wold-style-definition
3451 Warn if an old-style function definition is used. A warning is given
3452 even if there is a previous prototype.
3454 @item -Wmissing-parameter-type @r{(C only)}
3455 @opindex Wmissing-parameter-type
3456 A function parameter is declared without a type specifier in K&R-style
3463 This warning is also enabled by @option{-Wextra}.
3465 @item -Wmissing-prototypes @r{(C only)}
3466 @opindex Wmissing-prototypes
3467 Warn if a global function is defined without a previous prototype
3468 declaration. This warning is issued even if the definition itself
3469 provides a prototype. The aim is to detect global functions that fail
3470 to be declared in header files.
3472 @item -Wmissing-declarations @r{(C and C++ only)}
3473 @opindex Wmissing-declarations
3474 Warn if a global function is defined without a previous declaration.
3475 Do so even if the definition itself provides a prototype.
3476 Use this option to detect global functions that are not declared in
3477 header files. In C++, no warnings are issued for function templates,
3478 or for inline functions, or for functions in anonymous namespaces.
3480 @item -Wmissing-field-initializers
3481 @opindex Wmissing-field-initializers
3484 Warn if a structure's initializer has some fields missing. For
3485 example, the following code would cause such a warning, because
3486 @code{x.h} is implicitly zero:
3489 struct s @{ int f, g, h; @};
3490 struct s x = @{ 3, 4 @};
3493 This option does not warn about designated initializers, so the following
3494 modification would not trigger a warning:
3497 struct s @{ int f, g, h; @};
3498 struct s x = @{ .f = 3, .g = 4 @};
3501 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3502 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3504 @item -Wmissing-noreturn
3505 @opindex Wmissing-noreturn
3506 Warn about functions which might be candidates for attribute @code{noreturn}.
3507 Note these are only possible candidates, not absolute ones. Care should
3508 be taken to manually verify functions actually do not ever return before
3509 adding the @code{noreturn} attribute, otherwise subtle code generation
3510 bugs could be introduced. You will not get a warning for @code{main} in
3511 hosted C environments.
3513 @item -Wmissing-format-attribute
3514 @opindex Wmissing-format-attribute
3516 Warn about function pointers which might be candidates for @code{format}
3517 attributes. Note these are only possible candidates, not absolute ones.
3518 GCC will guess that function pointers with @code{format} attributes that
3519 are used in assignment, initialization, parameter passing or return
3520 statements should have a corresponding @code{format} attribute in the
3521 resulting type. I.e.@: the left-hand side of the assignment or
3522 initialization, the type of the parameter variable, or the return type
3523 of the containing function respectively should also have a @code{format}
3524 attribute to avoid the warning.
3526 GCC will also warn about function definitions which might be
3527 candidates for @code{format} attributes. Again, these are only
3528 possible candidates. GCC will guess that @code{format} attributes
3529 might be appropriate for any function that calls a function like
3530 @code{vprintf} or @code{vscanf}, but this might not always be the
3531 case, and some functions for which @code{format} attributes are
3532 appropriate may not be detected.
3534 @item -Wno-multichar
3535 @opindex Wno-multichar
3537 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3538 Usually they indicate a typo in the user's code, as they have
3539 implementation-defined values, and should not be used in portable code.
3541 @item -Wnormalized=<none|id|nfc|nfkc>
3542 @opindex Wnormalized
3545 @cindex character set, input normalization
3546 In ISO C and ISO C++, two identifiers are different if they are
3547 different sequences of characters. However, sometimes when characters
3548 outside the basic ASCII character set are used, you can have two
3549 different character sequences that look the same. To avoid confusion,
3550 the ISO 10646 standard sets out some @dfn{normalization rules} which
3551 when applied ensure that two sequences that look the same are turned into
3552 the same sequence. GCC can warn you if you are using identifiers which
3553 have not been normalized; this option controls that warning.
3555 There are four levels of warning that GCC supports. The default is
3556 @option{-Wnormalized=nfc}, which warns about any identifier which is
3557 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3558 recommended form for most uses.
3560 Unfortunately, there are some characters which ISO C and ISO C++ allow
3561 in identifiers that when turned into NFC aren't allowable as
3562 identifiers. That is, there's no way to use these symbols in portable
3563 ISO C or C++ and have all your identifiers in NFC.
3564 @option{-Wnormalized=id} suppresses the warning for these characters.
3565 It is hoped that future versions of the standards involved will correct
3566 this, which is why this option is not the default.
3568 You can switch the warning off for all characters by writing
3569 @option{-Wnormalized=none}. You would only want to do this if you
3570 were using some other normalization scheme (like ``D''), because
3571 otherwise you can easily create bugs that are literally impossible to see.
3573 Some characters in ISO 10646 have distinct meanings but look identical
3574 in some fonts or display methodologies, especially once formatting has
3575 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3576 LETTER N'', will display just like a regular @code{n} which has been
3577 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3578 normalization scheme to convert all these into a standard form as
3579 well, and GCC will warn if your code is not in NFKC if you use
3580 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3581 about every identifier that contains the letter O because it might be
3582 confused with the digit 0, and so is not the default, but may be
3583 useful as a local coding convention if the programming environment is
3584 unable to be fixed to display these characters distinctly.
3586 @item -Wno-deprecated-declarations
3587 @opindex Wno-deprecated-declarations
3588 Do not warn about uses of functions (@pxref{Function Attributes}),
3589 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3590 Attributes}) marked as deprecated by using the @code{deprecated}
3594 @opindex Wno-overflow
3595 Do not warn about compile-time overflow in constant expressions.
3597 @item -Woverride-init
3598 @opindex Woverride-init
3601 Warn if an initialized field without side effects is overridden when
3602 using designated initializers (@pxref{Designated Inits, , Designated
3605 This warning is included in @option{-Wextra}. To get other
3606 @option{-Wextra} warnings without this one, use @samp{-Wextra
3607 -Wno-override-init}.
3611 Warn if a structure is given the packed attribute, but the packed
3612 attribute has no effect on the layout or size of the structure.
3613 Such structures may be mis-aligned for little benefit. For
3614 instance, in this code, the variable @code{f.x} in @code{struct bar}
3615 will be misaligned even though @code{struct bar} does not itself
3616 have the packed attribute:
3623 @} __attribute__((packed));
3633 Warn if padding is included in a structure, either to align an element
3634 of the structure or to align the whole structure. Sometimes when this
3635 happens it is possible to rearrange the fields of the structure to
3636 reduce the padding and so make the structure smaller.
3638 @item -Wredundant-decls
3639 @opindex Wredundant-decls
3640 Warn if anything is declared more than once in the same scope, even in
3641 cases where multiple declaration is valid and changes nothing.
3643 @item -Wnested-externs @r{(C only)}
3644 @opindex Wnested-externs
3645 Warn if an @code{extern} declaration is encountered within a function.
3647 @item -Wunreachable-code
3648 @opindex Wunreachable-code
3649 Warn if the compiler detects that code will never be executed.
3651 This option is intended to warn when the compiler detects that at
3652 least a whole line of source code will never be executed, because
3653 some condition is never satisfied or because it is after a
3654 procedure that never returns.
3656 It is possible for this option to produce a warning even though there
3657 are circumstances under which part of the affected line can be executed,
3658 so care should be taken when removing apparently-unreachable code.
3660 For instance, when a function is inlined, a warning may mean that the
3661 line is unreachable in only one inlined copy of the function.
3663 This option is not made part of @option{-Wall} because in a debugging
3664 version of a program there is often substantial code which checks
3665 correct functioning of the program and is, hopefully, unreachable
3666 because the program does work. Another common use of unreachable
3667 code is to provide behavior which is selectable at compile-time.
3671 Warn if a function can not be inlined and it was declared as inline.
3672 Even with this option, the compiler will not warn about failures to
3673 inline functions declared in system headers.
3675 The compiler uses a variety of heuristics to determine whether or not
3676 to inline a function. For example, the compiler takes into account
3677 the size of the function being inlined and the amount of inlining
3678 that has already been done in the current function. Therefore,
3679 seemingly insignificant changes in the source program can cause the
3680 warnings produced by @option{-Winline} to appear or disappear.
3682 @item -Wno-invalid-offsetof @r{(C++ only)}
3683 @opindex Wno-invalid-offsetof
3684 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3685 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3686 to a non-POD type is undefined. In existing C++ implementations,
3687 however, @samp{offsetof} typically gives meaningful results even when
3688 applied to certain kinds of non-POD types. (Such as a simple
3689 @samp{struct} that fails to be a POD type only by virtue of having a
3690 constructor.) This flag is for users who are aware that they are
3691 writing nonportable code and who have deliberately chosen to ignore the
3694 The restrictions on @samp{offsetof} may be relaxed in a future version
3695 of the C++ standard.
3697 @item -Wno-int-to-pointer-cast @r{(C only)}
3698 @opindex Wno-int-to-pointer-cast
3699 Suppress warnings from casts to pointer type of an integer of a
3702 @item -Wno-pointer-to-int-cast @r{(C only)}
3703 @opindex Wno-pointer-to-int-cast
3704 Suppress warnings from casts from a pointer to an integer type of a
3708 @opindex Winvalid-pch
3709 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3710 the search path but can't be used.
3714 @opindex Wno-long-long
3715 Warn if @samp{long long} type is used. This is default. To inhibit
3716 the warning messages, use @option{-Wno-long-long}. Flags
3717 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3718 only when @option{-pedantic} flag is used.
3720 @item -Wvariadic-macros
3721 @opindex Wvariadic-macros
3722 @opindex Wno-variadic-macros
3723 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3724 alternate syntax when in pedantic ISO C99 mode. This is default.
3725 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3727 @item -Wvolatile-register-var
3728 @opindex Wvolatile-register-var
3729 @opindex Wno-volatile-register-var
3730 Warn if a register variable is declared volatile. The volatile
3731 modifier does not inhibit all optimizations that may eliminate reads
3732 and/or writes to register variables.
3734 @item -Wdisabled-optimization
3735 @opindex Wdisabled-optimization
3736 Warn if a requested optimization pass is disabled. This warning does
3737 not generally indicate that there is anything wrong with your code; it
3738 merely indicates that GCC's optimizers were unable to handle the code
3739 effectively. Often, the problem is that your code is too big or too
3740 complex; GCC will refuse to optimize programs when the optimization
3741 itself is likely to take inordinate amounts of time.
3743 @item -Wpointer-sign
3744 @opindex Wpointer-sign
3745 @opindex Wno-pointer-sign
3746 Warn for pointer argument passing or assignment with different signedness.
3747 This option is only supported for C and Objective-C@. It is implied by
3748 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3749 @option{-Wno-pointer-sign}.
3753 Make all warnings into errors.
3757 Make the specified warning into an errors. The specifier for a
3758 warning is appended, for example @option{-Werror=switch} turns the
3759 warnings controlled by @option{-Wswitch} into errors. This switch
3760 takes a negative form, to be used to negate @option{-Werror} for
3761 specific warnings, for example @option{-Wno-error=switch} makes
3762 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3763 is in effect. You can use the @option{-fdiagnostics-show-option}
3764 option to have each controllable warning amended with the option which
3765 controls it, to determine what to use with this option.
3767 Note that specifying @option{-Werror=}@var{foo} automatically implies
3768 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3771 @item -Wstack-protector
3772 @opindex Wstack-protector
3773 This option is only active when @option{-fstack-protector} is active. It
3774 warns about functions that will not be protected against stack smashing.
3776 @item -Woverlength-strings
3777 @opindex Woverlength-strings
3778 Warn about string constants which are longer than the ``minimum
3779 maximum'' length specified in the C standard. Modern compilers
3780 generally allow string constants which are much longer than the
3781 standard's minimum limit, but very portable programs should avoid
3782 using longer strings.
3784 The limit applies @emph{after} string constant concatenation, and does
3785 not count the trailing NUL@. In C89, the limit was 509 characters; in
3786 C99, it was raised to 4095. C++98 does not specify a normative
3787 minimum maximum, so we do not diagnose overlength strings in C++@.
3789 This option is implied by @option{-pedantic}, and can be disabled with
3790 @option{-Wno-overlength-strings}.
3793 @node Debugging Options
3794 @section Options for Debugging Your Program or GCC
3795 @cindex options, debugging
3796 @cindex debugging information options
3798 GCC has various special options that are used for debugging
3799 either your program or GCC:
3804 Produce debugging information in the operating system's native format
3805 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3808 On most systems that use stabs format, @option{-g} enables use of extra
3809 debugging information that only GDB can use; this extra information
3810 makes debugging work better in GDB but will probably make other debuggers
3812 refuse to read the program. If you want to control for certain whether
3813 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3814 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3816 GCC allows you to use @option{-g} with
3817 @option{-O}. The shortcuts taken by optimized code may occasionally
3818 produce surprising results: some variables you declared may not exist
3819 at all; flow of control may briefly move where you did not expect it;
3820 some statements may not be executed because they compute constant
3821 results or their values were already at hand; some statements may
3822 execute in different places because they were moved out of loops.
3824 Nevertheless it proves possible to debug optimized output. This makes
3825 it reasonable to use the optimizer for programs that might have bugs.
3827 The following options are useful when GCC is generated with the
3828 capability for more than one debugging format.
3832 Produce debugging information for use by GDB@. This means to use the
3833 most expressive format available (DWARF 2, stabs, or the native format
3834 if neither of those are supported), including GDB extensions if at all
3839 Produce debugging information in stabs format (if that is supported),
3840 without GDB extensions. This is the format used by DBX on most BSD
3841 systems. On MIPS, Alpha and System V Release 4 systems this option
3842 produces stabs debugging output which is not understood by DBX or SDB@.
3843 On System V Release 4 systems this option requires the GNU assembler.
3845 @item -feliminate-unused-debug-symbols
3846 @opindex feliminate-unused-debug-symbols
3847 Produce debugging information in stabs format (if that is supported),
3848 for only symbols that are actually used.
3850 @item -femit-class-debug-always
3851 Instead of emitting debugging information for a C++ class in only one
3852 object file, emit it in all object files using the class. This option
3853 should be used only with debuggers that are unable to handle the way GCC
3854 normally emits debugging information for classes because using this
3855 option will increase the size of debugging information by as much as a
3860 Produce debugging information in stabs format (if that is supported),
3861 using GNU extensions understood only by the GNU debugger (GDB)@. The
3862 use of these extensions is likely to make other debuggers crash or
3863 refuse to read the program.
3867 Produce debugging information in COFF format (if that is supported).
3868 This is the format used by SDB on most System V systems prior to
3873 Produce debugging information in XCOFF format (if that is supported).
3874 This is the format used by the DBX debugger on IBM RS/6000 systems.
3878 Produce debugging information in XCOFF format (if that is supported),
3879 using GNU extensions understood only by the GNU debugger (GDB)@. The
3880 use of these extensions is likely to make other debuggers crash or
3881 refuse to read the program, and may cause assemblers other than the GNU
3882 assembler (GAS) to fail with an error.
3886 Produce debugging information in DWARF version 2 format (if that is
3887 supported). This is the format used by DBX on IRIX 6. With this
3888 option, GCC uses features of DWARF version 3 when they are useful;
3889 version 3 is upward compatible with version 2, but may still cause
3890 problems for older debuggers.
3894 Produce debugging information in VMS debug format (if that is
3895 supported). This is the format used by DEBUG on VMS systems.
3898 @itemx -ggdb@var{level}
3899 @itemx -gstabs@var{level}
3900 @itemx -gcoff@var{level}
3901 @itemx -gxcoff@var{level}
3902 @itemx -gvms@var{level}
3903 Request debugging information and also use @var{level} to specify how
3904 much information. The default level is 2.
3906 Level 0 produces no debug information at all. Thus, @option{-g0} negates
3909 Level 1 produces minimal information, enough for making backtraces in
3910 parts of the program that you don't plan to debug. This includes
3911 descriptions of functions and external variables, but no information
3912 about local variables and no line numbers.
3914 Level 3 includes extra information, such as all the macro definitions
3915 present in the program. Some debuggers support macro expansion when
3916 you use @option{-g3}.
3918 @option{-gdwarf-2} does not accept a concatenated debug level, because
3919 GCC used to support an option @option{-gdwarf} that meant to generate
3920 debug information in version 1 of the DWARF format (which is very
3921 different from version 2), and it would have been too confusing. That
3922 debug format is long obsolete, but the option cannot be changed now.
3923 Instead use an additional @option{-g@var{level}} option to change the
3924 debug level for DWARF2.
3926 @item -feliminate-dwarf2-dups
3927 @opindex feliminate-dwarf2-dups
3928 Compress DWARF2 debugging information by eliminating duplicated
3929 information about each symbol. This option only makes sense when
3930 generating DWARF2 debugging information with @option{-gdwarf-2}.
3932 @cindex @command{prof}
3935 Generate extra code to write profile information suitable for the
3936 analysis program @command{prof}. You must use this option when compiling
3937 the source files you want data about, and you must also use it when
3940 @cindex @command{gprof}
3943 Generate extra code to write profile information suitable for the
3944 analysis program @command{gprof}. You must use this option when compiling
3945 the source files you want data about, and you must also use it when
3950 Makes the compiler print out each function name as it is compiled, and
3951 print some statistics about each pass when it finishes.
3954 @opindex ftime-report
3955 Makes the compiler print some statistics about the time consumed by each
3956 pass when it finishes.
3959 @opindex fmem-report
3960 Makes the compiler print some statistics about permanent memory
3961 allocation when it finishes.
3963 @item -fpre-ipa-mem-report
3964 @opindex fpre-ipa-mem-report
3965 @item -fpost-ipa-mem-report
3966 @opindex fpost-ipa-mem-report
3967 Makes the compiler print some statistics about permanent memory
3968 allocation before or after interprocedural optimization.
3970 @item -fprofile-arcs
3971 @opindex fprofile-arcs
3972 Add code so that program flow @dfn{arcs} are instrumented. During
3973 execution the program records how many times each branch and call is
3974 executed and how many times it is taken or returns. When the compiled
3975 program exits it saves this data to a file called
3976 @file{@var{auxname}.gcda} for each source file. The data may be used for
3977 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3978 test coverage analysis (@option{-ftest-coverage}). Each object file's
3979 @var{auxname} is generated from the name of the output file, if
3980 explicitly specified and it is not the final executable, otherwise it is
3981 the basename of the source file. In both cases any suffix is removed
3982 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3983 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3984 @xref{Cross-profiling}.
3986 @cindex @command{gcov}
3990 This option is used to compile and link code instrumented for coverage
3991 analysis. The option is a synonym for @option{-fprofile-arcs}
3992 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3993 linking). See the documentation for those options for more details.
3998 Compile the source files with @option{-fprofile-arcs} plus optimization
3999 and code generation options. For test coverage analysis, use the
4000 additional @option{-ftest-coverage} option. You do not need to profile
4001 every source file in a program.
4004 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4005 (the latter implies the former).
4008 Run the program on a representative workload to generate the arc profile
4009 information. This may be repeated any number of times. You can run
4010 concurrent instances of your program, and provided that the file system
4011 supports locking, the data files will be correctly updated. Also
4012 @code{fork} calls are detected and correctly handled (double counting
4016 For profile-directed optimizations, compile the source files again with
4017 the same optimization and code generation options plus
4018 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4019 Control Optimization}).
4022 For test coverage analysis, use @command{gcov} to produce human readable
4023 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4024 @command{gcov} documentation for further information.
4028 With @option{-fprofile-arcs}, for each function of your program GCC
4029 creates a program flow graph, then finds a spanning tree for the graph.
4030 Only arcs that are not on the spanning tree have to be instrumented: the
4031 compiler adds code to count the number of times that these arcs are
4032 executed. When an arc is the only exit or only entrance to a block, the
4033 instrumentation code can be added to the block; otherwise, a new basic
4034 block must be created to hold the instrumentation code.
4037 @item -ftest-coverage
4038 @opindex ftest-coverage
4039 Produce a notes file that the @command{gcov} code-coverage utility
4040 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4041 show program coverage. Each source file's note file is called
4042 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4043 above for a description of @var{auxname} and instructions on how to
4044 generate test coverage data. Coverage data will match the source files
4045 more closely, if you do not optimize.
4047 @item -d@var{letters}
4048 @item -fdump-rtl-@var{pass}
4050 Says to make debugging dumps during compilation at times specified by
4051 @var{letters}. This is used for debugging the RTL-based passes of the
4052 compiler. The file names for most of the dumps are made by appending a
4053 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4054 from the name of the output file, if explicitly specified and it is not
4055 an executable, otherwise it is the basename of the source file.
4057 Most debug dumps can be enabled either passing a letter to the @option{-d}
4058 option, or with a long @option{-fdump-rtl} switch; here are the possible
4059 letters for use in @var{letters} and @var{pass}, and their meanings:
4064 Annotate the assembler output with miscellaneous debugging information.
4067 @itemx -fdump-rtl-bbro
4069 @opindex fdump-rtl-bbro
4070 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4073 @itemx -fdump-rtl-combine
4075 @opindex fdump-rtl-combine
4076 Dump after the RTL instruction combination pass, to the file
4077 @file{@var{file}.129r.combine}.
4080 @itemx -fdump-rtl-ce1
4081 @itemx -fdump-rtl-ce2
4083 @opindex fdump-rtl-ce1
4084 @opindex fdump-rtl-ce2
4085 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
4086 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
4087 and @option{-fdump-rtl-ce2} enable dumping after the second if
4088 conversion, to the file @file{@var{file}.130r.ce2}.
4091 @itemx -fdump-rtl-btl
4092 @itemx -fdump-rtl-dbr
4094 @opindex fdump-rtl-btl
4095 @opindex fdump-rtl-dbr
4096 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
4097 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
4098 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
4099 scheduling, to @file{@var{file}.36.dbr}.
4103 Dump all macro definitions, at the end of preprocessing, in addition to
4107 @itemx -fdump-rtl-ce3
4109 @opindex fdump-rtl-ce3
4110 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4113 @itemx -fdump-rtl-cfg
4114 @itemx -fdump-rtl-life
4116 @opindex fdump-rtl-cfg
4117 @opindex fdump-rtl-life
4118 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
4119 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
4120 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4121 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4124 @itemx -fdump-rtl-greg
4126 @opindex fdump-rtl-greg
4127 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4130 @itemx -fdump-rtl-gcse
4131 @itemx -fdump-rtl-bypass
4133 @opindex fdump-rtl-gcse
4134 @opindex fdump-rtl-bypass
4135 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4136 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
4137 enable dumping after jump bypassing and control flow optimizations, to
4138 @file{@var{file}.115r.bypass}.
4141 @itemx -fdump-rtl-eh
4143 @opindex fdump-rtl-eh
4144 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4147 @itemx -fdump-rtl-sibling
4149 @opindex fdump-rtl-sibling
4150 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4153 @itemx -fdump-rtl-jump
4155 @opindex fdump-rtl-jump
4156 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4159 @itemx -fdump-rtl-stack
4161 @opindex fdump-rtl-stack
4162 Dump after conversion from GCC's "flat register file" registers to the
4163 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4166 @itemx -fdump-rtl-lreg
4168 @opindex fdump-rtl-lreg
4169 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4172 @itemx -fdump-rtl-loop2
4174 @opindex fdump-rtl-loop2
4175 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4176 loop optimization pass, to @file{@var{file}.119r.loop2},
4177 @file{@var{file}.120r.loop2_init},
4178 @file{@var{file}.121r.loop2_invariant}, and
4179 @file{@var{file}.125r.loop2_done}.
4182 @itemx -fdump-rtl-sms
4184 @opindex fdump-rtl-sms
4185 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4188 @itemx -fdump-rtl-mach
4190 @opindex fdump-rtl-mach
4191 Dump after performing the machine dependent reorganization pass, to
4192 @file{@var{file}.155r.mach}.
4195 @itemx -fdump-rtl-rnreg
4197 @opindex fdump-rtl-rnreg
4198 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4201 @itemx -fdump-rtl-regmove
4203 @opindex fdump-rtl-regmove
4204 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4207 @itemx -fdump-rtl-postreload
4209 @opindex fdump-rtl-postreload
4210 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4213 @itemx -fdump-rtl-expand
4215 @opindex fdump-rtl-expand
4216 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4219 @itemx -fdump-rtl-sched2
4221 @opindex fdump-rtl-sched2
4222 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4225 @itemx -fdump-rtl-cse
4227 @opindex fdump-rtl-cse
4228 Dump after CSE (including the jump optimization that sometimes follows
4229 CSE), to @file{@var{file}.113r.cse}.
4232 @itemx -fdump-rtl-sched1
4234 @opindex fdump-rtl-sched1
4235 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4238 @itemx -fdump-rtl-cse2
4240 @opindex fdump-rtl-cse2
4241 Dump after the second CSE pass (including the jump optimization that
4242 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4245 @itemx -fdump-rtl-tracer
4247 @opindex fdump-rtl-tracer
4248 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4251 @itemx -fdump-rtl-vpt
4252 @itemx -fdump-rtl-vartrack
4254 @opindex fdump-rtl-vpt
4255 @opindex fdump-rtl-vartrack
4256 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4257 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4258 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4259 to @file{@var{file}.154r.vartrack}.
4262 @itemx -fdump-rtl-flow2
4264 @opindex fdump-rtl-flow2
4265 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4268 @itemx -fdump-rtl-peephole2
4270 @opindex fdump-rtl-peephole2
4271 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4274 @itemx -fdump-rtl-web
4276 @opindex fdump-rtl-web
4277 Dump after live range splitting, to @file{@var{file}.126r.web}.
4280 @itemx -fdump-rtl-all
4282 @opindex fdump-rtl-all
4283 Produce all the dumps listed above.
4287 Produce a core dump whenever an error occurs.
4291 Print statistics on memory usage, at the end of the run, to
4296 Annotate the assembler output with a comment indicating which
4297 pattern and alternative was used. The length of each instruction is
4302 Dump the RTL in the assembler output as a comment before each instruction.
4303 Also turns on @option{-dp} annotation.
4307 For each of the other indicated dump files (either with @option{-d} or
4308 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4309 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4313 Just generate RTL for a function instead of compiling it. Usually used
4314 with @samp{r} (@option{-fdump-rtl-expand}).
4318 Dump debugging information during parsing, to standard error.
4322 @opindex fdump-noaddr
4323 When doing debugging dumps (see @option{-d} option above), suppress
4324 address output. This makes it more feasible to use diff on debugging
4325 dumps for compiler invocations with different compiler binaries and/or
4326 different text / bss / data / heap / stack / dso start locations.
4328 @item -fdump-unnumbered
4329 @opindex fdump-unnumbered
4330 When doing debugging dumps (see @option{-d} option above), suppress instruction
4331 numbers, line number note and address output. This makes it more feasible to
4332 use diff on debugging dumps for compiler invocations with different
4333 options, in particular with and without @option{-g}.
4335 @item -fdump-translation-unit @r{(C++ only)}
4336 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4337 @opindex fdump-translation-unit
4338 Dump a representation of the tree structure for the entire translation
4339 unit to a file. The file name is made by appending @file{.tu} to the
4340 source file name. If the @samp{-@var{options}} form is used, @var{options}
4341 controls the details of the dump as described for the
4342 @option{-fdump-tree} options.
4344 @item -fdump-class-hierarchy @r{(C++ only)}
4345 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4346 @opindex fdump-class-hierarchy
4347 Dump a representation of each class's hierarchy and virtual function
4348 table layout to a file. The file name is made by appending @file{.class}
4349 to the source file name. If the @samp{-@var{options}} form is used,
4350 @var{options} controls the details of the dump as described for the
4351 @option{-fdump-tree} options.
4353 @item -fdump-ipa-@var{switch}
4355 Control the dumping at various stages of inter-procedural analysis
4356 language tree to a file. The file name is generated by appending a switch
4357 specific suffix to the source file name. The following dumps are possible:
4361 Enables all inter-procedural analysis dumps; currently the only produced
4362 dump is the @samp{cgraph} dump.
4365 Dumps information about call-graph optimization, unused function removal,
4366 and inlining decisions.
4369 @item -fdump-tree-@var{switch}
4370 @itemx -fdump-tree-@var{switch}-@var{options}
4372 Control the dumping at various stages of processing the intermediate
4373 language tree to a file. The file name is generated by appending a switch
4374 specific suffix to the source file name. If the @samp{-@var{options}}
4375 form is used, @var{options} is a list of @samp{-} separated options that
4376 control the details of the dump. Not all options are applicable to all
4377 dumps, those which are not meaningful will be ignored. The following
4378 options are available
4382 Print the address of each node. Usually this is not meaningful as it
4383 changes according to the environment and source file. Its primary use
4384 is for tying up a dump file with a debug environment.
4386 Inhibit dumping of members of a scope or body of a function merely
4387 because that scope has been reached. Only dump such items when they
4388 are directly reachable by some other path. When dumping pretty-printed
4389 trees, this option inhibits dumping the bodies of control structures.
4391 Print a raw representation of the tree. By default, trees are
4392 pretty-printed into a C-like representation.
4394 Enable more detailed dumps (not honored by every dump option).
4396 Enable dumping various statistics about the pass (not honored by every dump
4399 Enable showing basic block boundaries (disabled in raw dumps).
4401 Enable showing virtual operands for every statement.
4403 Enable showing line numbers for statements.
4405 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4407 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4410 The following tree dumps are possible:
4414 Dump before any tree based optimization, to @file{@var{file}.original}.
4417 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4420 Dump after function inlining, to @file{@var{file}.inlined}.
4423 @opindex fdump-tree-gimple
4424 Dump each function before and after the gimplification pass to a file. The
4425 file name is made by appending @file{.gimple} to the source file name.
4428 @opindex fdump-tree-cfg
4429 Dump the control flow graph of each function to a file. The file name is
4430 made by appending @file{.cfg} to the source file name.
4433 @opindex fdump-tree-vcg
4434 Dump the control flow graph of each function to a file in VCG format. The
4435 file name is made by appending @file{.vcg} to the source file name. Note
4436 that if the file contains more than one function, the generated file cannot
4437 be used directly by VCG@. You will need to cut and paste each function's
4438 graph into its own separate file first.
4441 @opindex fdump-tree-ch
4442 Dump each function after copying loop headers. The file name is made by
4443 appending @file{.ch} to the source file name.
4446 @opindex fdump-tree-ssa
4447 Dump SSA related information to a file. The file name is made by appending
4448 @file{.ssa} to the source file name.
4451 @opindex fdump-tree-salias
4452 Dump structure aliasing variable information to a file. This file name
4453 is made by appending @file{.salias} to the source file name.
4456 @opindex fdump-tree-alias
4457 Dump aliasing information for each function. The file name is made by
4458 appending @file{.alias} to the source file name.
4461 @opindex fdump-tree-ccp
4462 Dump each function after CCP@. The file name is made by appending
4463 @file{.ccp} to the source file name.
4466 @opindex fdump-tree-storeccp
4467 Dump each function after STORE-CCP. The file name is made by appending
4468 @file{.storeccp} to the source file name.
4471 @opindex fdump-tree-pre
4472 Dump trees after partial redundancy elimination. The file name is made
4473 by appending @file{.pre} to the source file name.
4476 @opindex fdump-tree-fre
4477 Dump trees after full redundancy elimination. The file name is made
4478 by appending @file{.fre} to the source file name.
4481 @opindex fdump-tree-copyprop
4482 Dump trees after copy propagation. The file name is made
4483 by appending @file{.copyprop} to the source file name.
4485 @item store_copyprop
4486 @opindex fdump-tree-store_copyprop
4487 Dump trees after store copy-propagation. The file name is made
4488 by appending @file{.store_copyprop} to the source file name.
4491 @opindex fdump-tree-dce
4492 Dump each function after dead code elimination. The file name is made by
4493 appending @file{.dce} to the source file name.
4496 @opindex fdump-tree-mudflap
4497 Dump each function after adding mudflap instrumentation. The file name is
4498 made by appending @file{.mudflap} to the source file name.
4501 @opindex fdump-tree-sra
4502 Dump each function after performing scalar replacement of aggregates. The
4503 file name is made by appending @file{.sra} to the source file name.
4506 @opindex fdump-tree-sink
4507 Dump each function after performing code sinking. The file name is made
4508 by appending @file{.sink} to the source file name.
4511 @opindex fdump-tree-dom
4512 Dump each function after applying dominator tree optimizations. The file
4513 name is made by appending @file{.dom} to the source file name.
4516 @opindex fdump-tree-dse
4517 Dump each function after applying dead store elimination. The file
4518 name is made by appending @file{.dse} to the source file name.
4521 @opindex fdump-tree-phiopt
4522 Dump each function after optimizing PHI nodes into straightline code. The file
4523 name is made by appending @file{.phiopt} to the source file name.
4526 @opindex fdump-tree-forwprop
4527 Dump each function after forward propagating single use variables. The file
4528 name is made by appending @file{.forwprop} to the source file name.
4531 @opindex fdump-tree-copyrename
4532 Dump each function after applying the copy rename optimization. The file
4533 name is made by appending @file{.copyrename} to the source file name.
4536 @opindex fdump-tree-nrv
4537 Dump each function after applying the named return value optimization on
4538 generic trees. The file name is made by appending @file{.nrv} to the source
4542 @opindex fdump-tree-vect
4543 Dump each function after applying vectorization of loops. The file name is
4544 made by appending @file{.vect} to the source file name.
4547 @opindex fdump-tree-vrp
4548 Dump each function after Value Range Propagation (VRP). The file name
4549 is made by appending @file{.vrp} to the source file name.
4552 @opindex fdump-tree-all
4553 Enable all the available tree dumps with the flags provided in this option.
4556 @item -ftree-vectorizer-verbose=@var{n}
4557 @opindex ftree-vectorizer-verbose
4558 This option controls the amount of debugging output the vectorizer prints.
4559 This information is written to standard error, unless
4560 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4561 in which case it is output to the usual dump listing file, @file{.vect}.
4562 For @var{n}=0 no diagnostic information is reported.
4563 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4564 and the total number of loops that got vectorized.
4565 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4566 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4567 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4568 level that @option{-fdump-tree-vect-stats} uses.
4569 Higher verbosity levels mean either more information dumped for each
4570 reported loop, or same amount of information reported for more loops:
4571 If @var{n}=3, alignment related information is added to the reports.
4572 If @var{n}=4, data-references related information (e.g. memory dependences,
4573 memory access-patterns) is added to the reports.
4574 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4575 that did not pass the first analysis phase (i.e. may not be countable, or
4576 may have complicated control-flow).
4577 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4578 For @var{n}=7, all the information the vectorizer generates during its
4579 analysis and transformation is reported. This is the same verbosity level
4580 that @option{-fdump-tree-vect-details} uses.
4582 @item -frandom-seed=@var{string}
4583 @opindex frandom-string
4584 This option provides a seed that GCC uses when it would otherwise use
4585 random numbers. It is used to generate certain symbol names
4586 that have to be different in every compiled file. It is also used to
4587 place unique stamps in coverage data files and the object files that
4588 produce them. You can use the @option{-frandom-seed} option to produce
4589 reproducibly identical object files.
4591 The @var{string} should be different for every file you compile.
4593 @item -fsched-verbose=@var{n}
4594 @opindex fsched-verbose
4595 On targets that use instruction scheduling, this option controls the
4596 amount of debugging output the scheduler prints. This information is
4597 written to standard error, unless @option{-dS} or @option{-dR} is
4598 specified, in which case it is output to the usual dump
4599 listing file, @file{.sched} or @file{.sched2} respectively. However
4600 for @var{n} greater than nine, the output is always printed to standard
4603 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4604 same information as @option{-dRS}. For @var{n} greater than one, it
4605 also output basic block probabilities, detailed ready list information
4606 and unit/insn info. For @var{n} greater than two, it includes RTL
4607 at abort point, control-flow and regions info. And for @var{n} over
4608 four, @option{-fsched-verbose} also includes dependence info.
4612 Store the usual ``temporary'' intermediate files permanently; place them
4613 in the current directory and name them based on the source file. Thus,
4614 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4615 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4616 preprocessed @file{foo.i} output file even though the compiler now
4617 normally uses an integrated preprocessor.
4619 When used in combination with the @option{-x} command line option,
4620 @option{-save-temps} is sensible enough to avoid over writing an
4621 input source file with the same extension as an intermediate file.
4622 The corresponding intermediate file may be obtained by renaming the
4623 source file before using @option{-save-temps}.
4627 Report the CPU time taken by each subprocess in the compilation
4628 sequence. For C source files, this is the compiler proper and assembler
4629 (plus the linker if linking is done). The output looks like this:
4636 The first number on each line is the ``user time'', that is time spent
4637 executing the program itself. The second number is ``system time'',
4638 time spent executing operating system routines on behalf of the program.
4639 Both numbers are in seconds.
4641 @item -fvar-tracking
4642 @opindex fvar-tracking
4643 Run variable tracking pass. It computes where variables are stored at each
4644 position in code. Better debugging information is then generated
4645 (if the debugging information format supports this information).
4647 It is enabled by default when compiling with optimization (@option{-Os},
4648 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4649 the debug info format supports it.
4651 @item -print-file-name=@var{library}
4652 @opindex print-file-name
4653 Print the full absolute name of the library file @var{library} that
4654 would be used when linking---and don't do anything else. With this
4655 option, GCC does not compile or link anything; it just prints the
4658 @item -print-multi-directory
4659 @opindex print-multi-directory
4660 Print the directory name corresponding to the multilib selected by any
4661 other switches present in the command line. This directory is supposed
4662 to exist in @env{GCC_EXEC_PREFIX}.
4664 @item -print-multi-lib
4665 @opindex print-multi-lib
4666 Print the mapping from multilib directory names to compiler switches
4667 that enable them. The directory name is separated from the switches by
4668 @samp{;}, and each switch starts with an @samp{@@} instead of the
4669 @samp{-}, without spaces between multiple switches. This is supposed to
4670 ease shell-processing.
4672 @item -print-prog-name=@var{program}
4673 @opindex print-prog-name
4674 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4676 @item -print-libgcc-file-name
4677 @opindex print-libgcc-file-name
4678 Same as @option{-print-file-name=libgcc.a}.
4680 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4681 but you do want to link with @file{libgcc.a}. You can do
4684 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4687 @item -print-search-dirs
4688 @opindex print-search-dirs
4689 Print the name of the configured installation directory and a list of
4690 program and library directories @command{gcc} will search---and don't do anything else.
4692 This is useful when @command{gcc} prints the error message
4693 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4694 To resolve this you either need to put @file{cpp0} and the other compiler
4695 components where @command{gcc} expects to find them, or you can set the environment
4696 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4697 Don't forget the trailing @samp{/}.
4698 @xref{Environment Variables}.
4701 @opindex dumpmachine
4702 Print the compiler's target machine (for example,
4703 @samp{i686-pc-linux-gnu})---and don't do anything else.
4706 @opindex dumpversion
4707 Print the compiler version (for example, @samp{3.0})---and don't do
4712 Print the compiler's built-in specs---and don't do anything else. (This
4713 is used when GCC itself is being built.) @xref{Spec Files}.
4715 @item -feliminate-unused-debug-types
4716 @opindex feliminate-unused-debug-types
4717 Normally, when producing DWARF2 output, GCC will emit debugging
4718 information for all types declared in a compilation
4719 unit, regardless of whether or not they are actually used
4720 in that compilation unit. Sometimes this is useful, such as
4721 if, in the debugger, you want to cast a value to a type that is
4722 not actually used in your program (but is declared). More often,
4723 however, this results in a significant amount of wasted space.
4724 With this option, GCC will avoid producing debug symbol output
4725 for types that are nowhere used in the source file being compiled.
4728 @node Optimize Options
4729 @section Options That Control Optimization
4730 @cindex optimize options
4731 @cindex options, optimization
4733 These options control various sorts of optimizations.
4735 Without any optimization option, the compiler's goal is to reduce the
4736 cost of compilation and to make debugging produce the expected
4737 results. Statements are independent: if you stop the program with a
4738 breakpoint between statements, you can then assign a new value to any
4739 variable or change the program counter to any other statement in the
4740 function and get exactly the results you would expect from the source
4743 Turning on optimization flags makes the compiler attempt to improve
4744 the performance and/or code size at the expense of compilation time
4745 and possibly the ability to debug the program.
4747 The compiler performs optimization based on the knowledge it has of
4748 the program. Optimization levels @option{-O} and above, in
4749 particular, enable @emph{unit-at-a-time} mode, which allows the
4750 compiler to consider information gained from later functions in
4751 the file when compiling a function. Compiling multiple files at
4752 once to a single output file in @emph{unit-at-a-time} mode allows
4753 the compiler to use information gained from all of the files when
4754 compiling each of them.
4756 Not all optimizations are controlled directly by a flag. Only
4757 optimizations that have a flag are listed.
4764 Optimize. Optimizing compilation takes somewhat more time, and a lot
4765 more memory for a large function.
4767 With @option{-O}, the compiler tries to reduce code size and execution
4768 time, without performing any optimizations that take a great deal of
4771 @option{-O} turns on the following optimization flags:
4772 @gccoptlist{-fdefer-pop @gol
4773 -fdelayed-branch @gol
4774 -fguess-branch-probability @gol
4775 -fcprop-registers @gol
4776 -fif-conversion @gol
4777 -fif-conversion2 @gol
4778 -fsplit-wide-types @gol
4781 -ftree-dominator-opts @gol
4785 -ftree-copyrename @gol
4788 -funit-at-a-time @gol
4791 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4792 where doing so does not interfere with debugging.
4796 Optimize even more. GCC performs nearly all supported optimizations
4797 that do not involve a space-speed tradeoff. The compiler does not
4798 perform loop unrolling or function inlining when you specify @option{-O2}.
4799 As compared to @option{-O}, this option increases both compilation time
4800 and the performance of the generated code.
4802 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4803 also turns on the following optimization flags:
4804 @gccoptlist{-fthread-jumps @gol
4806 -foptimize-sibling-calls @gol
4807 -fcse-follow-jumps -fcse-skip-blocks @gol
4808 -fgcse -fgcse-lm @gol
4809 -fexpensive-optimizations @gol
4810 -frerun-cse-after-loop @gol
4813 -fschedule-insns -fschedule-insns2 @gol
4814 -fsched-interblock -fsched-spec @gol
4816 -fstrict-aliasing -fstrict-overflow @gol
4817 -fdelete-null-pointer-checks @gol
4818 -freorder-blocks -freorder-functions @gol
4819 -falign-functions -falign-jumps @gol
4820 -falign-loops -falign-labels @gol
4824 Please note the warning under @option{-fgcse} about
4825 invoking @option{-O2} on programs that use computed gotos.
4829 Optimize yet more. @option{-O3} turns on all optimizations specified by
4830 @option{-O2} and also turns on the @option{-finline-functions},
4831 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4835 Reduce compilation time and make debugging produce the expected
4836 results. This is the default.
4840 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4841 do not typically increase code size. It also performs further
4842 optimizations designed to reduce code size.
4844 @option{-Os} disables the following optimization flags:
4845 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4846 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4847 -fprefetch-loop-arrays -ftree-vect-loop-version}
4849 If you use multiple @option{-O} options, with or without level numbers,
4850 the last such option is the one that is effective.
4853 Options of the form @option{-f@var{flag}} specify machine-independent
4854 flags. Most flags have both positive and negative forms; the negative
4855 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4856 below, only one of the forms is listed---the one you typically will
4857 use. You can figure out the other form by either removing @samp{no-}
4860 The following options control specific optimizations. They are either
4861 activated by @option{-O} options or are related to ones that are. You
4862 can use the following flags in the rare cases when ``fine-tuning'' of
4863 optimizations to be performed is desired.
4866 @item -fno-default-inline
4867 @opindex fno-default-inline
4868 Do not make member functions inline by default merely because they are
4869 defined inside the class scope (C++ only). Otherwise, when you specify
4870 @w{@option{-O}}, member functions defined inside class scope are compiled
4871 inline by default; i.e., you don't need to add @samp{inline} in front of
4872 the member function name.
4874 @item -fno-defer-pop
4875 @opindex fno-defer-pop
4876 Always pop the arguments to each function call as soon as that function
4877 returns. For machines which must pop arguments after a function call,
4878 the compiler normally lets arguments accumulate on the stack for several
4879 function calls and pops them all at once.
4881 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4884 @opindex fforce-addr
4885 Force memory address constants to be copied into registers before
4886 doing arithmetic on them.
4888 @item -fforward-propagate
4889 @opindex fforward-propagate
4890 Perform a forward propagation pass on RTL. The pass tries to combine two
4891 instructions and checks if the result can be simplified. If loop unrolling
4892 is active, two passes are performed and the second is scheduled after
4895 This option is enabled by default at optimization levels @option{-O2},
4896 @option{-O3}, @option{-Os}.
4898 @item -fomit-frame-pointer
4899 @opindex fomit-frame-pointer
4900 Don't keep the frame pointer in a register for functions that
4901 don't need one. This avoids the instructions to save, set up and
4902 restore frame pointers; it also makes an extra register available
4903 in many functions. @strong{It also makes debugging impossible on
4906 On some machines, such as the VAX, this flag has no effect, because
4907 the standard calling sequence automatically handles the frame pointer
4908 and nothing is saved by pretending it doesn't exist. The
4909 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4910 whether a target machine supports this flag. @xref{Registers,,Register
4911 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4913 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4915 @item -foptimize-sibling-calls
4916 @opindex foptimize-sibling-calls
4917 Optimize sibling and tail recursive calls.
4919 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4923 Don't pay attention to the @code{inline} keyword. Normally this option
4924 is used to keep the compiler from expanding any functions inline.
4925 Note that if you are not optimizing, no functions can be expanded inline.
4927 @item -finline-functions
4928 @opindex finline-functions
4929 Integrate all simple functions into their callers. The compiler
4930 heuristically decides which functions are simple enough to be worth
4931 integrating in this way.
4933 If all calls to a given function are integrated, and the function is
4934 declared @code{static}, then the function is normally not output as
4935 assembler code in its own right.
4937 Enabled at level @option{-O3}.
4939 @item -finline-functions-called-once
4940 @opindex finline-functions-called-once
4941 Consider all @code{static} functions called once for inlining into their
4942 caller even if they are not marked @code{inline}. If a call to a given
4943 function is integrated, then the function is not output as assembler code
4946 Enabled if @option{-funit-at-a-time} is enabled.
4948 @item -fearly-inlining
4949 @opindex fearly-inlining
4950 Inline functions marked by @code{always_inline} and functions whose body seems
4951 smaller than the function call overhead early before doing
4952 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4953 makes profiling significantly cheaper and usually inlining faster on programs
4954 having large chains of nested wrapper functions.
4958 @item -finline-limit=@var{n}
4959 @opindex finline-limit
4960 By default, GCC limits the size of functions that can be inlined. This flag
4961 allows the control of this limit for functions that are explicitly marked as
4962 inline (i.e., marked with the inline keyword or defined within the class
4963 definition in c++). @var{n} is the size of functions that can be inlined in
4964 number of pseudo instructions (not counting parameter handling). The default
4965 value of @var{n} is 600.
4966 Increasing this value can result in more inlined code at
4967 the cost of compilation time and memory consumption. Decreasing usually makes
4968 the compilation faster and less code will be inlined (which presumably
4969 means slower programs). This option is particularly useful for programs that
4970 use inlining heavily such as those based on recursive templates with C++.
4972 Inlining is actually controlled by a number of parameters, which may be
4973 specified individually by using @option{--param @var{name}=@var{value}}.
4974 The @option{-finline-limit=@var{n}} option sets some of these parameters
4978 @item max-inline-insns-single
4979 is set to @var{n}/2.
4980 @item max-inline-insns-auto
4981 is set to @var{n}/2.
4982 @item min-inline-insns
4983 is set to 130 or @var{n}/4, whichever is smaller.
4984 @item max-inline-insns-rtl
4988 See below for a documentation of the individual
4989 parameters controlling inlining.
4991 @emph{Note:} pseudo instruction represents, in this particular context, an
4992 abstract measurement of function's size. In no way does it represent a count
4993 of assembly instructions and as such its exact meaning might change from one
4994 release to an another.
4996 @item -fkeep-inline-functions
4997 @opindex fkeep-inline-functions
4998 In C, emit @code{static} functions that are declared @code{inline}
4999 into the object file, even if the function has been inlined into all
5000 of its callers. This switch does not affect functions using the
5001 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5002 inline functions into the object file.
5004 @item -fkeep-static-consts
5005 @opindex fkeep-static-consts
5006 Emit variables declared @code{static const} when optimization isn't turned
5007 on, even if the variables aren't referenced.
5009 GCC enables this option by default. If you want to force the compiler to
5010 check if the variable was referenced, regardless of whether or not
5011 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5013 @item -fmerge-constants
5014 Attempt to merge identical constants (string constants and floating point
5015 constants) across compilation units.
5017 This option is the default for optimized compilation if the assembler and
5018 linker support it. Use @option{-fno-merge-constants} to inhibit this
5021 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5023 @item -fmerge-all-constants
5024 Attempt to merge identical constants and identical variables.
5026 This option implies @option{-fmerge-constants}. In addition to
5027 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5028 arrays or initialized constant variables with integral or floating point
5029 types. Languages like C or C++ require each non-automatic variable to
5030 have distinct location, so using this option will result in non-conforming
5033 @item -fmodulo-sched
5034 @opindex fmodulo-sched
5035 Perform swing modulo scheduling immediately before the first scheduling
5036 pass. This pass looks at innermost loops and reorders their
5037 instructions by overlapping different iterations.
5039 @item -fno-branch-count-reg
5040 @opindex fno-branch-count-reg
5041 Do not use ``decrement and branch'' instructions on a count register,
5042 but instead generate a sequence of instructions that decrement a
5043 register, compare it against zero, then branch based upon the result.
5044 This option is only meaningful on architectures that support such
5045 instructions, which include x86, PowerPC, IA-64 and S/390.
5047 The default is @option{-fbranch-count-reg}.
5049 @item -fno-function-cse
5050 @opindex fno-function-cse
5051 Do not put function addresses in registers; make each instruction that
5052 calls a constant function contain the function's address explicitly.
5054 This option results in less efficient code, but some strange hacks
5055 that alter the assembler output may be confused by the optimizations
5056 performed when this option is not used.
5058 The default is @option{-ffunction-cse}
5060 @item -fno-zero-initialized-in-bss
5061 @opindex fno-zero-initialized-in-bss
5062 If the target supports a BSS section, GCC by default puts variables that
5063 are initialized to zero into BSS@. This can save space in the resulting
5066 This option turns off this behavior because some programs explicitly
5067 rely on variables going to the data section. E.g., so that the
5068 resulting executable can find the beginning of that section and/or make
5069 assumptions based on that.
5071 The default is @option{-fzero-initialized-in-bss}.
5073 @item -fbounds-check
5074 @opindex fbounds-check
5075 For front-ends that support it, generate additional code to check that
5076 indices used to access arrays are within the declared range. This is
5077 currently only supported by the Java and Fortran front-ends, where
5078 this option defaults to true and false respectively.
5080 @item -fmudflap -fmudflapth -fmudflapir
5084 @cindex bounds checking
5086 For front-ends that support it (C and C++), instrument all risky
5087 pointer/array dereferencing operations, some standard library
5088 string/heap functions, and some other associated constructs with
5089 range/validity tests. Modules so instrumented should be immune to
5090 buffer overflows, invalid heap use, and some other classes of C/C++
5091 programming errors. The instrumentation relies on a separate runtime
5092 library (@file{libmudflap}), which will be linked into a program if
5093 @option{-fmudflap} is given at link time. Run-time behavior of the
5094 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5095 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5098 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5099 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5100 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5101 instrumentation should ignore pointer reads. This produces less
5102 instrumentation (and therefore faster execution) and still provides
5103 some protection against outright memory corrupting writes, but allows
5104 erroneously read data to propagate within a program.
5106 @item -fthread-jumps
5107 @opindex fthread-jumps
5108 Perform optimizations where we check to see if a jump branches to a
5109 location where another comparison subsumed by the first is found. If
5110 so, the first branch is redirected to either the destination of the
5111 second branch or a point immediately following it, depending on whether
5112 the condition is known to be true or false.
5114 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5116 @item -fsplit-wide-types
5117 @opindex -fsplit-wide-types
5118 When using a type that occupies multiple registers, such as @code{long
5119 long} on a 32-bit system, split the registers apart and allocate them
5120 independently. This normally generates better code for those types,
5121 but may make debugging more difficult.
5123 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5126 @item -fcse-follow-jumps
5127 @opindex fcse-follow-jumps
5128 In common subexpression elimination, scan through jump instructions
5129 when the target of the jump is not reached by any other path. For
5130 example, when CSE encounters an @code{if} statement with an
5131 @code{else} clause, CSE will follow the jump when the condition
5134 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5136 @item -fcse-skip-blocks
5137 @opindex fcse-skip-blocks
5138 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5139 follow jumps which conditionally skip over blocks. When CSE
5140 encounters a simple @code{if} statement with no else clause,
5141 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5142 body of the @code{if}.
5144 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5146 @item -frerun-cse-after-loop
5147 @opindex frerun-cse-after-loop
5148 Re-run common subexpression elimination after loop optimizations has been
5151 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5155 Perform a global common subexpression elimination pass.
5156 This pass also performs global constant and copy propagation.
5158 @emph{Note:} When compiling a program using computed gotos, a GCC
5159 extension, you may get better runtime performance if you disable
5160 the global common subexpression elimination pass by adding
5161 @option{-fno-gcse} to the command line.
5163 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5167 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5168 attempt to move loads which are only killed by stores into themselves. This
5169 allows a loop containing a load/store sequence to be changed to a load outside
5170 the loop, and a copy/store within the loop.
5172 Enabled by default when gcse is enabled.
5176 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5177 global common subexpression elimination. This pass will attempt to move
5178 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5179 loops containing a load/store sequence can be changed to a load before
5180 the loop and a store after the loop.
5182 Not enabled at any optimization level.
5186 When @option{-fgcse-las} is enabled, the global common subexpression
5187 elimination pass eliminates redundant loads that come after stores to the
5188 same memory location (both partial and full redundancies).
5190 Not enabled at any optimization level.
5192 @item -fgcse-after-reload
5193 @opindex fgcse-after-reload
5194 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5195 pass is performed after reload. The purpose of this pass is to cleanup
5198 @item -funsafe-loop-optimizations
5199 @opindex funsafe-loop-optimizations
5200 If given, the loop optimizer will assume that loop indices do not
5201 overflow, and that the loops with nontrivial exit condition are not
5202 infinite. This enables a wider range of loop optimizations even if
5203 the loop optimizer itself cannot prove that these assumptions are valid.
5204 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5205 if it finds this kind of loop.
5207 @item -fcrossjumping
5208 @opindex crossjumping
5209 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5210 resulting code may or may not perform better than without cross-jumping.
5212 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5214 @item -fif-conversion
5215 @opindex if-conversion
5216 Attempt to transform conditional jumps into branch-less equivalents. This
5217 include use of conditional moves, min, max, set flags and abs instructions, and
5218 some tricks doable by standard arithmetics. The use of conditional execution
5219 on chips where it is available is controlled by @code{if-conversion2}.
5221 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5223 @item -fif-conversion2
5224 @opindex if-conversion2
5225 Use conditional execution (where available) to transform conditional jumps into
5226 branch-less equivalents.
5228 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5230 @item -fdelete-null-pointer-checks
5231 @opindex fdelete-null-pointer-checks
5232 Use global dataflow analysis to identify and eliminate useless checks
5233 for null pointers. The compiler assumes that dereferencing a null
5234 pointer would have halted the program. If a pointer is checked after
5235 it has already been dereferenced, it cannot be null.
5237 In some environments, this assumption is not true, and programs can
5238 safely dereference null pointers. Use
5239 @option{-fno-delete-null-pointer-checks} to disable this optimization
5240 for programs which depend on that behavior.
5242 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5244 @item -fexpensive-optimizations
5245 @opindex fexpensive-optimizations
5246 Perform a number of minor optimizations that are relatively expensive.
5248 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5250 @item -foptimize-register-move
5252 @opindex foptimize-register-move
5254 Attempt to reassign register numbers in move instructions and as
5255 operands of other simple instructions in order to maximize the amount of
5256 register tying. This is especially helpful on machines with two-operand
5259 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5262 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5264 @item -fdelayed-branch
5265 @opindex fdelayed-branch
5266 If supported for the target machine, attempt to reorder instructions
5267 to exploit instruction slots available after delayed branch
5270 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5272 @item -fschedule-insns
5273 @opindex fschedule-insns
5274 If supported for the target machine, attempt to reorder instructions to
5275 eliminate execution stalls due to required data being unavailable. This
5276 helps machines that have slow floating point or memory load instructions
5277 by allowing other instructions to be issued until the result of the load
5278 or floating point instruction is required.
5280 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5282 @item -fschedule-insns2
5283 @opindex fschedule-insns2
5284 Similar to @option{-fschedule-insns}, but requests an additional pass of
5285 instruction scheduling after register allocation has been done. This is
5286 especially useful on machines with a relatively small number of
5287 registers and where memory load instructions take more than one cycle.
5289 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5291 @item -fno-sched-interblock
5292 @opindex fno-sched-interblock
5293 Don't schedule instructions across basic blocks. This is normally
5294 enabled by default when scheduling before register allocation, i.e.@:
5295 with @option{-fschedule-insns} or at @option{-O2} or higher.
5297 @item -fno-sched-spec
5298 @opindex fno-sched-spec
5299 Don't allow speculative motion of non-load instructions. This is normally
5300 enabled by default when scheduling before register allocation, i.e.@:
5301 with @option{-fschedule-insns} or at @option{-O2} or higher.
5303 @item -fsched-spec-load
5304 @opindex fsched-spec-load
5305 Allow speculative motion of some load instructions. This only makes
5306 sense when scheduling before register allocation, i.e.@: with
5307 @option{-fschedule-insns} or at @option{-O2} or higher.
5309 @item -fsched-spec-load-dangerous
5310 @opindex fsched-spec-load-dangerous
5311 Allow speculative motion of more load instructions. This only makes
5312 sense when scheduling before register allocation, i.e.@: with
5313 @option{-fschedule-insns} or at @option{-O2} or higher.
5315 @item -fsched-stalled-insns=@var{n}
5316 @opindex fsched-stalled-insns
5317 Define how many insns (if any) can be moved prematurely from the queue
5318 of stalled insns into the ready list, during the second scheduling pass.
5320 @item -fsched-stalled-insns-dep=@var{n}
5321 @opindex fsched-stalled-insns-dep
5322 Define how many insn groups (cycles) will be examined for a dependency
5323 on a stalled insn that is candidate for premature removal from the queue
5324 of stalled insns. Has an effect only during the second scheduling pass,
5325 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5327 @item -fsched2-use-superblocks
5328 @opindex fsched2-use-superblocks
5329 When scheduling after register allocation, do use superblock scheduling
5330 algorithm. Superblock scheduling allows motion across basic block boundaries
5331 resulting on faster schedules. This option is experimental, as not all machine
5332 descriptions used by GCC model the CPU closely enough to avoid unreliable
5333 results from the algorithm.
5335 This only makes sense when scheduling after register allocation, i.e.@: with
5336 @option{-fschedule-insns2} or at @option{-O2} or higher.
5338 @item -fsched2-use-traces
5339 @opindex fsched2-use-traces
5340 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5341 allocation and additionally perform code duplication in order to increase the
5342 size of superblocks using tracer pass. See @option{-ftracer} for details on
5345 This mode should produce faster but significantly longer programs. Also
5346 without @option{-fbranch-probabilities} the traces constructed may not
5347 match the reality and hurt the performance. This only makes
5348 sense when scheduling after register allocation, i.e.@: with
5349 @option{-fschedule-insns2} or at @option{-O2} or higher.
5353 Eliminates redundant extension instructions and move the non redundant
5354 ones to optimal placement using LCM.
5356 @item -freschedule-modulo-scheduled-loops
5357 @opindex fscheduling-in-modulo-scheduled-loops
5358 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5359 we may want to prevent the later scheduling passes from changing its schedule, we use this
5360 option to control that.
5362 @item -fcaller-saves
5363 @opindex fcaller-saves
5364 Enable values to be allocated in registers that will be clobbered by
5365 function calls, by emitting extra instructions to save and restore the
5366 registers around such calls. Such allocation is done only when it
5367 seems to result in better code than would otherwise be produced.
5369 This option is always enabled by default on certain machines, usually
5370 those which have no call-preserved registers to use instead.
5372 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5375 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5376 enabled by default at @option{-O2} and @option{-O3}.
5379 Perform Full Redundancy Elimination (FRE) on trees. The difference
5380 between FRE and PRE is that FRE only considers expressions
5381 that are computed on all paths leading to the redundant computation.
5382 This analysis faster than PRE, though it exposes fewer redundancies.
5383 This flag is enabled by default at @option{-O} and higher.
5385 @item -ftree-copy-prop
5386 Perform copy propagation on trees. This pass eliminates unnecessary
5387 copy operations. This flag is enabled by default at @option{-O} and
5390 @item -ftree-store-copy-prop
5391 Perform copy propagation of memory loads and stores. This pass
5392 eliminates unnecessary copy operations in memory references
5393 (structures, global variables, arrays, etc). This flag is enabled by
5394 default at @option{-O2} and higher.
5397 Perform structural alias analysis on trees. This flag
5398 is enabled by default at @option{-O} and higher.
5401 Perform interprocedural pointer analysis.
5404 Perform forward store motion on trees. This flag is
5405 enabled by default at @option{-O} and higher.
5408 Perform sparse conditional constant propagation (CCP) on trees. This
5409 pass only operates on local scalar variables and is enabled by default
5410 at @option{-O} and higher.
5412 @item -ftree-store-ccp
5413 Perform sparse conditional constant propagation (CCP) on trees. This
5414 pass operates on both local scalar variables and memory stores and
5415 loads (global variables, structures, arrays, etc). This flag is
5416 enabled by default at @option{-O2} and higher.
5419 Perform dead code elimination (DCE) on trees. This flag is enabled by
5420 default at @option{-O} and higher.
5422 @item -ftree-dominator-opts
5423 Perform a variety of simple scalar cleanups (constant/copy
5424 propagation, redundancy elimination, range propagation and expression
5425 simplification) based on a dominator tree traversal. This also
5426 performs jump threading (to reduce jumps to jumps). This flag is
5427 enabled by default at @option{-O} and higher.
5430 Perform loop header copying on trees. This is beneficial since it increases
5431 effectiveness of code motion optimizations. It also saves one jump. This flag
5432 is enabled by default at @option{-O} and higher. It is not enabled
5433 for @option{-Os}, since it usually increases code size.
5435 @item -ftree-loop-optimize
5436 Perform loop optimizations on trees. This flag is enabled by default
5437 at @option{-O} and higher.
5439 @item -ftree-loop-linear
5440 Perform linear loop transformations on tree. This flag can improve cache
5441 performance and allow further loop optimizations to take place.
5443 @item -ftree-loop-im
5444 Perform loop invariant motion on trees. This pass moves only invariants that
5445 would be hard to handle at RTL level (function calls, operations that expand to
5446 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5447 operands of conditions that are invariant out of the loop, so that we can use
5448 just trivial invariantness analysis in loop unswitching. The pass also includes
5451 @item -ftree-loop-ivcanon
5452 Create a canonical counter for number of iterations in the loop for that
5453 determining number of iterations requires complicated analysis. Later
5454 optimizations then may determine the number easily. Useful especially
5455 in connection with unrolling.
5458 Perform induction variable optimizations (strength reduction, induction
5459 variable merging and induction variable elimination) on trees.
5462 Perform scalar replacement of aggregates. This pass replaces structure
5463 references with scalars to prevent committing structures to memory too
5464 early. This flag is enabled by default at @option{-O} and higher.
5466 @item -ftree-copyrename
5467 Perform copy renaming on trees. This pass attempts to rename compiler
5468 temporaries to other variables at copy locations, usually resulting in
5469 variable names which more closely resemble the original variables. This flag
5470 is enabled by default at @option{-O} and higher.
5473 Perform temporary expression replacement during the SSA->normal phase. Single
5474 use/single def temporaries are replaced at their use location with their
5475 defining expression. This results in non-GIMPLE code, but gives the expanders
5476 much more complex trees to work on resulting in better RTL generation. This is
5477 enabled by default at @option{-O} and higher.
5479 @item -ftree-vectorize
5480 Perform loop vectorization on trees.
5482 @item -ftree-vect-loop-version
5483 @opindex ftree-vect-loop-version
5484 Perform loop versioning when doing loop vectorization on trees. When a loop
5485 appears to be vectorizable except that data alignment or data dependence cannot
5486 be determined at compile time then vectorized and non-vectorized versions of
5487 the loop are generated along with runtime checks for alignment or dependence
5488 to control which version is executed. This option is enabled by default
5489 except at level @option{-Os} where it is disabled.
5492 Perform Value Range Propagation on trees. This is similar to the
5493 constant propagation pass, but instead of values, ranges of values are
5494 propagated. This allows the optimizers to remove unnecessary range
5495 checks like array bound checks and null pointer checks. This is
5496 enabled by default at @option{-O2} and higher. Null pointer check
5497 elimination is only done if @option{-fdelete-null-pointer-checks} is
5502 Perform tail duplication to enlarge superblock size. This transformation
5503 simplifies the control flow of the function allowing other optimizations to do
5506 @item -funroll-loops
5507 @opindex funroll-loops
5508 Unroll loops whose number of iterations can be determined at compile
5509 time or upon entry to the loop. @option{-funroll-loops} implies
5510 @option{-frerun-cse-after-loop}. This option makes code larger,
5511 and may or may not make it run faster.
5513 @item -funroll-all-loops
5514 @opindex funroll-all-loops
5515 Unroll all loops, even if their number of iterations is uncertain when
5516 the loop is entered. This usually makes programs run more slowly.
5517 @option{-funroll-all-loops} implies the same options as
5518 @option{-funroll-loops},
5520 @item -fsplit-ivs-in-unroller
5521 @opindex -fsplit-ivs-in-unroller
5522 Enables expressing of values of induction variables in later iterations
5523 of the unrolled loop using the value in the first iteration. This breaks
5524 long dependency chains, thus improving efficiency of the scheduling passes.
5526 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5527 same effect. However in cases the loop body is more complicated than
5528 a single basic block, this is not reliable. It also does not work at all
5529 on some of the architectures due to restrictions in the CSE pass.
5531 This optimization is enabled by default.
5533 @item -fvariable-expansion-in-unroller
5534 @opindex -fvariable-expansion-in-unroller
5535 With this option, the compiler will create multiple copies of some
5536 local variables when unrolling a loop which can result in superior code.
5538 @item -fprefetch-loop-arrays
5539 @opindex fprefetch-loop-arrays
5540 If supported by the target machine, generate instructions to prefetch
5541 memory to improve the performance of loops that access large arrays.
5543 This option may generate better or worse code; results are highly
5544 dependent on the structure of loops within the source code.
5546 Disabled at level @option{-Os}.
5549 @itemx -fno-peephole2
5550 @opindex fno-peephole
5551 @opindex fno-peephole2
5552 Disable any machine-specific peephole optimizations. The difference
5553 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5554 are implemented in the compiler; some targets use one, some use the
5555 other, a few use both.
5557 @option{-fpeephole} is enabled by default.
5558 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5560 @item -fno-guess-branch-probability
5561 @opindex fno-guess-branch-probability
5562 Do not guess branch probabilities using heuristics.
5564 GCC will use heuristics to guess branch probabilities if they are
5565 not provided by profiling feedback (@option{-fprofile-arcs}). These
5566 heuristics are based on the control flow graph. If some branch probabilities
5567 are specified by @samp{__builtin_expect}, then the heuristics will be
5568 used to guess branch probabilities for the rest of the control flow graph,
5569 taking the @samp{__builtin_expect} info into account. The interactions
5570 between the heuristics and @samp{__builtin_expect} can be complex, and in
5571 some cases, it may be useful to disable the heuristics so that the effects
5572 of @samp{__builtin_expect} are easier to understand.
5574 The default is @option{-fguess-branch-probability} at levels
5575 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5577 @item -freorder-blocks
5578 @opindex freorder-blocks
5579 Reorder basic blocks in the compiled function in order to reduce number of
5580 taken branches and improve code locality.
5582 Enabled at levels @option{-O2}, @option{-O3}.
5584 @item -freorder-blocks-and-partition
5585 @opindex freorder-blocks-and-partition
5586 In addition to reordering basic blocks in the compiled function, in order
5587 to reduce number of taken branches, partitions hot and cold basic blocks
5588 into separate sections of the assembly and .o files, to improve
5589 paging and cache locality performance.
5591 This optimization is automatically turned off in the presence of
5592 exception handling, for linkonce sections, for functions with a user-defined
5593 section attribute and on any architecture that does not support named
5596 @item -freorder-functions
5597 @opindex freorder-functions
5598 Reorder functions in the object file in order to
5599 improve code locality. This is implemented by using special
5600 subsections @code{.text.hot} for most frequently executed functions and
5601 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5602 the linker so object file format must support named sections and linker must
5603 place them in a reasonable way.
5605 Also profile feedback must be available in to make this option effective. See
5606 @option{-fprofile-arcs} for details.
5608 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5610 @item -fstrict-aliasing
5611 @opindex fstrict-aliasing
5612 Allows the compiler to assume the strictest aliasing rules applicable to
5613 the language being compiled. For C (and C++), this activates
5614 optimizations based on the type of expressions. In particular, an
5615 object of one type is assumed never to reside at the same address as an
5616 object of a different type, unless the types are almost the same. For
5617 example, an @code{unsigned int} can alias an @code{int}, but not a
5618 @code{void*} or a @code{double}. A character type may alias any other
5621 Pay special attention to code like this:
5634 The practice of reading from a different union member than the one most
5635 recently written to (called ``type-punning'') is common. Even with
5636 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5637 is accessed through the union type. So, the code above will work as
5638 expected. However, this code might not:
5649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5651 @item -fstrict-overflow
5652 @opindex fstrict-overflow
5653 Allow the compiler to assume strict signed overflow rules, depending
5654 on the language being compiled. For C (and C++) this means that
5655 overflow when doing arithmetic with signed numbers is undefined, which
5656 means that the compiler may assume that it will not happen. This
5657 permits various optimizations. For example, the compiler will assume
5658 that an expression like @code{i + 10 > i} will always be true for
5659 signed @code{i}. This assumption is only valid if signed overflow is
5660 undefined, as the expression is false if @code{i + 10} overflows when
5661 using twos complement arithmetic. When this option is in effect any
5662 attempt to determine whether an operation on signed numbers will
5663 overflow must be written carefully to not actually involve overflow.
5665 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5666 that signed overflow is fully defined: it wraps. When
5667 @option{-fwrapv} is used, there is no difference between
5668 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5669 @option{-fwrapv} certain types of overflow are permitted. For
5670 example, if the compiler gets an overflow when doing arithmetic on
5671 constants, the overflowed value can still be used with
5672 @option{-fwrapv}, but not otherwise.
5674 The @option{-fstrict-overflow} option is enabled at levels
5675 @option{-O2}, @option{-O3}, @option{-Os}.
5677 @item -falign-functions
5678 @itemx -falign-functions=@var{n}
5679 @opindex falign-functions
5680 Align the start of functions to the next power-of-two greater than
5681 @var{n}, skipping up to @var{n} bytes. For instance,
5682 @option{-falign-functions=32} aligns functions to the next 32-byte
5683 boundary, but @option{-falign-functions=24} would align to the next
5684 32-byte boundary only if this can be done by skipping 23 bytes or less.
5686 @option{-fno-align-functions} and @option{-falign-functions=1} are
5687 equivalent and mean that functions will not be aligned.
5689 Some assemblers only support this flag when @var{n} is a power of two;
5690 in that case, it is rounded up.
5692 If @var{n} is not specified or is zero, use a machine-dependent default.
5694 Enabled at levels @option{-O2}, @option{-O3}.
5696 @item -falign-labels
5697 @itemx -falign-labels=@var{n}
5698 @opindex falign-labels
5699 Align all branch targets to a power-of-two boundary, skipping up to
5700 @var{n} bytes like @option{-falign-functions}. This option can easily
5701 make code slower, because it must insert dummy operations for when the
5702 branch target is reached in the usual flow of the code.
5704 @option{-fno-align-labels} and @option{-falign-labels=1} are
5705 equivalent and mean that labels will not be aligned.
5707 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5708 are greater than this value, then their values are used instead.
5710 If @var{n} is not specified or is zero, use a machine-dependent default
5711 which is very likely to be @samp{1}, meaning no alignment.
5713 Enabled at levels @option{-O2}, @option{-O3}.
5716 @itemx -falign-loops=@var{n}
5717 @opindex falign-loops
5718 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5719 like @option{-falign-functions}. The hope is that the loop will be
5720 executed many times, which will make up for any execution of the dummy
5723 @option{-fno-align-loops} and @option{-falign-loops=1} are
5724 equivalent and mean that loops will not be aligned.
5726 If @var{n} is not specified or is zero, use a machine-dependent default.
5728 Enabled at levels @option{-O2}, @option{-O3}.
5731 @itemx -falign-jumps=@var{n}
5732 @opindex falign-jumps
5733 Align branch targets to a power-of-two boundary, for branch targets
5734 where the targets can only be reached by jumping, skipping up to @var{n}
5735 bytes like @option{-falign-functions}. In this case, no dummy operations
5738 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5739 equivalent and mean that loops will not be aligned.
5741 If @var{n} is not specified or is zero, use a machine-dependent default.
5743 Enabled at levels @option{-O2}, @option{-O3}.
5745 @item -funit-at-a-time
5746 @opindex funit-at-a-time
5747 Parse the whole compilation unit before starting to produce code.
5748 This allows some extra optimizations to take place but consumes
5749 more memory (in general). There are some compatibility issues
5750 with @emph{unit-at-a-time} mode:
5753 enabling @emph{unit-at-a-time} mode may change the order
5754 in which functions, variables, and top-level @code{asm} statements
5755 are emitted, and will likely break code relying on some particular
5756 ordering. The majority of such top-level @code{asm} statements,
5757 though, can be replaced by @code{section} attributes. The
5758 @option{fno-toplevel-reorder} option may be used to keep the ordering
5759 used in the input file, at the cost of some optimizations.
5762 @emph{unit-at-a-time} mode removes unreferenced static variables
5763 and functions. This may result in undefined references
5764 when an @code{asm} statement refers directly to variables or functions
5765 that are otherwise unused. In that case either the variable/function
5766 shall be listed as an operand of the @code{asm} statement operand or,
5767 in the case of top-level @code{asm} statements the attribute @code{used}
5768 shall be used on the declaration.
5771 Static functions now can use non-standard passing conventions that
5772 may break @code{asm} statements calling functions directly. Again,
5773 attribute @code{used} will prevent this behavior.
5776 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5777 but this scheme may not be supported by future releases of GCC@.
5779 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5781 @item -fno-toplevel-reorder
5782 Do not reorder top-level functions, variables, and @code{asm}
5783 statements. Output them in the same order that they appear in the
5784 input file. When this option is used, unreferenced static variables
5785 will not be removed. This option is intended to support existing code
5786 which relies on a particular ordering. For new code, it is better to
5791 Constructs webs as commonly used for register allocation purposes and assign
5792 each web individual pseudo register. This allows the register allocation pass
5793 to operate on pseudos directly, but also strengthens several other optimization
5794 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5795 however, make debugging impossible, since variables will no longer stay in a
5798 Enabled by default with @option{-funroll-loops}.
5800 @item -fwhole-program
5801 @opindex fwhole-program
5802 Assume that the current compilation unit represents whole program being
5803 compiled. All public functions and variables with the exception of @code{main}
5804 and those merged by attribute @code{externally_visible} become static functions
5805 and in a affect gets more aggressively optimized by interprocedural optimizers.
5806 While this option is equivalent to proper use of @code{static} keyword for
5807 programs consisting of single file, in combination with option
5808 @option{--combine} this flag can be used to compile most of smaller scale C
5809 programs since the functions and variables become local for the whole combined
5810 compilation unit, not for the single source file itself.
5813 @item -fno-cprop-registers
5814 @opindex fno-cprop-registers
5815 After register allocation and post-register allocation instruction splitting,
5816 we perform a copy-propagation pass to try to reduce scheduling dependencies
5817 and occasionally eliminate the copy.
5819 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5821 @item -fprofile-generate
5822 @opindex fprofile-generate
5824 Enable options usually used for instrumenting application to produce
5825 profile useful for later recompilation with profile feedback based
5826 optimization. You must use @option{-fprofile-generate} both when
5827 compiling and when linking your program.
5829 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5832 @opindex fprofile-use
5833 Enable profile feedback directed optimizations, and optimizations
5834 generally profitable only with profile feedback available.
5836 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5837 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5839 By default, GCC emits an error message if the feedback profiles do not
5840 match the source code. This error can be turned into a warning by using
5841 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
5845 The following options control compiler behavior regarding floating
5846 point arithmetic. These options trade off between speed and
5847 correctness. All must be specifically enabled.
5851 @opindex ffloat-store
5852 Do not store floating point variables in registers, and inhibit other
5853 options that might change whether a floating point value is taken from a
5856 @cindex floating point precision
5857 This option prevents undesirable excess precision on machines such as
5858 the 68000 where the floating registers (of the 68881) keep more
5859 precision than a @code{double} is supposed to have. Similarly for the
5860 x86 architecture. For most programs, the excess precision does only
5861 good, but a few programs rely on the precise definition of IEEE floating
5862 point. Use @option{-ffloat-store} for such programs, after modifying
5863 them to store all pertinent intermediate computations into variables.
5867 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5868 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5869 @option{-fno-rounding-math}, @option{-fno-signaling-nans},
5870 @option{-fno-signed-zeros} and @option{fcx-limited-range}.
5872 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5874 This option should never be turned on by any @option{-O} option since
5875 it can result in incorrect output for programs which depend on
5876 an exact implementation of IEEE or ISO rules/specifications for
5879 @item -fno-math-errno
5880 @opindex fno-math-errno
5881 Do not set ERRNO after calling math functions that are executed
5882 with a single instruction, e.g., sqrt. A program that relies on
5883 IEEE exceptions for math error handling may want to use this flag
5884 for speed while maintaining IEEE arithmetic compatibility.
5886 This option should never be turned on by any @option{-O} option since
5887 it can result in incorrect output for programs which depend on
5888 an exact implementation of IEEE or ISO rules/specifications for
5891 The default is @option{-fmath-errno}.
5893 On Darwin systems, the math library never sets @code{errno}. There is
5894 therefore no reason for the compiler to consider the possibility that
5895 it might, and @option{-fno-math-errno} is the default.
5897 @item -funsafe-math-optimizations
5898 @opindex funsafe-math-optimizations
5899 Allow optimizations for floating-point arithmetic that (a) assume
5900 that arguments and results are valid and (b) may violate IEEE or
5901 ANSI standards. When used at link-time, it may include libraries
5902 or startup files that change the default FPU control word or other
5903 similar optimizations.
5905 This option should never be turned on by any @option{-O} option since
5906 it can result in incorrect output for programs which depend on
5907 an exact implementation of IEEE or ISO rules/specifications for
5910 The default is @option{-fno-unsafe-math-optimizations}.
5912 @item -ffinite-math-only
5913 @opindex ffinite-math-only
5914 Allow optimizations for floating-point arithmetic that assume
5915 that arguments and results are not NaNs or +-Infs.
5917 This option should never be turned on by any @option{-O} option since
5918 it can result in incorrect output for programs which depend on
5919 an exact implementation of IEEE or ISO rules/specifications.
5921 The default is @option{-fno-finite-math-only}.
5923 @item -fno-signed-zeros
5924 @opindex fno-signed-zeros
5925 Allow optimizations for floating point arithmetic that ignore the
5926 signedness of zero. IEEE arithmetic specifies the behavior of
5927 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
5928 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
5929 This option implies that the sign of a zero result isn't significant.
5931 The default is @option{-fsigned-zeros}.
5933 @item -fno-trapping-math
5934 @opindex fno-trapping-math
5935 Compile code assuming that floating-point operations cannot generate
5936 user-visible traps. These traps include division by zero, overflow,
5937 underflow, inexact result and invalid operation. This option implies
5938 @option{-fno-signaling-nans}. Setting this option may allow faster
5939 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5941 This option should never be turned on by any @option{-O} option since
5942 it can result in incorrect output for programs which depend on
5943 an exact implementation of IEEE or ISO rules/specifications for
5946 The default is @option{-ftrapping-math}.
5948 @item -frounding-math
5949 @opindex frounding-math
5950 Disable transformations and optimizations that assume default floating
5951 point rounding behavior. This is round-to-zero for all floating point
5952 to integer conversions, and round-to-nearest for all other arithmetic
5953 truncations. This option should be specified for programs that change
5954 the FP rounding mode dynamically, or that may be executed with a
5955 non-default rounding mode. This option disables constant folding of
5956 floating point expressions at compile-time (which may be affected by
5957 rounding mode) and arithmetic transformations that are unsafe in the
5958 presence of sign-dependent rounding modes.
5960 The default is @option{-fno-rounding-math}.
5962 This option is experimental and does not currently guarantee to
5963 disable all GCC optimizations that are affected by rounding mode.
5964 Future versions of GCC may provide finer control of this setting
5965 using C99's @code{FENV_ACCESS} pragma. This command line option
5966 will be used to specify the default state for @code{FENV_ACCESS}.
5968 @item -frtl-abstract-sequences
5969 @opindex frtl-abstract-sequences
5970 It is a size optimization method. This option is to find identical
5971 sequences of code, which can be turned into pseudo-procedures and
5972 then replace all occurrences with calls to the newly created
5973 subroutine. It is kind of an opposite of @option{-finline-functions}.
5974 This optimization runs at RTL level.
5976 @item -fsignaling-nans
5977 @opindex fsignaling-nans
5978 Compile code assuming that IEEE signaling NaNs may generate user-visible
5979 traps during floating-point operations. Setting this option disables
5980 optimizations that may change the number of exceptions visible with
5981 signaling NaNs. This option implies @option{-ftrapping-math}.
5983 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5986 The default is @option{-fno-signaling-nans}.
5988 This option is experimental and does not currently guarantee to
5989 disable all GCC optimizations that affect signaling NaN behavior.
5991 @item -fsingle-precision-constant
5992 @opindex fsingle-precision-constant
5993 Treat floating point constant as single precision constant instead of
5994 implicitly converting it to double precision constant.
5996 @item -fcx-limited-range
5997 @itemx -fno-cx-limited-range
5998 @opindex fcx-limited-range
5999 @opindex fno-cx-limited-range
6000 When enabled, this option states that a range reduction step is not
6001 needed when performing complex division. The default is
6002 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
6004 This option controls the default setting of the ISO C99
6005 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6010 The following options control optimizations that may improve
6011 performance, but are not enabled by any @option{-O} options. This
6012 section includes experimental options that may produce broken code.
6015 @item -fbranch-probabilities
6016 @opindex fbranch-probabilities
6017 After running a program compiled with @option{-fprofile-arcs}
6018 (@pxref{Debugging Options,, Options for Debugging Your Program or
6019 @command{gcc}}), you can compile it a second time using
6020 @option{-fbranch-probabilities}, to improve optimizations based on
6021 the number of times each branch was taken. When the program
6022 compiled with @option{-fprofile-arcs} exits it saves arc execution
6023 counts to a file called @file{@var{sourcename}.gcda} for each source
6024 file. The information in this data file is very dependent on the
6025 structure of the generated code, so you must use the same source code
6026 and the same optimization options for both compilations.
6028 With @option{-fbranch-probabilities}, GCC puts a
6029 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6030 These can be used to improve optimization. Currently, they are only
6031 used in one place: in @file{reorg.c}, instead of guessing which path a
6032 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6033 exactly determine which path is taken more often.
6035 @item -fprofile-values
6036 @opindex fprofile-values
6037 If combined with @option{-fprofile-arcs}, it adds code so that some
6038 data about values of expressions in the program is gathered.
6040 With @option{-fbranch-probabilities}, it reads back the data gathered
6041 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6042 notes to instructions for their later usage in optimizations.
6044 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6048 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6049 a code to gather information about values of expressions.
6051 With @option{-fbranch-probabilities}, it reads back the data gathered
6052 and actually performs the optimizations based on them.
6053 Currently the optimizations include specialization of division operation
6054 using the knowledge about the value of the denominator.
6056 @item -frename-registers
6057 @opindex frename-registers
6058 Attempt to avoid false dependencies in scheduled code by making use
6059 of registers left over after register allocation. This optimization
6060 will most benefit processors with lots of registers. Depending on the
6061 debug information format adopted by the target, however, it can
6062 make debugging impossible, since variables will no longer stay in
6063 a ``home register''.
6065 Enabled by default with @option{-funroll-loops}.
6069 Perform tail duplication to enlarge superblock size. This transformation
6070 simplifies the control flow of the function allowing other optimizations to do
6073 Enabled with @option{-fprofile-use}.
6075 @item -funroll-loops
6076 @opindex funroll-loops
6077 Unroll loops whose number of iterations can be determined at compile time or
6078 upon entry to the loop. @option{-funroll-loops} implies
6079 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6080 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6081 small constant number of iterations). This option makes code larger, and may
6082 or may not make it run faster.
6084 Enabled with @option{-fprofile-use}.
6086 @item -funroll-all-loops
6087 @opindex funroll-all-loops
6088 Unroll all loops, even if their number of iterations is uncertain when
6089 the loop is entered. This usually makes programs run more slowly.
6090 @option{-funroll-all-loops} implies the same options as
6091 @option{-funroll-loops}.
6094 @opindex fpeel-loops
6095 Peels the loops for that there is enough information that they do not
6096 roll much (from profile feedback). It also turns on complete loop peeling
6097 (i.e.@: complete removal of loops with small constant number of iterations).
6099 Enabled with @option{-fprofile-use}.
6101 @item -fmove-loop-invariants
6102 @opindex fmove-loop-invariants
6103 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6104 at level @option{-O1}
6106 @item -funswitch-loops
6107 @opindex funswitch-loops
6108 Move branches with loop invariant conditions out of the loop, with duplicates
6109 of the loop on both branches (modified according to result of the condition).
6111 @item -ffunction-sections
6112 @itemx -fdata-sections
6113 @opindex ffunction-sections
6114 @opindex fdata-sections
6115 Place each function or data item into its own section in the output
6116 file if the target supports arbitrary sections. The name of the
6117 function or the name of the data item determines the section's name
6120 Use these options on systems where the linker can perform optimizations
6121 to improve locality of reference in the instruction space. Most systems
6122 using the ELF object format and SPARC processors running Solaris 2 have
6123 linkers with such optimizations. AIX may have these optimizations in
6126 Only use these options when there are significant benefits from doing
6127 so. When you specify these options, the assembler and linker will
6128 create larger object and executable files and will also be slower.
6129 You will not be able to use @code{gprof} on all systems if you
6130 specify this option and you may have problems with debugging if
6131 you specify both this option and @option{-g}.
6133 @item -fbranch-target-load-optimize
6134 @opindex fbranch-target-load-optimize
6135 Perform branch target register load optimization before prologue / epilogue
6137 The use of target registers can typically be exposed only during reload,
6138 thus hoisting loads out of loops and doing inter-block scheduling needs
6139 a separate optimization pass.
6141 @item -fbranch-target-load-optimize2
6142 @opindex fbranch-target-load-optimize2
6143 Perform branch target register load optimization after prologue / epilogue
6146 @item -fbtr-bb-exclusive
6147 @opindex fbtr-bb-exclusive
6148 When performing branch target register load optimization, don't reuse
6149 branch target registers in within any basic block.
6151 @item -fstack-protector
6152 Emit extra code to check for buffer overflows, such as stack smashing
6153 attacks. This is done by adding a guard variable to functions with
6154 vulnerable objects. This includes functions that call alloca, and
6155 functions with buffers larger than 8 bytes. The guards are initialized
6156 when a function is entered and then checked when the function exits.
6157 If a guard check fails, an error message is printed and the program exits.
6159 @item -fstack-protector-all
6160 Like @option{-fstack-protector} except that all functions are protected.
6162 @item -fsection-anchors
6163 @opindex fsection-anchors
6164 Try to reduce the number of symbolic address calculations by using
6165 shared ``anchor'' symbols to address nearby objects. This transformation
6166 can help to reduce the number of GOT entries and GOT accesses on some
6169 For example, the implementation of the following function @code{foo}:
6173 int foo (void) @{ return a + b + c; @}
6176 would usually calculate the addresses of all three variables, but if you
6177 compile it with @option{-fsection-anchors}, it will access the variables
6178 from a common anchor point instead. The effect is similar to the
6179 following pseudocode (which isn't valid C):
6184 register int *xr = &x;
6185 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6189 Not all targets support this option.
6191 @item --param @var{name}=@var{value}
6193 In some places, GCC uses various constants to control the amount of
6194 optimization that is done. For example, GCC will not inline functions
6195 that contain more that a certain number of instructions. You can
6196 control some of these constants on the command-line using the
6197 @option{--param} option.
6199 The names of specific parameters, and the meaning of the values, are
6200 tied to the internals of the compiler, and are subject to change
6201 without notice in future releases.
6203 In each case, the @var{value} is an integer. The allowable choices for
6204 @var{name} are given in the following table:
6207 @item salias-max-implicit-fields
6208 The maximum number of fields in a variable without direct
6209 structure accesses for which structure aliasing will consider trying
6210 to track each field. The default is 5
6212 @item salias-max-array-elements
6213 The maximum number of elements an array can have and its elements
6214 still be tracked individually by structure aliasing. The default is 4
6216 @item sra-max-structure-size
6217 The maximum structure size, in bytes, at which the scalar replacement
6218 of aggregates (SRA) optimization will perform block copies. The
6219 default value, 0, implies that GCC will select the most appropriate
6222 @item sra-field-structure-ratio
6223 The threshold ratio (as a percentage) between instantiated fields and
6224 the complete structure size. We say that if the ratio of the number
6225 of bytes in instantiated fields to the number of bytes in the complete
6226 structure exceeds this parameter, then block copies are not used. The
6229 @item max-crossjump-edges
6230 The maximum number of incoming edges to consider for crossjumping.
6231 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6232 the number of edges incoming to each block. Increasing values mean
6233 more aggressive optimization, making the compile time increase with
6234 probably small improvement in executable size.
6236 @item min-crossjump-insns
6237 The minimum number of instructions which must be matched at the end
6238 of two blocks before crossjumping will be performed on them. This
6239 value is ignored in the case where all instructions in the block being
6240 crossjumped from are matched. The default value is 5.
6242 @item max-grow-copy-bb-insns
6243 The maximum code size expansion factor when copying basic blocks
6244 instead of jumping. The expansion is relative to a jump instruction.
6245 The default value is 8.
6247 @item max-goto-duplication-insns
6248 The maximum number of instructions to duplicate to a block that jumps
6249 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6250 passes, GCC factors computed gotos early in the compilation process,
6251 and unfactors them as late as possible. Only computed jumps at the
6252 end of a basic blocks with no more than max-goto-duplication-insns are
6253 unfactored. The default value is 8.
6255 @item max-delay-slot-insn-search
6256 The maximum number of instructions to consider when looking for an
6257 instruction to fill a delay slot. If more than this arbitrary number of
6258 instructions is searched, the time savings from filling the delay slot
6259 will be minimal so stop searching. Increasing values mean more
6260 aggressive optimization, making the compile time increase with probably
6261 small improvement in executable run time.
6263 @item max-delay-slot-live-search
6264 When trying to fill delay slots, the maximum number of instructions to
6265 consider when searching for a block with valid live register
6266 information. Increasing this arbitrarily chosen value means more
6267 aggressive optimization, increasing the compile time. This parameter
6268 should be removed when the delay slot code is rewritten to maintain the
6271 @item max-gcse-memory
6272 The approximate maximum amount of memory that will be allocated in
6273 order to perform the global common subexpression elimination
6274 optimization. If more memory than specified is required, the
6275 optimization will not be done.
6277 @item max-gcse-passes
6278 The maximum number of passes of GCSE to run. The default is 1.
6280 @item max-pending-list-length
6281 The maximum number of pending dependencies scheduling will allow
6282 before flushing the current state and starting over. Large functions
6283 with few branches or calls can create excessively large lists which
6284 needlessly consume memory and resources.
6286 @item max-inline-insns-single
6287 Several parameters control the tree inliner used in gcc.
6288 This number sets the maximum number of instructions (counted in GCC's
6289 internal representation) in a single function that the tree inliner
6290 will consider for inlining. This only affects functions declared
6291 inline and methods implemented in a class declaration (C++).
6292 The default value is 450.
6294 @item max-inline-insns-auto
6295 When you use @option{-finline-functions} (included in @option{-O3}),
6296 a lot of functions that would otherwise not be considered for inlining
6297 by the compiler will be investigated. To those functions, a different
6298 (more restrictive) limit compared to functions declared inline can
6300 The default value is 90.
6302 @item large-function-insns
6303 The limit specifying really large functions. For functions larger than this
6304 limit after inlining inlining is constrained by
6305 @option{--param large-function-growth}. This parameter is useful primarily
6306 to avoid extreme compilation time caused by non-linear algorithms used by the
6308 This parameter is ignored when @option{-funit-at-a-time} is not used.
6309 The default value is 2700.
6311 @item large-function-growth
6312 Specifies maximal growth of large function caused by inlining in percents.
6313 This parameter is ignored when @option{-funit-at-a-time} is not used.
6314 The default value is 100 which limits large function growth to 2.0 times
6317 @item large-unit-insns
6318 The limit specifying large translation unit. Growth caused by inlining of
6319 units larger than this limit is limited by @option{--param inline-unit-growth}.
6320 For small units this might be too tight (consider unit consisting of function A
6321 that is inline and B that just calls A three time. If B is small relative to
6322 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6323 large units consisting of small inlininable functions however the overall unit
6324 growth limit is needed to avoid exponential explosion of code size. Thus for
6325 smaller units, the size is increased to @option{--param large-unit-insns}
6326 before applying @option{--param inline-unit-growth}. The default is 10000
6328 @item inline-unit-growth
6329 Specifies maximal overall growth of the compilation unit caused by inlining.
6330 This parameter is ignored when @option{-funit-at-a-time} is not used.
6331 The default value is 30 which limits unit growth to 1.3 times the original
6334 @item large-stack-frame
6335 The limit specifying large stack frames. While inlining the algorithm is trying
6336 to not grow past this limit too much. Default value is 256 bytes.
6338 @item large-stack-frame-growth
6339 Specifies maximal growth of large stack frames caused by inlining in percents.
6340 The default value is 1000 which limits large stack frame growth to 11 times
6343 @item max-inline-insns-recursive
6344 @itemx max-inline-insns-recursive-auto
6345 Specifies maximum number of instructions out-of-line copy of self recursive inline
6346 function can grow into by performing recursive inlining.
6348 For functions declared inline @option{--param max-inline-insns-recursive} is
6349 taken into account. For function not declared inline, recursive inlining
6350 happens only when @option{-finline-functions} (included in @option{-O3}) is
6351 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6352 default value is 450.
6354 @item max-inline-recursive-depth
6355 @itemx max-inline-recursive-depth-auto
6356 Specifies maximum recursion depth used by the recursive inlining.
6358 For functions declared inline @option{--param max-inline-recursive-depth} is
6359 taken into account. For function not declared inline, recursive inlining
6360 happens only when @option{-finline-functions} (included in @option{-O3}) is
6361 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6362 default value is 450.
6364 @item min-inline-recursive-probability
6365 Recursive inlining is profitable only for function having deep recursion
6366 in average and can hurt for function having little recursion depth by
6367 increasing the prologue size or complexity of function body to other
6370 When profile feedback is available (see @option{-fprofile-generate}) the actual
6371 recursion depth can be guessed from probability that function will recurse via
6372 given call expression. This parameter limits inlining only to call expression
6373 whose probability exceeds given threshold (in percents). The default value is
6376 @item inline-call-cost
6377 Specify cost of call instruction relative to simple arithmetics operations
6378 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6379 functions and at the same time increases size of leaf function that is believed to
6380 reduce function size by being inlined. In effect it increases amount of
6381 inlining for code having large abstraction penalty (many functions that just
6382 pass the arguments to other functions) and decrease inlining for code with low
6383 abstraction penalty. The default value is 16.
6385 @item min-vect-loop-bound
6386 The minimum number of iterations under which a loop will not get vectorized
6387 when @option{-ftree-vectorize} is used. The number of iterations after
6388 vectorization needs to be greater than the value specified by this option
6389 to allow vectorization. The default value is 0.
6391 @item max-unrolled-insns
6392 The maximum number of instructions that a loop should have if that loop
6393 is unrolled, and if the loop is unrolled, it determines how many times
6394 the loop code is unrolled.
6396 @item max-average-unrolled-insns
6397 The maximum number of instructions biased by probabilities of their execution
6398 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6399 it determines how many times the loop code is unrolled.
6401 @item max-unroll-times
6402 The maximum number of unrollings of a single loop.
6404 @item max-peeled-insns
6405 The maximum number of instructions that a loop should have if that loop
6406 is peeled, and if the loop is peeled, it determines how many times
6407 the loop code is peeled.
6409 @item max-peel-times
6410 The maximum number of peelings of a single loop.
6412 @item max-completely-peeled-insns
6413 The maximum number of insns of a completely peeled loop.
6415 @item max-completely-peel-times
6416 The maximum number of iterations of a loop to be suitable for complete peeling.
6418 @item max-unswitch-insns
6419 The maximum number of insns of an unswitched loop.
6421 @item max-unswitch-level
6422 The maximum number of branches unswitched in a single loop.
6425 The minimum cost of an expensive expression in the loop invariant motion.
6427 @item iv-consider-all-candidates-bound
6428 Bound on number of candidates for induction variables below that
6429 all candidates are considered for each use in induction variable
6430 optimizations. Only the most relevant candidates are considered
6431 if there are more candidates, to avoid quadratic time complexity.
6433 @item iv-max-considered-uses
6434 The induction variable optimizations give up on loops that contain more
6435 induction variable uses.
6437 @item iv-always-prune-cand-set-bound
6438 If number of candidates in the set is smaller than this value,
6439 we always try to remove unnecessary ivs from the set during its
6440 optimization when a new iv is added to the set.
6442 @item scev-max-expr-size
6443 Bound on size of expressions used in the scalar evolutions analyzer.
6444 Large expressions slow the analyzer.
6446 @item vect-max-version-checks
6447 The maximum number of runtime checks that can be performed when doing
6448 loop versioning in the vectorizer. See option ftree-vect-loop-version
6449 for more information.
6451 @item max-iterations-to-track
6453 The maximum number of iterations of a loop the brute force algorithm
6454 for analysis of # of iterations of the loop tries to evaluate.
6456 @item hot-bb-count-fraction
6457 Select fraction of the maximal count of repetitions of basic block in program
6458 given basic block needs to have to be considered hot.
6460 @item hot-bb-frequency-fraction
6461 Select fraction of the maximal frequency of executions of basic block in
6462 function given basic block needs to have to be considered hot
6464 @item max-predicted-iterations
6465 The maximum number of loop iterations we predict statically. This is useful
6466 in cases where function contain single loop with known bound and other loop
6467 with unknown. We predict the known number of iterations correctly, while
6468 the unknown number of iterations average to roughly 10. This means that the
6469 loop without bounds would appear artificially cold relative to the other one.
6471 @item tracer-dynamic-coverage
6472 @itemx tracer-dynamic-coverage-feedback
6474 This value is used to limit superblock formation once the given percentage of
6475 executed instructions is covered. This limits unnecessary code size
6478 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6479 feedback is available. The real profiles (as opposed to statically estimated
6480 ones) are much less balanced allowing the threshold to be larger value.
6482 @item tracer-max-code-growth
6483 Stop tail duplication once code growth has reached given percentage. This is
6484 rather hokey argument, as most of the duplicates will be eliminated later in
6485 cross jumping, so it may be set to much higher values than is the desired code
6488 @item tracer-min-branch-ratio
6490 Stop reverse growth when the reverse probability of best edge is less than this
6491 threshold (in percent).
6493 @item tracer-min-branch-ratio
6494 @itemx tracer-min-branch-ratio-feedback
6496 Stop forward growth if the best edge do have probability lower than this
6499 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6500 compilation for profile feedback and one for compilation without. The value
6501 for compilation with profile feedback needs to be more conservative (higher) in
6502 order to make tracer effective.
6504 @item max-cse-path-length
6506 Maximum number of basic blocks on path that cse considers. The default is 10.
6509 The maximum instructions CSE process before flushing. The default is 1000.
6511 @item max-aliased-vops
6513 Maximum number of virtual operands per statement allowed to represent
6514 aliases before triggering the alias grouping heuristic. Alias
6515 grouping reduces compile times and memory consumption needed for
6516 aliasing at the expense of precision loss in alias information.
6518 @item ggc-min-expand
6520 GCC uses a garbage collector to manage its own memory allocation. This
6521 parameter specifies the minimum percentage by which the garbage
6522 collector's heap should be allowed to expand between collections.
6523 Tuning this may improve compilation speed; it has no effect on code
6526 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6527 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6528 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6529 GCC is not able to calculate RAM on a particular platform, the lower
6530 bound of 30% is used. Setting this parameter and
6531 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6532 every opportunity. This is extremely slow, but can be useful for
6535 @item ggc-min-heapsize
6537 Minimum size of the garbage collector's heap before it begins bothering
6538 to collect garbage. The first collection occurs after the heap expands
6539 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6540 tuning this may improve compilation speed, and has no effect on code
6543 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6544 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6545 with a lower bound of 4096 (four megabytes) and an upper bound of
6546 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6547 particular platform, the lower bound is used. Setting this parameter
6548 very large effectively disables garbage collection. Setting this
6549 parameter and @option{ggc-min-expand} to zero causes a full collection
6550 to occur at every opportunity.
6552 @item max-reload-search-insns
6553 The maximum number of instruction reload should look backward for equivalent
6554 register. Increasing values mean more aggressive optimization, making the
6555 compile time increase with probably slightly better performance. The default
6558 @item max-cselib-memory-locations
6559 The maximum number of memory locations cselib should take into account.
6560 Increasing values mean more aggressive optimization, making the compile time
6561 increase with probably slightly better performance. The default value is 500.
6563 @item max-flow-memory-locations
6564 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6565 The default value is 100.
6567 @item reorder-blocks-duplicate
6568 @itemx reorder-blocks-duplicate-feedback
6570 Used by basic block reordering pass to decide whether to use unconditional
6571 branch or duplicate the code on its destination. Code is duplicated when its
6572 estimated size is smaller than this value multiplied by the estimated size of
6573 unconditional jump in the hot spots of the program.
6575 The @option{reorder-block-duplicate-feedback} is used only when profile
6576 feedback is available and may be set to higher values than
6577 @option{reorder-block-duplicate} since information about the hot spots is more
6580 @item max-sched-ready-insns
6581 The maximum number of instructions ready to be issued the scheduler should
6582 consider at any given time during the first scheduling pass. Increasing
6583 values mean more thorough searches, making the compilation time increase
6584 with probably little benefit. The default value is 100.
6586 @item max-sched-region-blocks
6587 The maximum number of blocks in a region to be considered for
6588 interblock scheduling. The default value is 10.
6590 @item max-sched-region-insns
6591 The maximum number of insns in a region to be considered for
6592 interblock scheduling. The default value is 100.
6595 The minimum probability (in percents) of reaching a source block
6596 for interblock speculative scheduling. The default value is 40.
6598 @item max-sched-extend-regions-iters
6599 The maximum number of iterations through CFG to extend regions.
6600 0 - disable region extension,
6601 N - do at most N iterations.
6602 The default value is 0.
6604 @item max-sched-insn-conflict-delay
6605 The maximum conflict delay for an insn to be considered for speculative motion.
6606 The default value is 3.
6608 @item sched-spec-prob-cutoff
6609 The minimal probability of speculation success (in percents), so that
6610 speculative insn will be scheduled.
6611 The default value is 40.
6613 @item max-last-value-rtl
6615 The maximum size measured as number of RTLs that can be recorded in an expression
6616 in combiner for a pseudo register as last known value of that register. The default
6619 @item integer-share-limit
6620 Small integer constants can use a shared data structure, reducing the
6621 compiler's memory usage and increasing its speed. This sets the maximum
6622 value of a shared integer constant's. The default value is 256.
6624 @item min-virtual-mappings
6625 Specifies the minimum number of virtual mappings in the incremental
6626 SSA updater that should be registered to trigger the virtual mappings
6627 heuristic defined by virtual-mappings-ratio. The default value is
6630 @item virtual-mappings-ratio
6631 If the number of virtual mappings is virtual-mappings-ratio bigger
6632 than the number of virtual symbols to be updated, then the incremental
6633 SSA updater switches to a full update for those symbols. The default
6636 @item ssp-buffer-size
6637 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6638 protection when @option{-fstack-protection} is used.
6640 @item max-jump-thread-duplication-stmts
6641 Maximum number of statements allowed in a block that needs to be
6642 duplicated when threading jumps.
6644 @item max-fields-for-field-sensitive
6645 Maximum number of fields in a structure we will treat in
6646 a field sensitive manner during pointer analysis.
6648 @item prefetch-latency
6649 Estimate on average number of instructions that are executed before
6650 prefetch finishes. The distance we prefetch ahead is proportional
6651 to this constant. Increasing this number may also lead to less
6652 streams being prefetched (see @option{simultaneous-prefetches}).
6654 @item simultaneous-prefetches
6655 Maximum number of prefetches that can run at the same time.
6657 @item l1-cache-line-size
6658 The size of cache line in L1 cache, in bytes.
6661 The number of cache lines in L1 cache.
6663 @item verify-canonical-types
6664 Whether the compiler should verify the ``canonical'' types used for
6665 type equality comparisons within the C++ and Objective-C++ front
6666 ends. Set to 1 (the default when GCC is configured with
6667 --enable-checking) to enable verification, 0 to disable verification
6668 (the default when GCC is configured with --disable-checking).
6673 @node Preprocessor Options
6674 @section Options Controlling the Preprocessor
6675 @cindex preprocessor options
6676 @cindex options, preprocessor
6678 These options control the C preprocessor, which is run on each C source
6679 file before actual compilation.
6681 If you use the @option{-E} option, nothing is done except preprocessing.
6682 Some of these options make sense only together with @option{-E} because
6683 they cause the preprocessor output to be unsuitable for actual
6688 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6689 and pass @var{option} directly through to the preprocessor. If
6690 @var{option} contains commas, it is split into multiple options at the
6691 commas. However, many options are modified, translated or interpreted
6692 by the compiler driver before being passed to the preprocessor, and
6693 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6694 interface is undocumented and subject to change, so whenever possible
6695 you should avoid using @option{-Wp} and let the driver handle the
6698 @item -Xpreprocessor @var{option}
6699 @opindex preprocessor
6700 Pass @var{option} as an option to the preprocessor. You can use this to
6701 supply system-specific preprocessor options which GCC does not know how to
6704 If you want to pass an option that takes an argument, you must use
6705 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6708 @include cppopts.texi
6710 @node Assembler Options
6711 @section Passing Options to the Assembler
6713 @c prevent bad page break with this line
6714 You can pass options to the assembler.
6717 @item -Wa,@var{option}
6719 Pass @var{option} as an option to the assembler. If @var{option}
6720 contains commas, it is split into multiple options at the commas.
6722 @item -Xassembler @var{option}
6724 Pass @var{option} as an option to the assembler. You can use this to
6725 supply system-specific assembler options which GCC does not know how to
6728 If you want to pass an option that takes an argument, you must use
6729 @option{-Xassembler} twice, once for the option and once for the argument.
6734 @section Options for Linking
6735 @cindex link options
6736 @cindex options, linking
6738 These options come into play when the compiler links object files into
6739 an executable output file. They are meaningless if the compiler is
6740 not doing a link step.
6744 @item @var{object-file-name}
6745 A file name that does not end in a special recognized suffix is
6746 considered to name an object file or library. (Object files are
6747 distinguished from libraries by the linker according to the file
6748 contents.) If linking is done, these object files are used as input
6757 If any of these options is used, then the linker is not run, and
6758 object file names should not be used as arguments. @xref{Overall
6762 @item -l@var{library}
6763 @itemx -l @var{library}
6765 Search the library named @var{library} when linking. (The second
6766 alternative with the library as a separate argument is only for
6767 POSIX compliance and is not recommended.)
6769 It makes a difference where in the command you write this option; the
6770 linker searches and processes libraries and object files in the order they
6771 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6772 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6773 to functions in @samp{z}, those functions may not be loaded.
6775 The linker searches a standard list of directories for the library,
6776 which is actually a file named @file{lib@var{library}.a}. The linker
6777 then uses this file as if it had been specified precisely by name.
6779 The directories searched include several standard system directories
6780 plus any that you specify with @option{-L}.
6782 Normally the files found this way are library files---archive files
6783 whose members are object files. The linker handles an archive file by
6784 scanning through it for members which define symbols that have so far
6785 been referenced but not defined. But if the file that is found is an
6786 ordinary object file, it is linked in the usual fashion. The only
6787 difference between using an @option{-l} option and specifying a file name
6788 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6789 and searches several directories.
6793 You need this special case of the @option{-l} option in order to
6794 link an Objective-C or Objective-C++ program.
6797 @opindex nostartfiles
6798 Do not use the standard system startup files when linking.
6799 The standard system libraries are used normally, unless @option{-nostdlib}
6800 or @option{-nodefaultlibs} is used.
6802 @item -nodefaultlibs
6803 @opindex nodefaultlibs
6804 Do not use the standard system libraries when linking.
6805 Only the libraries you specify will be passed to the linker.
6806 The standard startup files are used normally, unless @option{-nostartfiles}
6807 is used. The compiler may generate calls to @code{memcmp},
6808 @code{memset}, @code{memcpy} and @code{memmove}.
6809 These entries are usually resolved by entries in
6810 libc. These entry points should be supplied through some other
6811 mechanism when this option is specified.
6815 Do not use the standard system startup files or libraries when linking.
6816 No startup files and only the libraries you specify will be passed to
6817 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6818 @code{memcpy} and @code{memmove}.
6819 These entries are usually resolved by entries in
6820 libc. These entry points should be supplied through some other
6821 mechanism when this option is specified.
6823 @cindex @option{-lgcc}, use with @option{-nostdlib}
6824 @cindex @option{-nostdlib} and unresolved references
6825 @cindex unresolved references and @option{-nostdlib}
6826 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6827 @cindex @option{-nodefaultlibs} and unresolved references
6828 @cindex unresolved references and @option{-nodefaultlibs}
6829 One of the standard libraries bypassed by @option{-nostdlib} and
6830 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6831 that GCC uses to overcome shortcomings of particular machines, or special
6832 needs for some languages.
6833 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6834 Collection (GCC) Internals},
6835 for more discussion of @file{libgcc.a}.)
6836 In most cases, you need @file{libgcc.a} even when you want to avoid
6837 other standard libraries. In other words, when you specify @option{-nostdlib}
6838 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6839 This ensures that you have no unresolved references to internal GCC
6840 library subroutines. (For example, @samp{__main}, used to ensure C++
6841 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6842 GNU Compiler Collection (GCC) Internals}.)
6846 Produce a position independent executable on targets which support it.
6847 For predictable results, you must also specify the same set of options
6848 that were used to generate code (@option{-fpie}, @option{-fPIE},
6849 or model suboptions) when you specify this option.
6853 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6854 that support it. This instructs the linker to add all symbols, not
6855 only used ones, to the dynamic symbol table. This option is needed
6856 for some uses of @code{dlopen} or to allow obtaining backtraces
6857 from within a program.
6861 Remove all symbol table and relocation information from the executable.
6865 On systems that support dynamic linking, this prevents linking with the shared
6866 libraries. On other systems, this option has no effect.
6870 Produce a shared object which can then be linked with other objects to
6871 form an executable. Not all systems support this option. For predictable
6872 results, you must also specify the same set of options that were used to
6873 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6874 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6875 needs to build supplementary stub code for constructors to work. On
6876 multi-libbed systems, @samp{gcc -shared} must select the correct support
6877 libraries to link against. Failing to supply the correct flags may lead
6878 to subtle defects. Supplying them in cases where they are not necessary
6881 @item -shared-libgcc
6882 @itemx -static-libgcc
6883 @opindex shared-libgcc
6884 @opindex static-libgcc
6885 On systems that provide @file{libgcc} as a shared library, these options
6886 force the use of either the shared or static version respectively.
6887 If no shared version of @file{libgcc} was built when the compiler was
6888 configured, these options have no effect.
6890 There are several situations in which an application should use the
6891 shared @file{libgcc} instead of the static version. The most common
6892 of these is when the application wishes to throw and catch exceptions
6893 across different shared libraries. In that case, each of the libraries
6894 as well as the application itself should use the shared @file{libgcc}.
6896 Therefore, the G++ and GCJ drivers automatically add
6897 @option{-shared-libgcc} whenever you build a shared library or a main
6898 executable, because C++ and Java programs typically use exceptions, so
6899 this is the right thing to do.
6901 If, instead, you use the GCC driver to create shared libraries, you may
6902 find that they will not always be linked with the shared @file{libgcc}.
6903 If GCC finds, at its configuration time, that you have a non-GNU linker
6904 or a GNU linker that does not support option @option{--eh-frame-hdr},
6905 it will link the shared version of @file{libgcc} into shared libraries
6906 by default. Otherwise, it will take advantage of the linker and optimize
6907 away the linking with the shared version of @file{libgcc}, linking with
6908 the static version of libgcc by default. This allows exceptions to
6909 propagate through such shared libraries, without incurring relocation
6910 costs at library load time.
6912 However, if a library or main executable is supposed to throw or catch
6913 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6914 for the languages used in the program, or using the option
6915 @option{-shared-libgcc}, such that it is linked with the shared
6920 Bind references to global symbols when building a shared object. Warn
6921 about any unresolved references (unless overridden by the link editor
6922 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6925 @item -Xlinker @var{option}
6927 Pass @var{option} as an option to the linker. You can use this to
6928 supply system-specific linker options which GCC does not know how to
6931 If you want to pass an option that takes an argument, you must use
6932 @option{-Xlinker} twice, once for the option and once for the argument.
6933 For example, to pass @option{-assert definitions}, you must write
6934 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6935 @option{-Xlinker "-assert definitions"}, because this passes the entire
6936 string as a single argument, which is not what the linker expects.
6938 @item -Wl,@var{option}
6940 Pass @var{option} as an option to the linker. If @var{option} contains
6941 commas, it is split into multiple options at the commas.
6943 @item -u @var{symbol}
6945 Pretend the symbol @var{symbol} is undefined, to force linking of
6946 library modules to define it. You can use @option{-u} multiple times with
6947 different symbols to force loading of additional library modules.
6950 @node Directory Options
6951 @section Options for Directory Search
6952 @cindex directory options
6953 @cindex options, directory search
6956 These options specify directories to search for header files, for
6957 libraries and for parts of the compiler:
6962 Add the directory @var{dir} to the head of the list of directories to be
6963 searched for header files. This can be used to override a system header
6964 file, substituting your own version, since these directories are
6965 searched before the system header file directories. However, you should
6966 not use this option to add directories that contain vendor-supplied
6967 system header files (use @option{-isystem} for that). If you use more than
6968 one @option{-I} option, the directories are scanned in left-to-right
6969 order; the standard system directories come after.
6971 If a standard system include directory, or a directory specified with
6972 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6973 option will be ignored. The directory will still be searched but as a
6974 system directory at its normal position in the system include chain.
6975 This is to ensure that GCC's procedure to fix buggy system headers and
6976 the ordering for the include_next directive are not inadvertently changed.
6977 If you really need to change the search order for system directories,
6978 use the @option{-nostdinc} and/or @option{-isystem} options.
6980 @item -iquote@var{dir}
6982 Add the directory @var{dir} to the head of the list of directories to
6983 be searched for header files only for the case of @samp{#include
6984 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6985 otherwise just like @option{-I}.
6989 Add directory @var{dir} to the list of directories to be searched
6992 @item -B@var{prefix}
6994 This option specifies where to find the executables, libraries,
6995 include files, and data files of the compiler itself.
6997 The compiler driver program runs one or more of the subprograms
6998 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6999 @var{prefix} as a prefix for each program it tries to run, both with and
7000 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7002 For each subprogram to be run, the compiler driver first tries the
7003 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7004 was not specified, the driver tries two standard prefixes, which are
7005 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7006 those results in a file name that is found, the unmodified program
7007 name is searched for using the directories specified in your
7008 @env{PATH} environment variable.
7010 The compiler will check to see if the path provided by the @option{-B}
7011 refers to a directory, and if necessary it will add a directory
7012 separator character at the end of the path.
7014 @option{-B} prefixes that effectively specify directory names also apply
7015 to libraries in the linker, because the compiler translates these
7016 options into @option{-L} options for the linker. They also apply to
7017 includes files in the preprocessor, because the compiler translates these
7018 options into @option{-isystem} options for the preprocessor. In this case,
7019 the compiler appends @samp{include} to the prefix.
7021 The run-time support file @file{libgcc.a} can also be searched for using
7022 the @option{-B} prefix, if needed. If it is not found there, the two
7023 standard prefixes above are tried, and that is all. The file is left
7024 out of the link if it is not found by those means.
7026 Another way to specify a prefix much like the @option{-B} prefix is to use
7027 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7030 As a special kludge, if the path provided by @option{-B} is
7031 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7032 9, then it will be replaced by @file{[dir/]include}. This is to help
7033 with boot-strapping the compiler.
7035 @item -specs=@var{file}
7037 Process @var{file} after the compiler reads in the standard @file{specs}
7038 file, in order to override the defaults that the @file{gcc} driver
7039 program uses when determining what switches to pass to @file{cc1},
7040 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7041 @option{-specs=@var{file}} can be specified on the command line, and they
7042 are processed in order, from left to right.
7044 @item --sysroot=@var{dir}
7046 Use @var{dir} as the logical root directory for headers and libraries.
7047 For example, if the compiler would normally search for headers in
7048 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7049 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7051 If you use both this option and the @option{-isysroot} option, then
7052 the @option{--sysroot} option will apply to libraries, but the
7053 @option{-isysroot} option will apply to header files.
7055 The GNU linker (beginning with version 2.16) has the necessary support
7056 for this option. If your linker does not support this option, the
7057 header file aspect of @option{--sysroot} will still work, but the
7058 library aspect will not.
7062 This option has been deprecated. Please use @option{-iquote} instead for
7063 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7064 Any directories you specify with @option{-I} options before the @option{-I-}
7065 option are searched only for the case of @samp{#include "@var{file}"};
7066 they are not searched for @samp{#include <@var{file}>}.
7068 If additional directories are specified with @option{-I} options after
7069 the @option{-I-}, these directories are searched for all @samp{#include}
7070 directives. (Ordinarily @emph{all} @option{-I} directories are used
7073 In addition, the @option{-I-} option inhibits the use of the current
7074 directory (where the current input file came from) as the first search
7075 directory for @samp{#include "@var{file}"}. There is no way to
7076 override this effect of @option{-I-}. With @option{-I.} you can specify
7077 searching the directory which was current when the compiler was
7078 invoked. That is not exactly the same as what the preprocessor does
7079 by default, but it is often satisfactory.
7081 @option{-I-} does not inhibit the use of the standard system directories
7082 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7089 @section Specifying subprocesses and the switches to pass to them
7092 @command{gcc} is a driver program. It performs its job by invoking a
7093 sequence of other programs to do the work of compiling, assembling and
7094 linking. GCC interprets its command-line parameters and uses these to
7095 deduce which programs it should invoke, and which command-line options
7096 it ought to place on their command lines. This behavior is controlled
7097 by @dfn{spec strings}. In most cases there is one spec string for each
7098 program that GCC can invoke, but a few programs have multiple spec
7099 strings to control their behavior. The spec strings built into GCC can
7100 be overridden by using the @option{-specs=} command-line switch to specify
7103 @dfn{Spec files} are plaintext files that are used to construct spec
7104 strings. They consist of a sequence of directives separated by blank
7105 lines. The type of directive is determined by the first non-whitespace
7106 character on the line and it can be one of the following:
7109 @item %@var{command}
7110 Issues a @var{command} to the spec file processor. The commands that can
7114 @item %include <@var{file}>
7116 Search for @var{file} and insert its text at the current point in the
7119 @item %include_noerr <@var{file}>
7120 @cindex %include_noerr
7121 Just like @samp{%include}, but do not generate an error message if the include
7122 file cannot be found.
7124 @item %rename @var{old_name} @var{new_name}
7126 Rename the spec string @var{old_name} to @var{new_name}.
7130 @item *[@var{spec_name}]:
7131 This tells the compiler to create, override or delete the named spec
7132 string. All lines after this directive up to the next directive or
7133 blank line are considered to be the text for the spec string. If this
7134 results in an empty string then the spec will be deleted. (Or, if the
7135 spec did not exist, then nothing will happened.) Otherwise, if the spec
7136 does not currently exist a new spec will be created. If the spec does
7137 exist then its contents will be overridden by the text of this
7138 directive, unless the first character of that text is the @samp{+}
7139 character, in which case the text will be appended to the spec.
7141 @item [@var{suffix}]:
7142 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7143 and up to the next directive or blank line are considered to make up the
7144 spec string for the indicated suffix. When the compiler encounters an
7145 input file with the named suffix, it will processes the spec string in
7146 order to work out how to compile that file. For example:
7153 This says that any input file whose name ends in @samp{.ZZ} should be
7154 passed to the program @samp{z-compile}, which should be invoked with the
7155 command-line switch @option{-input} and with the result of performing the
7156 @samp{%i} substitution. (See below.)
7158 As an alternative to providing a spec string, the text that follows a
7159 suffix directive can be one of the following:
7162 @item @@@var{language}
7163 This says that the suffix is an alias for a known @var{language}. This is
7164 similar to using the @option{-x} command-line switch to GCC to specify a
7165 language explicitly. For example:
7172 Says that .ZZ files are, in fact, C++ source files.
7175 This causes an error messages saying:
7178 @var{name} compiler not installed on this system.
7182 GCC already has an extensive list of suffixes built into it.
7183 This directive will add an entry to the end of the list of suffixes, but
7184 since the list is searched from the end backwards, it is effectively
7185 possible to override earlier entries using this technique.
7189 GCC has the following spec strings built into it. Spec files can
7190 override these strings or create their own. Note that individual
7191 targets can also add their own spec strings to this list.
7194 asm Options to pass to the assembler
7195 asm_final Options to pass to the assembler post-processor
7196 cpp Options to pass to the C preprocessor
7197 cc1 Options to pass to the C compiler
7198 cc1plus Options to pass to the C++ compiler
7199 endfile Object files to include at the end of the link
7200 link Options to pass to the linker
7201 lib Libraries to include on the command line to the linker
7202 libgcc Decides which GCC support library to pass to the linker
7203 linker Sets the name of the linker
7204 predefines Defines to be passed to the C preprocessor
7205 signed_char Defines to pass to CPP to say whether @code{char} is signed
7207 startfile Object files to include at the start of the link
7210 Here is a small example of a spec file:
7216 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7219 This example renames the spec called @samp{lib} to @samp{old_lib} and
7220 then overrides the previous definition of @samp{lib} with a new one.
7221 The new definition adds in some extra command-line options before
7222 including the text of the old definition.
7224 @dfn{Spec strings} are a list of command-line options to be passed to their
7225 corresponding program. In addition, the spec strings can contain
7226 @samp{%}-prefixed sequences to substitute variable text or to
7227 conditionally insert text into the command line. Using these constructs
7228 it is possible to generate quite complex command lines.
7230 Here is a table of all defined @samp{%}-sequences for spec
7231 strings. Note that spaces are not generated automatically around the
7232 results of expanding these sequences. Therefore you can concatenate them
7233 together or combine them with constant text in a single argument.
7237 Substitute one @samp{%} into the program name or argument.
7240 Substitute the name of the input file being processed.
7243 Substitute the basename of the input file being processed.
7244 This is the substring up to (and not including) the last period
7245 and not including the directory.
7248 This is the same as @samp{%b}, but include the file suffix (text after
7252 Marks the argument containing or following the @samp{%d} as a
7253 temporary file name, so that that file will be deleted if GCC exits
7254 successfully. Unlike @samp{%g}, this contributes no text to the
7257 @item %g@var{suffix}
7258 Substitute a file name that has suffix @var{suffix} and is chosen
7259 once per compilation, and mark the argument in the same way as
7260 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7261 name is now chosen in a way that is hard to predict even when previously
7262 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7263 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7264 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7265 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7266 was simply substituted with a file name chosen once per compilation,
7267 without regard to any appended suffix (which was therefore treated
7268 just like ordinary text), making such attacks more likely to succeed.
7270 @item %u@var{suffix}
7271 Like @samp{%g}, but generates a new temporary file name even if
7272 @samp{%u@var{suffix}} was already seen.
7274 @item %U@var{suffix}
7275 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7276 new one if there is no such last file name. In the absence of any
7277 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7278 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7279 would involve the generation of two distinct file names, one
7280 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7281 simply substituted with a file name chosen for the previous @samp{%u},
7282 without regard to any appended suffix.
7284 @item %j@var{suffix}
7285 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7286 writable, and if save-temps is off; otherwise, substitute the name
7287 of a temporary file, just like @samp{%u}. This temporary file is not
7288 meant for communication between processes, but rather as a junk
7291 @item %|@var{suffix}
7292 @itemx %m@var{suffix}
7293 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7294 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7295 all. These are the two most common ways to instruct a program that it
7296 should read from standard input or write to standard output. If you
7297 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7298 construct: see for example @file{f/lang-specs.h}.
7300 @item %.@var{SUFFIX}
7301 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7302 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7303 terminated by the next space or %.
7306 Marks the argument containing or following the @samp{%w} as the
7307 designated output file of this compilation. This puts the argument
7308 into the sequence of arguments that @samp{%o} will substitute later.
7311 Substitutes the names of all the output files, with spaces
7312 automatically placed around them. You should write spaces
7313 around the @samp{%o} as well or the results are undefined.
7314 @samp{%o} is for use in the specs for running the linker.
7315 Input files whose names have no recognized suffix are not compiled
7316 at all, but they are included among the output files, so they will
7320 Substitutes the suffix for object files. Note that this is
7321 handled specially when it immediately follows @samp{%g, %u, or %U},
7322 because of the need for those to form complete file names. The
7323 handling is such that @samp{%O} is treated exactly as if it had already
7324 been substituted, except that @samp{%g, %u, and %U} do not currently
7325 support additional @var{suffix} characters following @samp{%O} as they would
7326 following, for example, @samp{.o}.
7329 Substitutes the standard macro predefinitions for the
7330 current target machine. Use this when running @code{cpp}.
7333 Like @samp{%p}, but puts @samp{__} before and after the name of each
7334 predefined macro, except for macros that start with @samp{__} or with
7335 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7339 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7340 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7341 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7342 and @option{-imultilib} as necessary.
7345 Current argument is the name of a library or startup file of some sort.
7346 Search for that file in a standard list of directories and substitute
7347 the full name found.
7350 Print @var{str} as an error message. @var{str} is terminated by a newline.
7351 Use this when inconsistent options are detected.
7354 Substitute the contents of spec string @var{name} at this point.
7357 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7359 @item %x@{@var{option}@}
7360 Accumulate an option for @samp{%X}.
7363 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7367 Output the accumulated assembler options specified by @option{-Wa}.
7370 Output the accumulated preprocessor options specified by @option{-Wp}.
7373 Process the @code{asm} spec. This is used to compute the
7374 switches to be passed to the assembler.
7377 Process the @code{asm_final} spec. This is a spec string for
7378 passing switches to an assembler post-processor, if such a program is
7382 Process the @code{link} spec. This is the spec for computing the
7383 command line passed to the linker. Typically it will make use of the
7384 @samp{%L %G %S %D and %E} sequences.
7387 Dump out a @option{-L} option for each directory that GCC believes might
7388 contain startup files. If the target supports multilibs then the
7389 current multilib directory will be prepended to each of these paths.
7392 Process the @code{lib} spec. This is a spec string for deciding which
7393 libraries should be included on the command line to the linker.
7396 Process the @code{libgcc} spec. This is a spec string for deciding
7397 which GCC support library should be included on the command line to the linker.
7400 Process the @code{startfile} spec. This is a spec for deciding which
7401 object files should be the first ones passed to the linker. Typically
7402 this might be a file named @file{crt0.o}.
7405 Process the @code{endfile} spec. This is a spec string that specifies
7406 the last object files that will be passed to the linker.
7409 Process the @code{cpp} spec. This is used to construct the arguments
7410 to be passed to the C preprocessor.
7413 Process the @code{cc1} spec. This is used to construct the options to be
7414 passed to the actual C compiler (@samp{cc1}).
7417 Process the @code{cc1plus} spec. This is used to construct the options to be
7418 passed to the actual C++ compiler (@samp{cc1plus}).
7421 Substitute the variable part of a matched option. See below.
7422 Note that each comma in the substituted string is replaced by
7426 Remove all occurrences of @code{-S} from the command line. Note---this
7427 command is position dependent. @samp{%} commands in the spec string
7428 before this one will see @code{-S}, @samp{%} commands in the spec string
7429 after this one will not.
7431 @item %:@var{function}(@var{args})
7432 Call the named function @var{function}, passing it @var{args}.
7433 @var{args} is first processed as a nested spec string, then split
7434 into an argument vector in the usual fashion. The function returns
7435 a string which is processed as if it had appeared literally as part
7436 of the current spec.
7438 The following built-in spec functions are provided:
7442 The @code{getenv} spec function takes two arguments: an environment
7443 variable name and a string. If the environment variable is not
7444 defined, a fatal error is issued. Otherwise, the return value is the
7445 value of the environment variable concatenated with the string. For
7446 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
7449 %:getenv(TOPDIR /include)
7452 expands to @file{/path/to/top/include}.
7454 @item @code{if-exists}
7455 The @code{if-exists} spec function takes one argument, an absolute
7456 pathname to a file. If the file exists, @code{if-exists} returns the
7457 pathname. Here is a small example of its usage:
7461 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7464 @item @code{if-exists-else}
7465 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7466 spec function, except that it takes two arguments. The first argument is
7467 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7468 returns the pathname. If it does not exist, it returns the second argument.
7469 This way, @code{if-exists-else} can be used to select one file or another,
7470 based on the existence of the first. Here is a small example of its usage:
7474 crt0%O%s %:if-exists(crti%O%s) \
7475 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7478 @item @code{replace-outfile}
7479 The @code{replace-outfile} spec function takes two arguments. It looks for the
7480 first argument in the outfiles array and replaces it with the second argument. Here
7481 is a small example of its usage:
7484 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7490 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7491 If that switch was not specified, this substitutes nothing. Note that
7492 the leading dash is omitted when specifying this option, and it is
7493 automatically inserted if the substitution is performed. Thus the spec
7494 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7495 and would output the command line option @option{-foo}.
7497 @item %W@{@code{S}@}
7498 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7501 @item %@{@code{S}*@}
7502 Substitutes all the switches specified to GCC whose names start
7503 with @code{-S}, but which also take an argument. This is used for
7504 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7505 GCC considers @option{-o foo} as being
7506 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7507 text, including the space. Thus two arguments would be generated.
7509 @item %@{@code{S}*&@code{T}*@}
7510 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7511 (the order of @code{S} and @code{T} in the spec is not significant).
7512 There can be any number of ampersand-separated variables; for each the
7513 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7515 @item %@{@code{S}:@code{X}@}
7516 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7518 @item %@{!@code{S}:@code{X}@}
7519 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7521 @item %@{@code{S}*:@code{X}@}
7522 Substitutes @code{X} if one or more switches whose names start with
7523 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7524 once, no matter how many such switches appeared. However, if @code{%*}
7525 appears somewhere in @code{X}, then @code{X} will be substituted once
7526 for each matching switch, with the @code{%*} replaced by the part of
7527 that switch that matched the @code{*}.
7529 @item %@{.@code{S}:@code{X}@}
7530 Substitutes @code{X}, if processing a file with suffix @code{S}.
7532 @item %@{!.@code{S}:@code{X}@}
7533 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7535 @item %@{@code{S}|@code{P}:@code{X}@}
7536 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7537 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7538 although they have a stronger binding than the @samp{|}. If @code{%*}
7539 appears in @code{X}, all of the alternatives must be starred, and only
7540 the first matching alternative is substituted.
7542 For example, a spec string like this:
7545 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7548 will output the following command-line options from the following input
7549 command-line options:
7554 -d fred.c -foo -baz -boggle
7555 -d jim.d -bar -baz -boggle
7558 @item %@{S:X; T:Y; :D@}
7560 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7561 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7562 be as many clauses as you need. This may be combined with @code{.},
7563 @code{!}, @code{|}, and @code{*} as needed.
7568 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7569 construct may contain other nested @samp{%} constructs or spaces, or
7570 even newlines. They are processed as usual, as described above.
7571 Trailing white space in @code{X} is ignored. White space may also
7572 appear anywhere on the left side of the colon in these constructs,
7573 except between @code{.} or @code{*} and the corresponding word.
7575 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7576 handled specifically in these constructs. If another value of
7577 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7578 @option{-W} switch is found later in the command line, the earlier
7579 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7580 just one letter, which passes all matching options.
7582 The character @samp{|} at the beginning of the predicate text is used to
7583 indicate that a command should be piped to the following command, but
7584 only if @option{-pipe} is specified.
7586 It is built into GCC which switches take arguments and which do not.
7587 (You might think it would be useful to generalize this to allow each
7588 compiler's spec to say which switches take arguments. But this cannot
7589 be done in a consistent fashion. GCC cannot even decide which input
7590 files have been specified without knowing which switches take arguments,
7591 and it must know which input files to compile in order to tell which
7594 GCC also knows implicitly that arguments starting in @option{-l} are to be
7595 treated as compiler output files, and passed to the linker in their
7596 proper position among the other output files.
7598 @c man begin OPTIONS
7600 @node Target Options
7601 @section Specifying Target Machine and Compiler Version
7602 @cindex target options
7603 @cindex cross compiling
7604 @cindex specifying machine version
7605 @cindex specifying compiler version and target machine
7606 @cindex compiler version, specifying
7607 @cindex target machine, specifying
7609 The usual way to run GCC is to run the executable called @file{gcc}, or
7610 @file{<machine>-gcc} when cross-compiling, or
7611 @file{<machine>-gcc-<version>} to run a version other than the one that
7612 was installed last. Sometimes this is inconvenient, so GCC provides
7613 options that will switch to another cross-compiler or version.
7616 @item -b @var{machine}
7618 The argument @var{machine} specifies the target machine for compilation.
7620 The value to use for @var{machine} is the same as was specified as the
7621 machine type when configuring GCC as a cross-compiler. For
7622 example, if a cross-compiler was configured with @samp{configure
7623 arm-elf}, meaning to compile for an arm processor with elf binaries,
7624 then you would specify @option{-b arm-elf} to run that cross compiler.
7625 Because there are other options beginning with @option{-b}, the
7626 configuration must contain a hyphen.
7628 @item -V @var{version}
7630 The argument @var{version} specifies which version of GCC to run.
7631 This is useful when multiple versions are installed. For example,
7632 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7635 The @option{-V} and @option{-b} options work by running the
7636 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7637 use them if you can just run that directly.
7639 @node Submodel Options
7640 @section Hardware Models and Configurations
7641 @cindex submodel options
7642 @cindex specifying hardware config
7643 @cindex hardware models and configurations, specifying
7644 @cindex machine dependent options
7646 Earlier we discussed the standard option @option{-b} which chooses among
7647 different installed compilers for completely different target
7648 machines, such as VAX vs.@: 68000 vs.@: 80386.
7650 In addition, each of these target machine types can have its own
7651 special options, starting with @samp{-m}, to choose among various
7652 hardware models or configurations---for example, 68010 vs 68020,
7653 floating coprocessor or none. A single installed version of the
7654 compiler can compile for any model or configuration, according to the
7657 Some configurations of the compiler also support additional special
7658 options, usually for compatibility with other compilers on the same
7661 @c This list is ordered alphanumerically by subsection name.
7662 @c It should be the same order and spelling as these options are listed
7663 @c in Machine Dependent Options
7669 * Blackfin Options::
7673 * DEC Alpha Options::
7674 * DEC Alpha/VMS Options::
7676 * GNU/Linux Options::
7679 * i386 and x86-64 Options::
7692 * RS/6000 and PowerPC Options::
7693 * S/390 and zSeries Options::
7698 * System V Options::
7699 * TMS320C3x/C4x Options::
7703 * Xstormy16 Options::
7709 @subsection ARC Options
7712 These options are defined for ARC implementations:
7717 Compile code for little endian mode. This is the default.
7721 Compile code for big endian mode.
7724 @opindex mmangle-cpu
7725 Prepend the name of the cpu to all public symbol names.
7726 In multiple-processor systems, there are many ARC variants with different
7727 instruction and register set characteristics. This flag prevents code
7728 compiled for one cpu to be linked with code compiled for another.
7729 No facility exists for handling variants that are ``almost identical''.
7730 This is an all or nothing option.
7732 @item -mcpu=@var{cpu}
7734 Compile code for ARC variant @var{cpu}.
7735 Which variants are supported depend on the configuration.
7736 All variants support @option{-mcpu=base}, this is the default.
7738 @item -mtext=@var{text-section}
7739 @itemx -mdata=@var{data-section}
7740 @itemx -mrodata=@var{readonly-data-section}
7744 Put functions, data, and readonly data in @var{text-section},
7745 @var{data-section}, and @var{readonly-data-section} respectively
7746 by default. This can be overridden with the @code{section} attribute.
7747 @xref{Variable Attributes}.
7752 @subsection ARM Options
7755 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7759 @item -mabi=@var{name}
7761 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7762 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7765 @opindex mapcs-frame
7766 Generate a stack frame that is compliant with the ARM Procedure Call
7767 Standard for all functions, even if this is not strictly necessary for
7768 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7769 with this option will cause the stack frames not to be generated for
7770 leaf functions. The default is @option{-mno-apcs-frame}.
7774 This is a synonym for @option{-mapcs-frame}.
7777 @c not currently implemented
7778 @item -mapcs-stack-check
7779 @opindex mapcs-stack-check
7780 Generate code to check the amount of stack space available upon entry to
7781 every function (that actually uses some stack space). If there is
7782 insufficient space available then either the function
7783 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7784 called, depending upon the amount of stack space required. The run time
7785 system is required to provide these functions. The default is
7786 @option{-mno-apcs-stack-check}, since this produces smaller code.
7788 @c not currently implemented
7790 @opindex mapcs-float
7791 Pass floating point arguments using the float point registers. This is
7792 one of the variants of the APCS@. This option is recommended if the
7793 target hardware has a floating point unit or if a lot of floating point
7794 arithmetic is going to be performed by the code. The default is
7795 @option{-mno-apcs-float}, since integer only code is slightly increased in
7796 size if @option{-mapcs-float} is used.
7798 @c not currently implemented
7799 @item -mapcs-reentrant
7800 @opindex mapcs-reentrant
7801 Generate reentrant, position independent code. The default is
7802 @option{-mno-apcs-reentrant}.
7805 @item -mthumb-interwork
7806 @opindex mthumb-interwork
7807 Generate code which supports calling between the ARM and Thumb
7808 instruction sets. Without this option the two instruction sets cannot
7809 be reliably used inside one program. The default is
7810 @option{-mno-thumb-interwork}, since slightly larger code is generated
7811 when @option{-mthumb-interwork} is specified.
7813 @item -mno-sched-prolog
7814 @opindex mno-sched-prolog
7815 Prevent the reordering of instructions in the function prolog, or the
7816 merging of those instruction with the instructions in the function's
7817 body. This means that all functions will start with a recognizable set
7818 of instructions (or in fact one of a choice from a small set of
7819 different function prologues), and this information can be used to
7820 locate the start if functions inside an executable piece of code. The
7821 default is @option{-msched-prolog}.
7824 @opindex mhard-float
7825 Generate output containing floating point instructions. This is the
7829 @opindex msoft-float
7830 Generate output containing library calls for floating point.
7831 @strong{Warning:} the requisite libraries are not available for all ARM
7832 targets. Normally the facilities of the machine's usual C compiler are
7833 used, but this cannot be done directly in cross-compilation. You must make
7834 your own arrangements to provide suitable library functions for
7837 @option{-msoft-float} changes the calling convention in the output file;
7838 therefore, it is only useful if you compile @emph{all} of a program with
7839 this option. In particular, you need to compile @file{libgcc.a}, the
7840 library that comes with GCC, with @option{-msoft-float} in order for
7843 @item -mfloat-abi=@var{name}
7845 Specifies which ABI to use for floating point values. Permissible values
7846 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7848 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7849 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7850 of floating point instructions, but still uses the soft-float calling
7853 @item -mlittle-endian
7854 @opindex mlittle-endian
7855 Generate code for a processor running in little-endian mode. This is
7856 the default for all standard configurations.
7859 @opindex mbig-endian
7860 Generate code for a processor running in big-endian mode; the default is
7861 to compile code for a little-endian processor.
7863 @item -mwords-little-endian
7864 @opindex mwords-little-endian
7865 This option only applies when generating code for big-endian processors.
7866 Generate code for a little-endian word order but a big-endian byte
7867 order. That is, a byte order of the form @samp{32107654}. Note: this
7868 option should only be used if you require compatibility with code for
7869 big-endian ARM processors generated by versions of the compiler prior to
7872 @item -mcpu=@var{name}
7874 This specifies the name of the target ARM processor. GCC uses this name
7875 to determine what kind of instructions it can emit when generating
7876 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7877 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7878 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7879 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7880 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7881 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7882 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7883 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7884 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7885 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7886 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7887 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7888 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7889 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
7890 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
7891 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
7893 @itemx -mtune=@var{name}
7895 This option is very similar to the @option{-mcpu=} option, except that
7896 instead of specifying the actual target processor type, and hence
7897 restricting which instructions can be used, it specifies that GCC should
7898 tune the performance of the code as if the target were of the type
7899 specified in this option, but still choosing the instructions that it
7900 will generate based on the cpu specified by a @option{-mcpu=} option.
7901 For some ARM implementations better performance can be obtained by using
7904 @item -march=@var{name}
7906 This specifies the name of the target ARM architecture. GCC uses this
7907 name to determine what kind of instructions it can emit when generating
7908 assembly code. This option can be used in conjunction with or instead
7909 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7910 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7911 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7912 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
7913 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
7915 @item -mfpu=@var{name}
7916 @itemx -mfpe=@var{number}
7917 @itemx -mfp=@var{number}
7921 This specifies what floating point hardware (or hardware emulation) is
7922 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7923 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7924 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7925 with older versions of GCC@.
7927 If @option{-msoft-float} is specified this specifies the format of
7928 floating point values.
7930 @item -mstructure-size-boundary=@var{n}
7931 @opindex mstructure-size-boundary
7932 The size of all structures and unions will be rounded up to a multiple
7933 of the number of bits set by this option. Permissible values are 8, 32
7934 and 64. The default value varies for different toolchains. For the COFF
7935 targeted toolchain the default value is 8. A value of 64 is only allowed
7936 if the underlying ABI supports it.
7938 Specifying the larger number can produce faster, more efficient code, but
7939 can also increase the size of the program. Different values are potentially
7940 incompatible. Code compiled with one value cannot necessarily expect to
7941 work with code or libraries compiled with another value, if they exchange
7942 information using structures or unions.
7944 @item -mabort-on-noreturn
7945 @opindex mabort-on-noreturn
7946 Generate a call to the function @code{abort} at the end of a
7947 @code{noreturn} function. It will be executed if the function tries to
7951 @itemx -mno-long-calls
7952 @opindex mlong-calls
7953 @opindex mno-long-calls
7954 Tells the compiler to perform function calls by first loading the
7955 address of the function into a register and then performing a subroutine
7956 call on this register. This switch is needed if the target function
7957 will lie outside of the 64 megabyte addressing range of the offset based
7958 version of subroutine call instruction.
7960 Even if this switch is enabled, not all function calls will be turned
7961 into long calls. The heuristic is that static functions, functions
7962 which have the @samp{short-call} attribute, functions that are inside
7963 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7964 definitions have already been compiled within the current compilation
7965 unit, will not be turned into long calls. The exception to this rule is
7966 that weak function definitions, functions with the @samp{long-call}
7967 attribute or the @samp{section} attribute, and functions that are within
7968 the scope of a @samp{#pragma long_calls} directive, will always be
7969 turned into long calls.
7971 This feature is not enabled by default. Specifying
7972 @option{-mno-long-calls} will restore the default behavior, as will
7973 placing the function calls within the scope of a @samp{#pragma
7974 long_calls_off} directive. Note these switches have no effect on how
7975 the compiler generates code to handle function calls via function
7978 @item -mnop-fun-dllimport
7979 @opindex mnop-fun-dllimport
7980 Disable support for the @code{dllimport} attribute.
7982 @item -msingle-pic-base
7983 @opindex msingle-pic-base
7984 Treat the register used for PIC addressing as read-only, rather than
7985 loading it in the prologue for each function. The run-time system is
7986 responsible for initializing this register with an appropriate value
7987 before execution begins.
7989 @item -mpic-register=@var{reg}
7990 @opindex mpic-register
7991 Specify the register to be used for PIC addressing. The default is R10
7992 unless stack-checking is enabled, when R9 is used.
7994 @item -mcirrus-fix-invalid-insns
7995 @opindex mcirrus-fix-invalid-insns
7996 @opindex mno-cirrus-fix-invalid-insns
7997 Insert NOPs into the instruction stream to in order to work around
7998 problems with invalid Maverick instruction combinations. This option
7999 is only valid if the @option{-mcpu=ep9312} option has been used to
8000 enable generation of instructions for the Cirrus Maverick floating
8001 point co-processor. This option is not enabled by default, since the
8002 problem is only present in older Maverick implementations. The default
8003 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8006 @item -mpoke-function-name
8007 @opindex mpoke-function-name
8008 Write the name of each function into the text section, directly
8009 preceding the function prologue. The generated code is similar to this:
8013 .ascii "arm_poke_function_name", 0
8016 .word 0xff000000 + (t1 - t0)
8017 arm_poke_function_name
8019 stmfd sp!, @{fp, ip, lr, pc@}
8023 When performing a stack backtrace, code can inspect the value of
8024 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8025 location @code{pc - 12} and the top 8 bits are set, then we know that
8026 there is a function name embedded immediately preceding this location
8027 and has length @code{((pc[-3]) & 0xff000000)}.
8031 Generate code for the Thumb instruction set. The default is to
8032 use the 32-bit ARM instruction set.
8033 This option automatically enables either 16-bit Thumb-1 or
8034 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8035 and @option{-march=@var{name}} options.
8038 @opindex mtpcs-frame
8039 Generate a stack frame that is compliant with the Thumb Procedure Call
8040 Standard for all non-leaf functions. (A leaf function is one that does
8041 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8043 @item -mtpcs-leaf-frame
8044 @opindex mtpcs-leaf-frame
8045 Generate a stack frame that is compliant with the Thumb Procedure Call
8046 Standard for all leaf functions. (A leaf function is one that does
8047 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8049 @item -mcallee-super-interworking
8050 @opindex mcallee-super-interworking
8051 Gives all externally visible functions in the file being compiled an ARM
8052 instruction set header which switches to Thumb mode before executing the
8053 rest of the function. This allows these functions to be called from
8054 non-interworking code.
8056 @item -mcaller-super-interworking
8057 @opindex mcaller-super-interworking
8058 Allows calls via function pointers (including virtual functions) to
8059 execute correctly regardless of whether the target code has been
8060 compiled for interworking or not. There is a small overhead in the cost
8061 of executing a function pointer if this option is enabled.
8063 @item -mtp=@var{name}
8065 Specify the access model for the thread local storage pointer. The valid
8066 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8067 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8068 (supported in the arm6k architecture), and @option{auto}, which uses the
8069 best available method for the selected processor. The default setting is
8075 @subsection AVR Options
8078 These options are defined for AVR implementations:
8081 @item -mmcu=@var{mcu}
8083 Specify ATMEL AVR instruction set or MCU type.
8085 Instruction set avr1 is for the minimal AVR core, not supported by the C
8086 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8087 attiny11, attiny12, attiny15, attiny28).
8089 Instruction set avr2 (default) is for the classic AVR core with up to
8090 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8091 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8092 at90c8534, at90s8535).
8094 Instruction set avr3 is for the classic AVR core with up to 128K program
8095 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8097 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8098 memory space (MCU types: atmega8, atmega83, atmega85).
8100 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8101 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8102 atmega64, atmega128, at43usb355, at94k).
8106 Output instruction sizes to the asm file.
8108 @item -minit-stack=@var{N}
8109 @opindex minit-stack
8110 Specify the initial stack address, which may be a symbol or numeric value,
8111 @samp{__stack} is the default.
8113 @item -mno-interrupts
8114 @opindex mno-interrupts
8115 Generated code is not compatible with hardware interrupts.
8116 Code size will be smaller.
8118 @item -mcall-prologues
8119 @opindex mcall-prologues
8120 Functions prologues/epilogues expanded as call to appropriate
8121 subroutines. Code size will be smaller.
8123 @item -mno-tablejump
8124 @opindex mno-tablejump
8125 Do not generate tablejump insns which sometimes increase code size.
8128 @opindex mtiny-stack
8129 Change only the low 8 bits of the stack pointer.
8133 Assume int to be 8 bit integer. This affects the sizes of all types: A
8134 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8135 and long long will be 4 bytes. Please note that this option does not
8136 comply to the C standards, but it will provide you with smaller code
8140 @node Blackfin Options
8141 @subsection Blackfin Options
8142 @cindex Blackfin Options
8145 @item -momit-leaf-frame-pointer
8146 @opindex momit-leaf-frame-pointer
8147 Don't keep the frame pointer in a register for leaf functions. This
8148 avoids the instructions to save, set up and restore frame pointers and
8149 makes an extra register available in leaf functions. The option
8150 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8151 which might make debugging harder.
8153 @item -mspecld-anomaly
8154 @opindex mspecld-anomaly
8155 When enabled, the compiler will ensure that the generated code does not
8156 contain speculative loads after jump instructions. This option is enabled
8159 @item -mno-specld-anomaly
8160 @opindex mno-specld-anomaly
8161 Don't generate extra code to prevent speculative loads from occurring.
8163 @item -mcsync-anomaly
8164 @opindex mcsync-anomaly
8165 When enabled, the compiler will ensure that the generated code does not
8166 contain CSYNC or SSYNC instructions too soon after conditional branches.
8167 This option is enabled by default.
8169 @item -mno-csync-anomaly
8170 @opindex mno-csync-anomaly
8171 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8172 occurring too soon after a conditional branch.
8176 When enabled, the compiler is free to take advantage of the knowledge that
8177 the entire program fits into the low 64k of memory.
8180 @opindex mno-low-64k
8181 Assume that the program is arbitrarily large. This is the default.
8183 @item -mstack-check-l1
8184 @opindex mstack-check-l1
8185 Do stack checking using information placed into L1 scratchpad memory by the
8188 @item -mid-shared-library
8189 @opindex mid-shared-library
8190 Generate code that supports shared libraries via the library ID method.
8191 This allows for execute in place and shared libraries in an environment
8192 without virtual memory management. This option implies @option{-fPIC}.
8194 @item -mno-id-shared-library
8195 @opindex mno-id-shared-library
8196 Generate code that doesn't assume ID based shared libraries are being used.
8197 This is the default.
8199 @item -mleaf-id-shared-library
8200 @opindex mleaf-id-shared-library
8201 Generate code that supports shared libraries via the library ID method,
8202 but assumes that this library or executable won't link against any other
8203 ID shared libraries. That allows the compiler to use faster code for jumps
8206 @item -mno-leaf-id-shared-library
8207 @opindex mno-leaf-id-shared-library
8208 Do not assume that the code being compiled won't link against any ID shared
8209 libraries. Slower code will be generated for jump and call insns.
8211 @item -mshared-library-id=n
8212 @opindex mshared-library-id
8213 Specified the identification number of the ID based shared library being
8214 compiled. Specifying a value of 0 will generate more compact code, specifying
8215 other values will force the allocation of that number to the current
8216 library but is no more space or time efficient than omitting this option.
8220 Generate code that allows the data segment to be located in a different
8221 area of memory from the text segment. This allows for execute in place in
8222 an environment without virtual memory management by eliminating relocations
8223 against the text section.
8226 @opindex mno-sep-data
8227 Generate code that assumes that the data segment follows the text segment.
8228 This is the default.
8231 @itemx -mno-long-calls
8232 @opindex mlong-calls
8233 @opindex mno-long-calls
8234 Tells the compiler to perform function calls by first loading the
8235 address of the function into a register and then performing a subroutine
8236 call on this register. This switch is needed if the target function
8237 will lie outside of the 24 bit addressing range of the offset based
8238 version of subroutine call instruction.
8240 This feature is not enabled by default. Specifying
8241 @option{-mno-long-calls} will restore the default behavior. Note these
8242 switches have no effect on how the compiler generates code to handle
8243 function calls via function pointers.
8247 @subsection CRIS Options
8248 @cindex CRIS Options
8250 These options are defined specifically for the CRIS ports.
8253 @item -march=@var{architecture-type}
8254 @itemx -mcpu=@var{architecture-type}
8257 Generate code for the specified architecture. The choices for
8258 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8259 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8260 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8263 @item -mtune=@var{architecture-type}
8265 Tune to @var{architecture-type} everything applicable about the generated
8266 code, except for the ABI and the set of available instructions. The
8267 choices for @var{architecture-type} are the same as for
8268 @option{-march=@var{architecture-type}}.
8270 @item -mmax-stack-frame=@var{n}
8271 @opindex mmax-stack-frame
8272 Warn when the stack frame of a function exceeds @var{n} bytes.
8274 @item -melinux-stacksize=@var{n}
8275 @opindex melinux-stacksize
8276 Only available with the @samp{cris-axis-aout} target. Arranges for
8277 indications in the program to the kernel loader that the stack of the
8278 program should be set to @var{n} bytes.
8284 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8285 @option{-march=v3} and @option{-march=v8} respectively.
8287 @item -mmul-bug-workaround
8288 @itemx -mno-mul-bug-workaround
8289 @opindex mmul-bug-workaround
8290 @opindex mno-mul-bug-workaround
8291 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8292 models where it applies. This option is active by default.
8296 Enable CRIS-specific verbose debug-related information in the assembly
8297 code. This option also has the effect to turn off the @samp{#NO_APP}
8298 formatted-code indicator to the assembler at the beginning of the
8303 Do not use condition-code results from previous instruction; always emit
8304 compare and test instructions before use of condition codes.
8306 @item -mno-side-effects
8307 @opindex mno-side-effects
8308 Do not emit instructions with side-effects in addressing modes other than
8312 @itemx -mno-stack-align
8314 @itemx -mno-data-align
8315 @itemx -mconst-align
8316 @itemx -mno-const-align
8317 @opindex mstack-align
8318 @opindex mno-stack-align
8319 @opindex mdata-align
8320 @opindex mno-data-align
8321 @opindex mconst-align
8322 @opindex mno-const-align
8323 These options (no-options) arranges (eliminate arrangements) for the
8324 stack-frame, individual data and constants to be aligned for the maximum
8325 single data access size for the chosen CPU model. The default is to
8326 arrange for 32-bit alignment. ABI details such as structure layout are
8327 not affected by these options.
8335 Similar to the stack- data- and const-align options above, these options
8336 arrange for stack-frame, writable data and constants to all be 32-bit,
8337 16-bit or 8-bit aligned. The default is 32-bit alignment.
8339 @item -mno-prologue-epilogue
8340 @itemx -mprologue-epilogue
8341 @opindex mno-prologue-epilogue
8342 @opindex mprologue-epilogue
8343 With @option{-mno-prologue-epilogue}, the normal function prologue and
8344 epilogue that sets up the stack-frame are omitted and no return
8345 instructions or return sequences are generated in the code. Use this
8346 option only together with visual inspection of the compiled code: no
8347 warnings or errors are generated when call-saved registers must be saved,
8348 or storage for local variable needs to be allocated.
8354 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8355 instruction sequences that load addresses for functions from the PLT part
8356 of the GOT rather than (traditional on other architectures) calls to the
8357 PLT@. The default is @option{-mgotplt}.
8361 Legacy no-op option only recognized with the cris-axis-aout target.
8365 Legacy no-op option only recognized with the cris-axis-elf and
8366 cris-axis-linux-gnu targets.
8370 Only recognized with the cris-axis-aout target, where it selects a
8371 GNU/linux-like multilib, include files and instruction set for
8376 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8380 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8381 to link with input-output functions from a simulator library. Code,
8382 initialized data and zero-initialized data are allocated consecutively.
8386 Like @option{-sim}, but pass linker options to locate initialized data at
8387 0x40000000 and zero-initialized data at 0x80000000.
8391 @subsection CRX Options
8394 These options are defined specifically for the CRX ports.
8400 Enable the use of multiply-accumulate instructions. Disabled by default.
8404 Push instructions will be used to pass outgoing arguments when functions
8405 are called. Enabled by default.
8408 @node Darwin Options
8409 @subsection Darwin Options
8410 @cindex Darwin options
8412 These options are defined for all architectures running the Darwin operating
8415 FSF GCC on Darwin does not create ``fat'' object files; it will create
8416 an object file for the single architecture that it was built to
8417 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8418 @option{-arch} options are used; it does so by running the compiler or
8419 linker multiple times and joining the results together with
8422 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8423 @samp{i686}) is determined by the flags that specify the ISA
8424 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8425 @option{-force_cpusubtype_ALL} option can be used to override this.
8427 The Darwin tools vary in their behavior when presented with an ISA
8428 mismatch. The assembler, @file{as}, will only permit instructions to
8429 be used that are valid for the subtype of the file it is generating,
8430 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8431 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8432 and print an error if asked to create a shared library with a less
8433 restrictive subtype than its input files (for instance, trying to put
8434 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8435 for executables, @file{ld}, will quietly give the executable the most
8436 restrictive subtype of any of its input files.
8441 Add the framework directory @var{dir} to the head of the list of
8442 directories to be searched for header files. These directories are
8443 interleaved with those specified by @option{-I} options and are
8444 scanned in a left-to-right order.
8446 A framework directory is a directory with frameworks in it. A
8447 framework is a directory with a @samp{"Headers"} and/or
8448 @samp{"PrivateHeaders"} directory contained directly in it that ends
8449 in @samp{".framework"}. The name of a framework is the name of this
8450 directory excluding the @samp{".framework"}. Headers associated with
8451 the framework are found in one of those two directories, with
8452 @samp{"Headers"} being searched first. A subframework is a framework
8453 directory that is in a framework's @samp{"Frameworks"} directory.
8454 Includes of subframework headers can only appear in a header of a
8455 framework that contains the subframework, or in a sibling subframework
8456 header. Two subframeworks are siblings if they occur in the same
8457 framework. A subframework should not have the same name as a
8458 framework, a warning will be issued if this is violated. Currently a
8459 subframework cannot have subframeworks, in the future, the mechanism
8460 may be extended to support this. The standard frameworks can be found
8461 in @samp{"/System/Library/Frameworks"} and
8462 @samp{"/Library/Frameworks"}. An example include looks like
8463 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8464 the name of the framework and header.h is found in the
8465 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8467 @item -iframework@var{dir}
8469 Like @option{-F} except the directory is a treated as a system
8470 directory. The main difference between this @option{-iframework} and
8471 @option{-F} is that with @option{-iframework} the compiler does not
8472 warn about constructs contained within header files found via
8473 @var{dir}. This option is valid only for the C family of languages.
8477 Emit debugging information for symbols that are used. For STABS
8478 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8479 This is by default ON@.
8483 Emit debugging information for all symbols and types.
8485 @item -mmacosx-version-min=@var{version}
8486 The earliest version of MacOS X that this executable will run on
8487 is @var{version}. Typical values of @var{version} include @code{10.1},
8488 @code{10.2}, and @code{10.3.9}.
8490 The default for this option is to make choices that seem to be most
8495 Enable kernel development mode. The @option{-mkernel} option sets
8496 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8497 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8498 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8499 applicable. This mode also sets @option{-mno-altivec},
8500 @option{-msoft-float}, @option{-fno-builtin} and
8501 @option{-mlong-branch} for PowerPC targets.
8503 @item -mone-byte-bool
8504 @opindex -mone-byte-bool
8505 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8506 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8507 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8508 option has no effect on x86.
8510 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8511 to generate code that is not binary compatible with code generated
8512 without that switch. Using this switch may require recompiling all
8513 other modules in a program, including system libraries. Use this
8514 switch to conform to a non-default data model.
8516 @item -mfix-and-continue
8517 @itemx -ffix-and-continue
8518 @itemx -findirect-data
8519 @opindex mfix-and-continue
8520 @opindex ffix-and-continue
8521 @opindex findirect-data
8522 Generate code suitable for fast turn around development. Needed to
8523 enable gdb to dynamically load @code{.o} files into already running
8524 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8525 are provided for backwards compatibility.
8529 Loads all members of static archive libraries.
8530 See man ld(1) for more information.
8532 @item -arch_errors_fatal
8533 @opindex arch_errors_fatal
8534 Cause the errors having to do with files that have the wrong architecture
8538 @opindex bind_at_load
8539 Causes the output file to be marked such that the dynamic linker will
8540 bind all undefined references when the file is loaded or launched.
8544 Produce a Mach-o bundle format file.
8545 See man ld(1) for more information.
8547 @item -bundle_loader @var{executable}
8548 @opindex bundle_loader
8549 This option specifies the @var{executable} that will be loading the build
8550 output file being linked. See man ld(1) for more information.
8553 @opindex -dynamiclib
8554 When passed this option, GCC will produce a dynamic library instead of
8555 an executable when linking, using the Darwin @file{libtool} command.
8557 @item -force_cpusubtype_ALL
8558 @opindex -force_cpusubtype_ALL
8559 This causes GCC's output file to have the @var{ALL} subtype, instead of
8560 one controlled by the @option{-mcpu} or @option{-march} option.
8562 @item -allowable_client @var{client_name}
8564 @itemx -compatibility_version
8565 @itemx -current_version
8567 @itemx -dependency-file
8569 @itemx -dylinker_install_name
8571 @itemx -exported_symbols_list
8573 @itemx -flat_namespace
8574 @itemx -force_flat_namespace
8575 @itemx -headerpad_max_install_names
8578 @itemx -install_name
8579 @itemx -keep_private_externs
8580 @itemx -multi_module
8581 @itemx -multiply_defined
8582 @itemx -multiply_defined_unused
8584 @itemx -no_dead_strip_inits_and_terms
8585 @itemx -nofixprebinding
8588 @itemx -noseglinkedit
8589 @itemx -pagezero_size
8591 @itemx -prebind_all_twolevel_modules
8592 @itemx -private_bundle
8593 @itemx -read_only_relocs
8595 @itemx -sectobjectsymbols
8599 @itemx -sectobjectsymbols
8602 @itemx -segs_read_only_addr
8603 @itemx -segs_read_write_addr
8604 @itemx -seg_addr_table
8605 @itemx -seg_addr_table_filename
8608 @itemx -segs_read_only_addr
8609 @itemx -segs_read_write_addr
8610 @itemx -single_module
8613 @itemx -sub_umbrella
8614 @itemx -twolevel_namespace
8617 @itemx -unexported_symbols_list
8618 @itemx -weak_reference_mismatches
8621 @opindex allowable_client
8622 @opindex client_name
8623 @opindex compatibility_version
8624 @opindex current_version
8626 @opindex dependency-file
8628 @opindex dylinker_install_name
8630 @opindex exported_symbols_list
8632 @opindex flat_namespace
8633 @opindex force_flat_namespace
8634 @opindex headerpad_max_install_names
8637 @opindex install_name
8638 @opindex keep_private_externs
8639 @opindex multi_module
8640 @opindex multiply_defined
8641 @opindex multiply_defined_unused
8643 @opindex no_dead_strip_inits_and_terms
8644 @opindex nofixprebinding
8645 @opindex nomultidefs
8647 @opindex noseglinkedit
8648 @opindex pagezero_size
8650 @opindex prebind_all_twolevel_modules
8651 @opindex private_bundle
8652 @opindex read_only_relocs
8654 @opindex sectobjectsymbols
8658 @opindex sectobjectsymbols
8661 @opindex segs_read_only_addr
8662 @opindex segs_read_write_addr
8663 @opindex seg_addr_table
8664 @opindex seg_addr_table_filename
8665 @opindex seglinkedit
8667 @opindex segs_read_only_addr
8668 @opindex segs_read_write_addr
8669 @opindex single_module
8671 @opindex sub_library
8672 @opindex sub_umbrella
8673 @opindex twolevel_namespace
8676 @opindex unexported_symbols_list
8677 @opindex weak_reference_mismatches
8678 @opindex whatsloaded
8680 These options are passed to the Darwin linker. The Darwin linker man page
8681 describes them in detail.
8684 @node DEC Alpha Options
8685 @subsection DEC Alpha Options
8687 These @samp{-m} options are defined for the DEC Alpha implementations:
8690 @item -mno-soft-float
8692 @opindex mno-soft-float
8693 @opindex msoft-float
8694 Use (do not use) the hardware floating-point instructions for
8695 floating-point operations. When @option{-msoft-float} is specified,
8696 functions in @file{libgcc.a} will be used to perform floating-point
8697 operations. Unless they are replaced by routines that emulate the
8698 floating-point operations, or compiled in such a way as to call such
8699 emulations routines, these routines will issue floating-point
8700 operations. If you are compiling for an Alpha without floating-point
8701 operations, you must ensure that the library is built so as not to call
8704 Note that Alpha implementations without floating-point operations are
8705 required to have floating-point registers.
8710 @opindex mno-fp-regs
8711 Generate code that uses (does not use) the floating-point register set.
8712 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8713 register set is not used, floating point operands are passed in integer
8714 registers as if they were integers and floating-point results are passed
8715 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8716 so any function with a floating-point argument or return value called by code
8717 compiled with @option{-mno-fp-regs} must also be compiled with that
8720 A typical use of this option is building a kernel that does not use,
8721 and hence need not save and restore, any floating-point registers.
8725 The Alpha architecture implements floating-point hardware optimized for
8726 maximum performance. It is mostly compliant with the IEEE floating
8727 point standard. However, for full compliance, software assistance is
8728 required. This option generates code fully IEEE compliant code
8729 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8730 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8731 defined during compilation. The resulting code is less efficient but is
8732 able to correctly support denormalized numbers and exceptional IEEE
8733 values such as not-a-number and plus/minus infinity. Other Alpha
8734 compilers call this option @option{-ieee_with_no_inexact}.
8736 @item -mieee-with-inexact
8737 @opindex mieee-with-inexact
8738 This is like @option{-mieee} except the generated code also maintains
8739 the IEEE @var{inexact-flag}. Turning on this option causes the
8740 generated code to implement fully-compliant IEEE math. In addition to
8741 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8742 macro. On some Alpha implementations the resulting code may execute
8743 significantly slower than the code generated by default. Since there is
8744 very little code that depends on the @var{inexact-flag}, you should
8745 normally not specify this option. Other Alpha compilers call this
8746 option @option{-ieee_with_inexact}.
8748 @item -mfp-trap-mode=@var{trap-mode}
8749 @opindex mfp-trap-mode
8750 This option controls what floating-point related traps are enabled.
8751 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8752 The trap mode can be set to one of four values:
8756 This is the default (normal) setting. The only traps that are enabled
8757 are the ones that cannot be disabled in software (e.g., division by zero
8761 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8765 Like @samp{u}, but the instructions are marked to be safe for software
8766 completion (see Alpha architecture manual for details).
8769 Like @samp{su}, but inexact traps are enabled as well.
8772 @item -mfp-rounding-mode=@var{rounding-mode}
8773 @opindex mfp-rounding-mode
8774 Selects the IEEE rounding mode. Other Alpha compilers call this option
8775 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8780 Normal IEEE rounding mode. Floating point numbers are rounded towards
8781 the nearest machine number or towards the even machine number in case
8785 Round towards minus infinity.
8788 Chopped rounding mode. Floating point numbers are rounded towards zero.
8791 Dynamic rounding mode. A field in the floating point control register
8792 (@var{fpcr}, see Alpha architecture reference manual) controls the
8793 rounding mode in effect. The C library initializes this register for
8794 rounding towards plus infinity. Thus, unless your program modifies the
8795 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8798 @item -mtrap-precision=@var{trap-precision}
8799 @opindex mtrap-precision
8800 In the Alpha architecture, floating point traps are imprecise. This
8801 means without software assistance it is impossible to recover from a
8802 floating trap and program execution normally needs to be terminated.
8803 GCC can generate code that can assist operating system trap handlers
8804 in determining the exact location that caused a floating point trap.
8805 Depending on the requirements of an application, different levels of
8806 precisions can be selected:
8810 Program precision. This option is the default and means a trap handler
8811 can only identify which program caused a floating point exception.
8814 Function precision. The trap handler can determine the function that
8815 caused a floating point exception.
8818 Instruction precision. The trap handler can determine the exact
8819 instruction that caused a floating point exception.
8822 Other Alpha compilers provide the equivalent options called
8823 @option{-scope_safe} and @option{-resumption_safe}.
8825 @item -mieee-conformant
8826 @opindex mieee-conformant
8827 This option marks the generated code as IEEE conformant. You must not
8828 use this option unless you also specify @option{-mtrap-precision=i} and either
8829 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8830 is to emit the line @samp{.eflag 48} in the function prologue of the
8831 generated assembly file. Under DEC Unix, this has the effect that
8832 IEEE-conformant math library routines will be linked in.
8834 @item -mbuild-constants
8835 @opindex mbuild-constants
8836 Normally GCC examines a 32- or 64-bit integer constant to
8837 see if it can construct it from smaller constants in two or three
8838 instructions. If it cannot, it will output the constant as a literal and
8839 generate code to load it from the data segment at runtime.
8841 Use this option to require GCC to construct @emph{all} integer constants
8842 using code, even if it takes more instructions (the maximum is six).
8844 You would typically use this option to build a shared library dynamic
8845 loader. Itself a shared library, it must relocate itself in memory
8846 before it can find the variables and constants in its own data segment.
8852 Select whether to generate code to be assembled by the vendor-supplied
8853 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8871 Indicate whether GCC should generate code to use the optional BWX,
8872 CIX, FIX and MAX instruction sets. The default is to use the instruction
8873 sets supported by the CPU type specified via @option{-mcpu=} option or that
8874 of the CPU on which GCC was built if none was specified.
8879 @opindex mfloat-ieee
8880 Generate code that uses (does not use) VAX F and G floating point
8881 arithmetic instead of IEEE single and double precision.
8883 @item -mexplicit-relocs
8884 @itemx -mno-explicit-relocs
8885 @opindex mexplicit-relocs
8886 @opindex mno-explicit-relocs
8887 Older Alpha assemblers provided no way to generate symbol relocations
8888 except via assembler macros. Use of these macros does not allow
8889 optimal instruction scheduling. GNU binutils as of version 2.12
8890 supports a new syntax that allows the compiler to explicitly mark
8891 which relocations should apply to which instructions. This option
8892 is mostly useful for debugging, as GCC detects the capabilities of
8893 the assembler when it is built and sets the default accordingly.
8897 @opindex msmall-data
8898 @opindex mlarge-data
8899 When @option{-mexplicit-relocs} is in effect, static data is
8900 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8901 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8902 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8903 16-bit relocations off of the @code{$gp} register. This limits the
8904 size of the small data area to 64KB, but allows the variables to be
8905 directly accessed via a single instruction.
8907 The default is @option{-mlarge-data}. With this option the data area
8908 is limited to just below 2GB@. Programs that require more than 2GB of
8909 data must use @code{malloc} or @code{mmap} to allocate the data in the
8910 heap instead of in the program's data segment.
8912 When generating code for shared libraries, @option{-fpic} implies
8913 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8917 @opindex msmall-text
8918 @opindex mlarge-text
8919 When @option{-msmall-text} is used, the compiler assumes that the
8920 code of the entire program (or shared library) fits in 4MB, and is
8921 thus reachable with a branch instruction. When @option{-msmall-data}
8922 is used, the compiler can assume that all local symbols share the
8923 same @code{$gp} value, and thus reduce the number of instructions
8924 required for a function call from 4 to 1.
8926 The default is @option{-mlarge-text}.
8928 @item -mcpu=@var{cpu_type}
8930 Set the instruction set and instruction scheduling parameters for
8931 machine type @var{cpu_type}. You can specify either the @samp{EV}
8932 style name or the corresponding chip number. GCC supports scheduling
8933 parameters for the EV4, EV5 and EV6 family of processors and will
8934 choose the default values for the instruction set from the processor
8935 you specify. If you do not specify a processor type, GCC will default
8936 to the processor on which the compiler was built.
8938 Supported values for @var{cpu_type} are
8944 Schedules as an EV4 and has no instruction set extensions.
8948 Schedules as an EV5 and has no instruction set extensions.
8952 Schedules as an EV5 and supports the BWX extension.
8957 Schedules as an EV5 and supports the BWX and MAX extensions.
8961 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8965 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8968 @item -mtune=@var{cpu_type}
8970 Set only the instruction scheduling parameters for machine type
8971 @var{cpu_type}. The instruction set is not changed.
8973 @item -mmemory-latency=@var{time}
8974 @opindex mmemory-latency
8975 Sets the latency the scheduler should assume for typical memory
8976 references as seen by the application. This number is highly
8977 dependent on the memory access patterns used by the application
8978 and the size of the external cache on the machine.
8980 Valid options for @var{time} are
8984 A decimal number representing clock cycles.
8990 The compiler contains estimates of the number of clock cycles for
8991 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8992 (also called Dcache, Scache, and Bcache), as well as to main memory.
8993 Note that L3 is only valid for EV5.
8998 @node DEC Alpha/VMS Options
8999 @subsection DEC Alpha/VMS Options
9001 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9004 @item -mvms-return-codes
9005 @opindex mvms-return-codes
9006 Return VMS condition codes from main. The default is to return POSIX
9007 style condition (e.g.@ error) codes.
9011 @subsection FRV Options
9018 Only use the first 32 general purpose registers.
9023 Use all 64 general purpose registers.
9028 Use only the first 32 floating point registers.
9033 Use all 64 floating point registers
9036 @opindex mhard-float
9038 Use hardware instructions for floating point operations.
9041 @opindex msoft-float
9043 Use library routines for floating point operations.
9048 Dynamically allocate condition code registers.
9053 Do not try to dynamically allocate condition code registers, only
9054 use @code{icc0} and @code{fcc0}.
9059 Change ABI to use double word insns.
9064 Do not use double word instructions.
9069 Use floating point double instructions.
9074 Do not use floating point double instructions.
9079 Use media instructions.
9084 Do not use media instructions.
9089 Use multiply and add/subtract instructions.
9094 Do not use multiply and add/subtract instructions.
9099 Select the FDPIC ABI, that uses function descriptors to represent
9100 pointers to functions. Without any PIC/PIE-related options, it
9101 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9102 assumes GOT entries and small data are within a 12-bit range from the
9103 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9104 are computed with 32 bits.
9107 @opindex minline-plt
9109 Enable inlining of PLT entries in function calls to functions that are
9110 not known to bind locally. It has no effect without @option{-mfdpic}.
9111 It's enabled by default if optimizing for speed and compiling for
9112 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9113 optimization option such as @option{-O3} or above is present in the
9119 Assume a large TLS segment when generating thread-local code.
9124 Do not assume a large TLS segment when generating thread-local code.
9129 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9130 that is known to be in read-only sections. It's enabled by default,
9131 except for @option{-fpic} or @option{-fpie}: even though it may help
9132 make the global offset table smaller, it trades 1 instruction for 4.
9133 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9134 one of which may be shared by multiple symbols, and it avoids the need
9135 for a GOT entry for the referenced symbol, so it's more likely to be a
9136 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9138 @item -multilib-library-pic
9139 @opindex multilib-library-pic
9141 Link with the (library, not FD) pic libraries. It's implied by
9142 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9143 @option{-fpic} without @option{-mfdpic}. You should never have to use
9149 Follow the EABI requirement of always creating a frame pointer whenever
9150 a stack frame is allocated. This option is enabled by default and can
9151 be disabled with @option{-mno-linked-fp}.
9154 @opindex mlong-calls
9156 Use indirect addressing to call functions outside the current
9157 compilation unit. This allows the functions to be placed anywhere
9158 within the 32-bit address space.
9160 @item -malign-labels
9161 @opindex malign-labels
9163 Try to align labels to an 8-byte boundary by inserting nops into the
9164 previous packet. This option only has an effect when VLIW packing
9165 is enabled. It doesn't create new packets; it merely adds nops to
9169 @opindex mlibrary-pic
9171 Generate position-independent EABI code.
9176 Use only the first four media accumulator registers.
9181 Use all eight media accumulator registers.
9186 Pack VLIW instructions.
9191 Do not pack VLIW instructions.
9196 Do not mark ABI switches in e_flags.
9201 Enable the use of conditional-move instructions (default).
9203 This switch is mainly for debugging the compiler and will likely be removed
9204 in a future version.
9206 @item -mno-cond-move
9207 @opindex mno-cond-move
9209 Disable the use of conditional-move instructions.
9211 This switch is mainly for debugging the compiler and will likely be removed
9212 in a future version.
9217 Enable the use of conditional set instructions (default).
9219 This switch is mainly for debugging the compiler and will likely be removed
9220 in a future version.
9225 Disable the use of conditional set instructions.
9227 This switch is mainly for debugging the compiler and will likely be removed
9228 in a future version.
9233 Enable the use of conditional execution (default).
9235 This switch is mainly for debugging the compiler and will likely be removed
9236 in a future version.
9238 @item -mno-cond-exec
9239 @opindex mno-cond-exec
9241 Disable the use of conditional execution.
9243 This switch is mainly for debugging the compiler and will likely be removed
9244 in a future version.
9247 @opindex mvliw-branch
9249 Run a pass to pack branches into VLIW instructions (default).
9251 This switch is mainly for debugging the compiler and will likely be removed
9252 in a future version.
9254 @item -mno-vliw-branch
9255 @opindex mno-vliw-branch
9257 Do not run a pass to pack branches into VLIW instructions.
9259 This switch is mainly for debugging the compiler and will likely be removed
9260 in a future version.
9262 @item -mmulti-cond-exec
9263 @opindex mmulti-cond-exec
9265 Enable optimization of @code{&&} and @code{||} in conditional execution
9268 This switch is mainly for debugging the compiler and will likely be removed
9269 in a future version.
9271 @item -mno-multi-cond-exec
9272 @opindex mno-multi-cond-exec
9274 Disable optimization of @code{&&} and @code{||} in conditional execution.
9276 This switch is mainly for debugging the compiler and will likely be removed
9277 in a future version.
9279 @item -mnested-cond-exec
9280 @opindex mnested-cond-exec
9282 Enable nested conditional execution optimizations (default).
9284 This switch is mainly for debugging the compiler and will likely be removed
9285 in a future version.
9287 @item -mno-nested-cond-exec
9288 @opindex mno-nested-cond-exec
9290 Disable nested conditional execution optimizations.
9292 This switch is mainly for debugging the compiler and will likely be removed
9293 in a future version.
9295 @item -moptimize-membar
9296 @opindex moptimize-membar
9298 This switch removes redundant @code{membar} instructions from the
9299 compiler generated code. It is enabled by default.
9301 @item -mno-optimize-membar
9302 @opindex mno-optimize-membar
9304 This switch disables the automatic removal of redundant @code{membar}
9305 instructions from the generated code.
9307 @item -mtomcat-stats
9308 @opindex mtomcat-stats
9310 Cause gas to print out tomcat statistics.
9312 @item -mcpu=@var{cpu}
9315 Select the processor type for which to generate code. Possible values are
9316 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9317 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9321 @node GNU/Linux Options
9322 @subsection GNU/Linux Options
9324 These @samp{-m} options are defined for GNU/Linux targets:
9329 Use the GNU C library instead of uClibc. This is the default except
9330 on @samp{*-*-linux-*uclibc*} targets.
9334 Use uClibc instead of the GNU C library. This is the default on
9335 @samp{*-*-linux-*uclibc*} targets.
9338 @node H8/300 Options
9339 @subsection H8/300 Options
9341 These @samp{-m} options are defined for the H8/300 implementations:
9346 Shorten some address references at link time, when possible; uses the
9347 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9348 ld, Using ld}, for a fuller description.
9352 Generate code for the H8/300H@.
9356 Generate code for the H8S@.
9360 Generate code for the H8S and H8/300H in the normal mode. This switch
9361 must be used either with @option{-mh} or @option{-ms}.
9365 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9369 Make @code{int} data 32 bits by default.
9373 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9374 The default for the H8/300H and H8S is to align longs and floats on 4
9376 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9377 This option has no effect on the H8/300.
9381 @subsection HPPA Options
9382 @cindex HPPA Options
9384 These @samp{-m} options are defined for the HPPA family of computers:
9387 @item -march=@var{architecture-type}
9389 Generate code for the specified architecture. The choices for
9390 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9391 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9392 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9393 architecture option for your machine. Code compiled for lower numbered
9394 architectures will run on higher numbered architectures, but not the
9398 @itemx -mpa-risc-1-1
9399 @itemx -mpa-risc-2-0
9400 @opindex mpa-risc-1-0
9401 @opindex mpa-risc-1-1
9402 @opindex mpa-risc-2-0
9403 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9406 @opindex mbig-switch
9407 Generate code suitable for big switch tables. Use this option only if
9408 the assembler/linker complain about out of range branches within a switch
9411 @item -mjump-in-delay
9412 @opindex mjump-in-delay
9413 Fill delay slots of function calls with unconditional jump instructions
9414 by modifying the return pointer for the function call to be the target
9415 of the conditional jump.
9417 @item -mdisable-fpregs
9418 @opindex mdisable-fpregs
9419 Prevent floating point registers from being used in any manner. This is
9420 necessary for compiling kernels which perform lazy context switching of
9421 floating point registers. If you use this option and attempt to perform
9422 floating point operations, the compiler will abort.
9424 @item -mdisable-indexing
9425 @opindex mdisable-indexing
9426 Prevent the compiler from using indexing address modes. This avoids some
9427 rather obscure problems when compiling MIG generated code under MACH@.
9429 @item -mno-space-regs
9430 @opindex mno-space-regs
9431 Generate code that assumes the target has no space registers. This allows
9432 GCC to generate faster indirect calls and use unscaled index address modes.
9434 Such code is suitable for level 0 PA systems and kernels.
9436 @item -mfast-indirect-calls
9437 @opindex mfast-indirect-calls
9438 Generate code that assumes calls never cross space boundaries. This
9439 allows GCC to emit code which performs faster indirect calls.
9441 This option will not work in the presence of shared libraries or nested
9444 @item -mfixed-range=@var{register-range}
9445 @opindex mfixed-range
9446 Generate code treating the given register range as fixed registers.
9447 A fixed register is one that the register allocator can not use. This is
9448 useful when compiling kernel code. A register range is specified as
9449 two registers separated by a dash. Multiple register ranges can be
9450 specified separated by a comma.
9452 @item -mlong-load-store
9453 @opindex mlong-load-store
9454 Generate 3-instruction load and store sequences as sometimes required by
9455 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9458 @item -mportable-runtime
9459 @opindex mportable-runtime
9460 Use the portable calling conventions proposed by HP for ELF systems.
9464 Enable the use of assembler directives only GAS understands.
9466 @item -mschedule=@var{cpu-type}
9468 Schedule code according to the constraints for the machine type
9469 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9470 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9471 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9472 proper scheduling option for your machine. The default scheduling is
9476 @opindex mlinker-opt
9477 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9478 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9479 linkers in which they give bogus error messages when linking some programs.
9482 @opindex msoft-float
9483 Generate output containing library calls for floating point.
9484 @strong{Warning:} the requisite libraries are not available for all HPPA
9485 targets. Normally the facilities of the machine's usual C compiler are
9486 used, but this cannot be done directly in cross-compilation. You must make
9487 your own arrangements to provide suitable library functions for
9488 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9489 does provide software floating point support.
9491 @option{-msoft-float} changes the calling convention in the output file;
9492 therefore, it is only useful if you compile @emph{all} of a program with
9493 this option. In particular, you need to compile @file{libgcc.a}, the
9494 library that comes with GCC, with @option{-msoft-float} in order for
9499 Generate the predefine, @code{_SIO}, for server IO@. The default is
9500 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9501 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9502 options are available under HP-UX and HI-UX@.
9506 Use GNU ld specific options. This passes @option{-shared} to ld when
9507 building a shared library. It is the default when GCC is configured,
9508 explicitly or implicitly, with the GNU linker. This option does not
9509 have any affect on which ld is called, it only changes what parameters
9510 are passed to that ld. The ld that is called is determined by the
9511 @option{--with-ld} configure option, GCC's program search path, and
9512 finally by the user's @env{PATH}. The linker used by GCC can be printed
9513 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9514 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9518 Use HP ld specific options. This passes @option{-b} to ld when building
9519 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9520 links. It is the default when GCC is configured, explicitly or
9521 implicitly, with the HP linker. This option does not have any affect on
9522 which ld is called, it only changes what parameters are passed to that
9523 ld. The ld that is called is determined by the @option{--with-ld}
9524 configure option, GCC's program search path, and finally by the user's
9525 @env{PATH}. The linker used by GCC can be printed using @samp{which
9526 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9527 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9530 @opindex mno-long-calls
9531 Generate code that uses long call sequences. This ensures that a call
9532 is always able to reach linker generated stubs. The default is to generate
9533 long calls only when the distance from the call site to the beginning
9534 of the function or translation unit, as the case may be, exceeds a
9535 predefined limit set by the branch type being used. The limits for
9536 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9537 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9540 Distances are measured from the beginning of functions when using the
9541 @option{-ffunction-sections} option, or when using the @option{-mgas}
9542 and @option{-mno-portable-runtime} options together under HP-UX with
9545 It is normally not desirable to use this option as it will degrade
9546 performance. However, it may be useful in large applications,
9547 particularly when partial linking is used to build the application.
9549 The types of long calls used depends on the capabilities of the
9550 assembler and linker, and the type of code being generated. The
9551 impact on systems that support long absolute calls, and long pic
9552 symbol-difference or pc-relative calls should be relatively small.
9553 However, an indirect call is used on 32-bit ELF systems in pic code
9554 and it is quite long.
9556 @item -munix=@var{unix-std}
9558 Generate compiler predefines and select a startfile for the specified
9559 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9560 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9561 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9562 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9563 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9566 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9567 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9568 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9569 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9570 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9571 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9573 It is @emph{important} to note that this option changes the interfaces
9574 for various library routines. It also affects the operational behavior
9575 of the C library. Thus, @emph{extreme} care is needed in using this
9578 Library code that is intended to operate with more than one UNIX
9579 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9580 as appropriate. Most GNU software doesn't provide this capability.
9584 Suppress the generation of link options to search libdld.sl when the
9585 @option{-static} option is specified on HP-UX 10 and later.
9589 The HP-UX implementation of setlocale in libc has a dependency on
9590 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9591 when the @option{-static} option is specified, special link options
9592 are needed to resolve this dependency.
9594 On HP-UX 10 and later, the GCC driver adds the necessary options to
9595 link with libdld.sl when the @option{-static} option is specified.
9596 This causes the resulting binary to be dynamic. On the 64-bit port,
9597 the linkers generate dynamic binaries by default in any case. The
9598 @option{-nolibdld} option can be used to prevent the GCC driver from
9599 adding these link options.
9603 Add support for multithreading with the @dfn{dce thread} library
9604 under HP-UX@. This option sets flags for both the preprocessor and
9608 @node i386 and x86-64 Options
9609 @subsection Intel 386 and AMD x86-64 Options
9610 @cindex i386 Options
9611 @cindex x86-64 Options
9612 @cindex Intel 386 Options
9613 @cindex AMD x86-64 Options
9615 These @samp{-m} options are defined for the i386 and x86-64 family of
9619 @item -mtune=@var{cpu-type}
9621 Tune to @var{cpu-type} everything applicable about the generated code, except
9622 for the ABI and the set of available instructions. The choices for
9626 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9627 If you know the CPU on which your code will run, then you should use
9628 the corresponding @option{-mtune} option instead of
9629 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9630 of your application will have, then you should use this option.
9632 As new processors are deployed in the marketplace, the behavior of this
9633 option will change. Therefore, if you upgrade to a newer version of
9634 GCC, the code generated option will change to reflect the processors
9635 that were most common when that version of GCC was released.
9637 There is no @option{-march=generic} option because @option{-march}
9638 indicates the instruction set the compiler can use, and there is no
9639 generic instruction set applicable to all processors. In contrast,
9640 @option{-mtune} indicates the processor (or, in this case, collection of
9641 processors) for which the code is optimized.
9643 This selects the CPU to tune for at compilation time by determining
9644 the processor type of the compiling machine. Using @option{-mtune=native}
9645 will produce code optimized for the local machine under the constraints
9646 of the selected instruction set. Using @option{-march=native} will
9647 enable all instruction subsets supported by the local machine (hence
9648 the result might not run on different machines).
9650 Original Intel's i386 CPU@.
9652 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9654 Intel Pentium CPU with no MMX support.
9656 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9658 Intel PentiumPro CPU@.
9660 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9661 instruction set will be used, so the code will run on all i686 family chips.
9663 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9664 @item pentium3, pentium3m
9665 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9668 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9669 support. Used by Centrino notebooks.
9670 @item pentium4, pentium4m
9671 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9673 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9676 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9677 SSE2 and SSE3 instruction set support.
9679 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9680 instruction set support.
9682 AMD K6 CPU with MMX instruction set support.
9684 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9685 @item athlon, athlon-tbird
9686 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9688 @item athlon-4, athlon-xp, athlon-mp
9689 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9690 instruction set support.
9691 @item k8, opteron, athlon64, athlon-fx
9692 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9693 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9695 AMD Family 10 core based CPUs with x86-64 instruction set support. (This
9696 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9697 instruction set extensions.)
9699 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9702 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9703 instruction set support.
9705 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9706 implemented for this chip.)
9708 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9709 implemented for this chip.)
9711 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9714 While picking a specific @var{cpu-type} will schedule things appropriately
9715 for that particular chip, the compiler will not generate any code that
9716 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9719 @item -march=@var{cpu-type}
9721 Generate instructions for the machine type @var{cpu-type}. The choices
9722 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9723 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9725 @item -mcpu=@var{cpu-type}
9727 A deprecated synonym for @option{-mtune}.
9736 @opindex mpentiumpro
9737 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9738 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9739 These synonyms are deprecated.
9741 @item -mfpmath=@var{unit}
9743 Generate floating point arithmetics for selected unit @var{unit}. The choices
9748 Use the standard 387 floating point coprocessor present majority of chips and
9749 emulated otherwise. Code compiled with this option will run almost everywhere.
9750 The temporary results are computed in 80bit precision instead of precision
9751 specified by the type resulting in slightly different results compared to most
9752 of other chips. See @option{-ffloat-store} for more detailed description.
9754 This is the default choice for i386 compiler.
9757 Use scalar floating point instructions present in the SSE instruction set.
9758 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9759 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9760 instruction set supports only single precision arithmetics, thus the double and
9761 extended precision arithmetics is still done using 387. Later version, present
9762 only in Pentium4 and the future AMD x86-64 chips supports double precision
9765 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9766 or @option{-msse2} switches to enable SSE extensions and make this option
9767 effective. For the x86-64 compiler, these extensions are enabled by default.
9769 The resulting code should be considerably faster in the majority of cases and avoid
9770 the numerical instability problems of 387 code, but may break some existing
9771 code that expects temporaries to be 80bit.
9773 This is the default choice for the x86-64 compiler.
9776 Attempt to utilize both instruction sets at once. This effectively double the
9777 amount of available registers and on chips with separate execution units for
9778 387 and SSE the execution resources too. Use this option with care, as it is
9779 still experimental, because the GCC register allocator does not model separate
9780 functional units well resulting in instable performance.
9783 @item -masm=@var{dialect}
9784 @opindex masm=@var{dialect}
9785 Output asm instructions using selected @var{dialect}. Supported
9786 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9787 not support @samp{intel}.
9792 @opindex mno-ieee-fp
9793 Control whether or not the compiler uses IEEE floating point
9794 comparisons. These handle correctly the case where the result of a
9795 comparison is unordered.
9798 @opindex msoft-float
9799 Generate output containing library calls for floating point.
9800 @strong{Warning:} the requisite libraries are not part of GCC@.
9801 Normally the facilities of the machine's usual C compiler are used, but
9802 this can't be done directly in cross-compilation. You must make your
9803 own arrangements to provide suitable library functions for
9806 On machines where a function returns floating point results in the 80387
9807 register stack, some floating point opcodes may be emitted even if
9808 @option{-msoft-float} is used.
9810 @item -mno-fp-ret-in-387
9811 @opindex mno-fp-ret-in-387
9812 Do not use the FPU registers for return values of functions.
9814 The usual calling convention has functions return values of types
9815 @code{float} and @code{double} in an FPU register, even if there
9816 is no FPU@. The idea is that the operating system should emulate
9819 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9820 in ordinary CPU registers instead.
9822 @item -mno-fancy-math-387
9823 @opindex mno-fancy-math-387
9824 Some 387 emulators do not support the @code{sin}, @code{cos} and
9825 @code{sqrt} instructions for the 387. Specify this option to avoid
9826 generating those instructions. This option is the default on FreeBSD,
9827 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9828 indicates that the target cpu will always have an FPU and so the
9829 instruction will not need emulation. As of revision 2.6.1, these
9830 instructions are not generated unless you also use the
9831 @option{-funsafe-math-optimizations} switch.
9833 @item -malign-double
9834 @itemx -mno-align-double
9835 @opindex malign-double
9836 @opindex mno-align-double
9837 Control whether GCC aligns @code{double}, @code{long double}, and
9838 @code{long long} variables on a two word boundary or a one word
9839 boundary. Aligning @code{double} variables on a two word boundary will
9840 produce code that runs somewhat faster on a @samp{Pentium} at the
9841 expense of more memory.
9843 On x86-64, @option{-malign-double} is enabled by default.
9845 @strong{Warning:} if you use the @option{-malign-double} switch,
9846 structures containing the above types will be aligned differently than
9847 the published application binary interface specifications for the 386
9848 and will not be binary compatible with structures in code compiled
9849 without that switch.
9851 @item -m96bit-long-double
9852 @itemx -m128bit-long-double
9853 @opindex m96bit-long-double
9854 @opindex m128bit-long-double
9855 These switches control the size of @code{long double} type. The i386
9856 application binary interface specifies the size to be 96 bits,
9857 so @option{-m96bit-long-double} is the default in 32 bit mode.
9859 Modern architectures (Pentium and newer) would prefer @code{long double}
9860 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9861 conforming to the ABI, this would not be possible. So specifying a
9862 @option{-m128bit-long-double} will align @code{long double}
9863 to a 16 byte boundary by padding the @code{long double} with an additional
9866 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9867 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9869 Notice that neither of these options enable any extra precision over the x87
9870 standard of 80 bits for a @code{long double}.
9872 @strong{Warning:} if you override the default value for your target ABI, the
9873 structures and arrays containing @code{long double} variables will change
9874 their size as well as function calling convention for function taking
9875 @code{long double} will be modified. Hence they will not be binary
9876 compatible with arrays or structures in code compiled without that switch.
9878 @item -mmlarge-data-threshold=@var{number}
9879 @opindex mlarge-data-threshold=@var{number}
9880 When @option{-mcmodel=medium} is specified, the data greater than
9881 @var{threshold} are placed in large data section. This value must be the
9882 same across all object linked into the binary and defaults to 65535.
9885 @itemx -mno-svr3-shlib
9886 @opindex msvr3-shlib
9887 @opindex mno-svr3-shlib
9888 Control whether GCC places uninitialized local variables into the
9889 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9890 into @code{bss}. These options are meaningful only on System V Release 3.
9894 Use a different function-calling convention, in which functions that
9895 take a fixed number of arguments return with the @code{ret} @var{num}
9896 instruction, which pops their arguments while returning. This saves one
9897 instruction in the caller since there is no need to pop the arguments
9900 You can specify that an individual function is called with this calling
9901 sequence with the function attribute @samp{stdcall}. You can also
9902 override the @option{-mrtd} option by using the function attribute
9903 @samp{cdecl}. @xref{Function Attributes}.
9905 @strong{Warning:} this calling convention is incompatible with the one
9906 normally used on Unix, so you cannot use it if you need to call
9907 libraries compiled with the Unix compiler.
9909 Also, you must provide function prototypes for all functions that
9910 take variable numbers of arguments (including @code{printf});
9911 otherwise incorrect code will be generated for calls to those
9914 In addition, seriously incorrect code will result if you call a
9915 function with too many arguments. (Normally, extra arguments are
9916 harmlessly ignored.)
9918 @item -mregparm=@var{num}
9920 Control how many registers are used to pass integer arguments. By
9921 default, no registers are used to pass arguments, and at most 3
9922 registers can be used. You can control this behavior for a specific
9923 function by using the function attribute @samp{regparm}.
9924 @xref{Function Attributes}.
9926 @strong{Warning:} if you use this switch, and
9927 @var{num} is nonzero, then you must build all modules with the same
9928 value, including any libraries. This includes the system libraries and
9932 @opindex msseregparm
9933 Use SSE register passing conventions for float and double arguments
9934 and return values. You can control this behavior for a specific
9935 function by using the function attribute @samp{sseregparm}.
9936 @xref{Function Attributes}.
9938 @strong{Warning:} if you use this switch then you must build all
9939 modules with the same value, including any libraries. This includes
9940 the system libraries and startup modules.
9942 @item -mstackrealign
9943 @opindex mstackrealign
9944 Realign the stack at entry. On the Intel x86, the
9945 @option{-mstackrealign} option will generate an alternate prologue and
9946 epilogue that realigns the runtime stack. This supports mixing legacy
9947 codes that keep a 4-byte aligned stack with modern codes that keep a
9948 16-byte stack for SSE compatibility. The alternate prologue and
9949 epilogue are slower and bigger than the regular ones, and the
9950 alternate prologue requires an extra scratch register; this lowers the
9951 number of registers available if used in conjunction with the
9952 @code{regparm} attribute. The @option{-mstackrealign} option is
9953 incompatible with the nested function prologue; this is considered a
9954 hard error. See also the attribute @code{force_align_arg_pointer},
9955 applicable to individual functions.
9957 @item -mpreferred-stack-boundary=@var{num}
9958 @opindex mpreferred-stack-boundary
9959 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9960 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9961 the default is 4 (16 bytes or 128 bits).
9963 On Pentium and PentiumPro, @code{double} and @code{long double} values
9964 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9965 suffer significant run time performance penalties. On Pentium III, the
9966 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9967 properly if it is not 16 byte aligned.
9969 To ensure proper alignment of this values on the stack, the stack boundary
9970 must be as aligned as that required by any value stored on the stack.
9971 Further, every function must be generated such that it keeps the stack
9972 aligned. Thus calling a function compiled with a higher preferred
9973 stack boundary from a function compiled with a lower preferred stack
9974 boundary will most likely misalign the stack. It is recommended that
9975 libraries that use callbacks always use the default setting.
9977 This extra alignment does consume extra stack space, and generally
9978 increases code size. Code that is sensitive to stack space usage, such
9979 as embedded systems and operating system kernels, may want to reduce the
9980 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10006 These switches enable or disable the use of instructions in the MMX,
10007 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM or 3DNow! extended instruction sets.
10008 These extensions are also available as built-in functions: see
10009 @ref{X86 Built-in Functions}, for details of the functions enabled and
10010 disabled by these switches.
10012 To have SSE/SSE2 instructions generated automatically from floating-point
10013 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10015 These options will enable GCC to use these extended instructions in
10016 generated code, even without @option{-mfpmath=sse}. Applications which
10017 perform runtime CPU detection must compile separate files for each
10018 supported architecture, using the appropriate flags. In particular,
10019 the file containing the CPU detection code should be compiled without
10023 @itemx -mno-push-args
10024 @opindex mpush-args
10025 @opindex mno-push-args
10026 Use PUSH operations to store outgoing parameters. This method is shorter
10027 and usually equally fast as method using SUB/MOV operations and is enabled
10028 by default. In some cases disabling it may improve performance because of
10029 improved scheduling and reduced dependencies.
10031 @item -maccumulate-outgoing-args
10032 @opindex maccumulate-outgoing-args
10033 If enabled, the maximum amount of space required for outgoing arguments will be
10034 computed in the function prologue. This is faster on most modern CPUs
10035 because of reduced dependencies, improved scheduling and reduced stack usage
10036 when preferred stack boundary is not equal to 2. The drawback is a notable
10037 increase in code size. This switch implies @option{-mno-push-args}.
10041 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10042 on thread-safe exception handling must compile and link all code with the
10043 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10044 @option{-D_MT}; when linking, it links in a special thread helper library
10045 @option{-lmingwthrd} which cleans up per thread exception handling data.
10047 @item -mno-align-stringops
10048 @opindex mno-align-stringops
10049 Do not align destination of inlined string operations. This switch reduces
10050 code size and improves performance in case the destination is already aligned,
10051 but GCC doesn't know about it.
10053 @item -minline-all-stringops
10054 @opindex minline-all-stringops
10055 By default GCC inlines string operations only when destination is known to be
10056 aligned at least to 4 byte boundary. This enables more inlining, increase code
10057 size, but may improve performance of code that depends on fast memcpy, strlen
10058 and memset for short lengths.
10060 @item -minline-stringops-dynamically
10061 @opindex minline-stringops-dynamically
10062 For string operation of unknown size, inline runtime checks so for small
10063 blocks inline code is used, while for large blocks library call is used.
10065 @item -mstringop-strategy=@var{alg}
10066 @opindex mstringop-strategy=@var{alg}
10067 Overwrite internal decision heuristic about particular algorithm to inline
10068 string operation with. The allowed values are @code{rep_byte},
10069 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10070 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10071 expanding inline loop, @code{libcall} for always expanding library call.
10073 @item -momit-leaf-frame-pointer
10074 @opindex momit-leaf-frame-pointer
10075 Don't keep the frame pointer in a register for leaf functions. This
10076 avoids the instructions to save, set up and restore frame pointers and
10077 makes an extra register available in leaf functions. The option
10078 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10079 which might make debugging harder.
10081 @item -mtls-direct-seg-refs
10082 @itemx -mno-tls-direct-seg-refs
10083 @opindex mtls-direct-seg-refs
10084 Controls whether TLS variables may be accessed with offsets from the
10085 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10086 or whether the thread base pointer must be added. Whether or not this
10087 is legal depends on the operating system, and whether it maps the
10088 segment to cover the entire TLS area.
10090 For systems that use GNU libc, the default is on.
10093 These @samp{-m} switches are supported in addition to the above
10094 on AMD x86-64 processors in 64-bit environments.
10101 Generate code for a 32-bit or 64-bit environment.
10102 The 32-bit environment sets int, long and pointer to 32 bits and
10103 generates code that runs on any i386 system.
10104 The 64-bit environment sets int to 32 bits and long and pointer
10105 to 64 bits and generates code for AMD's x86-64 architecture.
10107 @item -mno-red-zone
10108 @opindex no-red-zone
10109 Do not use a so called red zone for x86-64 code. The red zone is mandated
10110 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10111 stack pointer that will not be modified by signal or interrupt handlers
10112 and therefore can be used for temporary data without adjusting the stack
10113 pointer. The flag @option{-mno-red-zone} disables this red zone.
10115 @item -mcmodel=small
10116 @opindex mcmodel=small
10117 Generate code for the small code model: the program and its symbols must
10118 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10119 Programs can be statically or dynamically linked. This is the default
10122 @item -mcmodel=kernel
10123 @opindex mcmodel=kernel
10124 Generate code for the kernel code model. The kernel runs in the
10125 negative 2 GB of the address space.
10126 This model has to be used for Linux kernel code.
10128 @item -mcmodel=medium
10129 @opindex mcmodel=medium
10130 Generate code for the medium model: The program is linked in the lower 2
10131 GB of the address space but symbols can be located anywhere in the
10132 address space. Programs can be statically or dynamically linked, but
10133 building of shared libraries are not supported with the medium model.
10135 @item -mcmodel=large
10136 @opindex mcmodel=large
10137 Generate code for the large model: This model makes no assumptions
10138 about addresses and sizes of sections. Currently GCC does not implement
10142 @node IA-64 Options
10143 @subsection IA-64 Options
10144 @cindex IA-64 Options
10146 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10150 @opindex mbig-endian
10151 Generate code for a big endian target. This is the default for HP-UX@.
10153 @item -mlittle-endian
10154 @opindex mlittle-endian
10155 Generate code for a little endian target. This is the default for AIX5
10161 @opindex mno-gnu-as
10162 Generate (or don't) code for the GNU assembler. This is the default.
10163 @c Also, this is the default if the configure option @option{--with-gnu-as}
10169 @opindex mno-gnu-ld
10170 Generate (or don't) code for the GNU linker. This is the default.
10171 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10176 Generate code that does not use a global pointer register. The result
10177 is not position independent code, and violates the IA-64 ABI@.
10179 @item -mvolatile-asm-stop
10180 @itemx -mno-volatile-asm-stop
10181 @opindex mvolatile-asm-stop
10182 @opindex mno-volatile-asm-stop
10183 Generate (or don't) a stop bit immediately before and after volatile asm
10186 @item -mregister-names
10187 @itemx -mno-register-names
10188 @opindex mregister-names
10189 @opindex mno-register-names
10190 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10191 the stacked registers. This may make assembler output more readable.
10197 Disable (or enable) optimizations that use the small data section. This may
10198 be useful for working around optimizer bugs.
10200 @item -mconstant-gp
10201 @opindex mconstant-gp
10202 Generate code that uses a single constant global pointer value. This is
10203 useful when compiling kernel code.
10207 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10208 This is useful when compiling firmware code.
10210 @item -minline-float-divide-min-latency
10211 @opindex minline-float-divide-min-latency
10212 Generate code for inline divides of floating point values
10213 using the minimum latency algorithm.
10215 @item -minline-float-divide-max-throughput
10216 @opindex minline-float-divide-max-throughput
10217 Generate code for inline divides of floating point values
10218 using the maximum throughput algorithm.
10220 @item -minline-int-divide-min-latency
10221 @opindex minline-int-divide-min-latency
10222 Generate code for inline divides of integer values
10223 using the minimum latency algorithm.
10225 @item -minline-int-divide-max-throughput
10226 @opindex minline-int-divide-max-throughput
10227 Generate code for inline divides of integer values
10228 using the maximum throughput algorithm.
10230 @item -minline-sqrt-min-latency
10231 @opindex minline-sqrt-min-latency
10232 Generate code for inline square roots
10233 using the minimum latency algorithm.
10235 @item -minline-sqrt-max-throughput
10236 @opindex minline-sqrt-max-throughput
10237 Generate code for inline square roots
10238 using the maximum throughput algorithm.
10240 @item -mno-dwarf2-asm
10241 @itemx -mdwarf2-asm
10242 @opindex mno-dwarf2-asm
10243 @opindex mdwarf2-asm
10244 Don't (or do) generate assembler code for the DWARF2 line number debugging
10245 info. This may be useful when not using the GNU assembler.
10247 @item -mearly-stop-bits
10248 @itemx -mno-early-stop-bits
10249 @opindex mearly-stop-bits
10250 @opindex mno-early-stop-bits
10251 Allow stop bits to be placed earlier than immediately preceding the
10252 instruction that triggered the stop bit. This can improve instruction
10253 scheduling, but does not always do so.
10255 @item -mfixed-range=@var{register-range}
10256 @opindex mfixed-range
10257 Generate code treating the given register range as fixed registers.
10258 A fixed register is one that the register allocator can not use. This is
10259 useful when compiling kernel code. A register range is specified as
10260 two registers separated by a dash. Multiple register ranges can be
10261 specified separated by a comma.
10263 @item -mtls-size=@var{tls-size}
10265 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
10268 @item -mtune=@var{cpu-type}
10270 Tune the instruction scheduling for a particular CPU, Valid values are
10271 itanium, itanium1, merced, itanium2, and mckinley.
10277 Add support for multithreading using the POSIX threads library. This
10278 option sets flags for both the preprocessor and linker. It does
10279 not affect the thread safety of object code produced by the compiler or
10280 that of libraries supplied with it. These are HP-UX specific flags.
10286 Generate code for a 32-bit or 64-bit environment.
10287 The 32-bit environment sets int, long and pointer to 32 bits.
10288 The 64-bit environment sets int to 32 bits and long and pointer
10289 to 64 bits. These are HP-UX specific flags.
10291 @item -mno-sched-br-data-spec
10292 @itemx -msched-br-data-spec
10293 @opindex -mno-sched-br-data-spec
10294 @opindex -msched-br-data-spec
10295 (Dis/En)able data speculative scheduling before reload.
10296 This will result in generation of the ld.a instructions and
10297 the corresponding check instructions (ld.c / chk.a).
10298 The default is 'disable'.
10300 @item -msched-ar-data-spec
10301 @itemx -mno-sched-ar-data-spec
10302 @opindex -msched-ar-data-spec
10303 @opindex -mno-sched-ar-data-spec
10304 (En/Dis)able data speculative scheduling after reload.
10305 This will result in generation of the ld.a instructions and
10306 the corresponding check instructions (ld.c / chk.a).
10307 The default is 'enable'.
10309 @item -mno-sched-control-spec
10310 @itemx -msched-control-spec
10311 @opindex -mno-sched-control-spec
10312 @opindex -msched-control-spec
10313 (Dis/En)able control speculative scheduling. This feature is
10314 available only during region scheduling (i.e. before reload).
10315 This will result in generation of the ld.s instructions and
10316 the corresponding check instructions chk.s .
10317 The default is 'disable'.
10319 @item -msched-br-in-data-spec
10320 @itemx -mno-sched-br-in-data-spec
10321 @opindex -msched-br-in-data-spec
10322 @opindex -mno-sched-br-in-data-spec
10323 (En/Dis)able speculative scheduling of the instructions that
10324 are dependent on the data speculative loads before reload.
10325 This is effective only with @option{-msched-br-data-spec} enabled.
10326 The default is 'enable'.
10328 @item -msched-ar-in-data-spec
10329 @itemx -mno-sched-ar-in-data-spec
10330 @opindex -msched-ar-in-data-spec
10331 @opindex -mno-sched-ar-in-data-spec
10332 (En/Dis)able speculative scheduling of the instructions that
10333 are dependent on the data speculative loads after reload.
10334 This is effective only with @option{-msched-ar-data-spec} enabled.
10335 The default is 'enable'.
10337 @item -msched-in-control-spec
10338 @itemx -mno-sched-in-control-spec
10339 @opindex -msched-in-control-spec
10340 @opindex -mno-sched-in-control-spec
10341 (En/Dis)able speculative scheduling of the instructions that
10342 are dependent on the control speculative loads.
10343 This is effective only with @option{-msched-control-spec} enabled.
10344 The default is 'enable'.
10347 @itemx -mno-sched-ldc
10348 @opindex -msched-ldc
10349 @opindex -mno-sched-ldc
10350 (En/Dis)able use of simple data speculation checks ld.c .
10351 If disabled, only chk.a instructions will be emitted to check
10352 data speculative loads.
10353 The default is 'enable'.
10355 @item -mno-sched-control-ldc
10356 @itemx -msched-control-ldc
10357 @opindex -mno-sched-control-ldc
10358 @opindex -msched-control-ldc
10359 (Dis/En)able use of ld.c instructions to check control speculative loads.
10360 If enabled, in case of control speculative load with no speculatively
10361 scheduled dependent instructions this load will be emitted as ld.sa and
10362 ld.c will be used to check it.
10363 The default is 'disable'.
10365 @item -mno-sched-spec-verbose
10366 @itemx -msched-spec-verbose
10367 @opindex -mno-sched-spec-verbose
10368 @opindex -msched-spec-verbose
10369 (Dis/En)able printing of the information about speculative motions.
10371 @item -mno-sched-prefer-non-data-spec-insns
10372 @itemx -msched-prefer-non-data-spec-insns
10373 @opindex -mno-sched-prefer-non-data-spec-insns
10374 @opindex -msched-prefer-non-data-spec-insns
10375 If enabled, data speculative instructions will be chosen for schedule
10376 only if there are no other choices at the moment. This will make
10377 the use of the data speculation much more conservative.
10378 The default is 'disable'.
10380 @item -mno-sched-prefer-non-control-spec-insns
10381 @itemx -msched-prefer-non-control-spec-insns
10382 @opindex -mno-sched-prefer-non-control-spec-insns
10383 @opindex -msched-prefer-non-control-spec-insns
10384 If enabled, control speculative instructions will be chosen for schedule
10385 only if there are no other choices at the moment. This will make
10386 the use of the control speculation much more conservative.
10387 The default is 'disable'.
10389 @item -mno-sched-count-spec-in-critical-path
10390 @itemx -msched-count-spec-in-critical-path
10391 @opindex -mno-sched-count-spec-in-critical-path
10392 @opindex -msched-count-spec-in-critical-path
10393 If enabled, speculative dependencies will be considered during
10394 computation of the instructions priorities. This will make the use of the
10395 speculation a bit more conservative.
10396 The default is 'disable'.
10401 @subsection M32C Options
10402 @cindex M32C options
10405 @item -mcpu=@var{name}
10407 Select the CPU for which code is generated. @var{name} may be one of
10408 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10409 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10410 the M32C/80 series.
10414 Specifies that the program will be run on the simulator. This causes
10415 an alternate runtime library to be linked in which supports, for
10416 example, file I/O. You must not use this option when generating
10417 programs that will run on real hardware; you must provide your own
10418 runtime library for whatever I/O functions are needed.
10420 @item -memregs=@var{number}
10422 Specifies the number of memory-based pseudo-registers GCC will use
10423 during code generation. These pseudo-registers will be used like real
10424 registers, so there is a tradeoff between GCC's ability to fit the
10425 code into available registers, and the performance penalty of using
10426 memory instead of registers. Note that all modules in a program must
10427 be compiled with the same value for this option. Because of that, you
10428 must not use this option with the default runtime libraries gcc
10433 @node M32R/D Options
10434 @subsection M32R/D Options
10435 @cindex M32R/D options
10437 These @option{-m} options are defined for Renesas M32R/D architectures:
10442 Generate code for the M32R/2@.
10446 Generate code for the M32R/X@.
10450 Generate code for the M32R@. This is the default.
10452 @item -mmodel=small
10453 @opindex mmodel=small
10454 Assume all objects live in the lower 16MB of memory (so that their addresses
10455 can be loaded with the @code{ld24} instruction), and assume all subroutines
10456 are reachable with the @code{bl} instruction.
10457 This is the default.
10459 The addressability of a particular object can be set with the
10460 @code{model} attribute.
10462 @item -mmodel=medium
10463 @opindex mmodel=medium
10464 Assume objects may be anywhere in the 32-bit address space (the compiler
10465 will generate @code{seth/add3} instructions to load their addresses), and
10466 assume all subroutines are reachable with the @code{bl} instruction.
10468 @item -mmodel=large
10469 @opindex mmodel=large
10470 Assume objects may be anywhere in the 32-bit address space (the compiler
10471 will generate @code{seth/add3} instructions to load their addresses), and
10472 assume subroutines may not be reachable with the @code{bl} instruction
10473 (the compiler will generate the much slower @code{seth/add3/jl}
10474 instruction sequence).
10477 @opindex msdata=none
10478 Disable use of the small data area. Variables will be put into
10479 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10480 @code{section} attribute has been specified).
10481 This is the default.
10483 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10484 Objects may be explicitly put in the small data area with the
10485 @code{section} attribute using one of these sections.
10487 @item -msdata=sdata
10488 @opindex msdata=sdata
10489 Put small global and static data in the small data area, but do not
10490 generate special code to reference them.
10493 @opindex msdata=use
10494 Put small global and static data in the small data area, and generate
10495 special instructions to reference them.
10499 @cindex smaller data references
10500 Put global and static objects less than or equal to @var{num} bytes
10501 into the small data or bss sections instead of the normal data or bss
10502 sections. The default value of @var{num} is 8.
10503 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10504 for this option to have any effect.
10506 All modules should be compiled with the same @option{-G @var{num}} value.
10507 Compiling with different values of @var{num} may or may not work; if it
10508 doesn't the linker will give an error message---incorrect code will not be
10513 Makes the M32R specific code in the compiler display some statistics
10514 that might help in debugging programs.
10516 @item -malign-loops
10517 @opindex malign-loops
10518 Align all loops to a 32-byte boundary.
10520 @item -mno-align-loops
10521 @opindex mno-align-loops
10522 Do not enforce a 32-byte alignment for loops. This is the default.
10524 @item -missue-rate=@var{number}
10525 @opindex missue-rate=@var{number}
10526 Issue @var{number} instructions per cycle. @var{number} can only be 1
10529 @item -mbranch-cost=@var{number}
10530 @opindex mbranch-cost=@var{number}
10531 @var{number} can only be 1 or 2. If it is 1 then branches will be
10532 preferred over conditional code, if it is 2, then the opposite will
10535 @item -mflush-trap=@var{number}
10536 @opindex mflush-trap=@var{number}
10537 Specifies the trap number to use to flush the cache. The default is
10538 12. Valid numbers are between 0 and 15 inclusive.
10540 @item -mno-flush-trap
10541 @opindex mno-flush-trap
10542 Specifies that the cache cannot be flushed by using a trap.
10544 @item -mflush-func=@var{name}
10545 @opindex mflush-func=@var{name}
10546 Specifies the name of the operating system function to call to flush
10547 the cache. The default is @emph{_flush_cache}, but a function call
10548 will only be used if a trap is not available.
10550 @item -mno-flush-func
10551 @opindex mno-flush-func
10552 Indicates that there is no OS function for flushing the cache.
10556 @node M680x0 Options
10557 @subsection M680x0 Options
10558 @cindex M680x0 options
10560 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
10561 The default settings depend on which architecture was selected when
10562 the compiler was configured; the defaults for the most common choices
10566 @item -march=@var{arch}
10568 Generate code for a specific M680x0 or ColdFire instruction set
10569 architecture. Permissible values of @var{arch} for M680x0
10570 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
10571 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
10572 architectures are selected according to Freescale's ISA classification
10573 and the permissible values are: @samp{isaa}, @samp{isaaplus},
10574 @samp{isab} and @samp{isac}.
10576 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
10577 code for a ColdFire target. The @var{arch} in this macro is one of the
10578 @option{-march} arguments given above.
10580 When used together, @option{-march} and @option{-mtune} select code
10581 that runs on a family of similar processors but that is optimized
10582 for a particular microarchitecture.
10584 @item -mcpu=@var{cpu}
10586 Generate code for a specific M680x0 or ColdFire processor.
10587 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
10588 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
10589 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
10590 below, which also classifies the CPUs into families:
10592 @multitable @columnfractions 0.20 0.80
10593 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
10594 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
10595 @item @samp{5206e} @tab @samp{5206e}
10596 @item @samp{5208} @tab @samp{5207} @samp{5208}
10597 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
10598 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
10599 @item @samp{5216} @tab @samp{5214} @samp{5216}
10600 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
10601 @item @samp{5225} @tab @samp{5224} @samp{5225}
10602 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
10603 @item @samp{5249} @tab @samp{5249}
10604 @item @samp{5250} @tab @samp{5250}
10605 @item @samp{5271} @tab @samp{5270} @samp{5271}
10606 @item @samp{5272} @tab @samp{5272}
10607 @item @samp{5275} @tab @samp{5274} @samp{5275}
10608 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
10609 @item @samp{5307} @tab @samp{5307}
10610 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
10611 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
10612 @item @samp{5407} @tab @samp{5407}
10613 @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}
10616 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
10617 @var{arch} is compatible with @var{cpu}. Other combinations of
10618 @option{-mcpu} and @option{-march} are rejected.
10620 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
10621 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
10622 where the value of @var{family} is given by the table above.
10624 @item -mtune=@var{tune}
10626 Tune the code for a particular microarchitecture, within the
10627 constraints set by @option{-march} and @option{-mcpu}.
10628 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
10629 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
10630 and @samp{cpu32}. The ColdFire microarchitectures
10631 are: @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
10633 You can also use @option{-mtune=68020-40} for code that needs
10634 to run relatively well on 68020, 68030 and 68040 targets.
10635 @option{-mtune=68020-60} is similar but includes 68060 targets
10636 as well. These two options select the same tuning decisions as
10637 @option{-m68020-40} and @option{-m68020-60} respectively.
10639 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
10640 when tuning for 680x0 architecture @var{arch}. It also defines
10641 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
10642 option is used. If gcc is tuning for a range of architectures,
10643 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
10644 it defines the macros for every architecture in the range.
10646 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
10647 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
10648 of the arguments given above.
10654 Generate output for a 68000. This is the default
10655 when the compiler is configured for 68000-based systems.
10656 It is equivalent to @option{-march=68000}.
10658 Use this option for microcontrollers with a 68000 or EC000 core,
10659 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10663 Generate output for a 68010. This is the default
10664 when the compiler is configured for 68010-based systems.
10665 It is equivalent to @option{-march=68010}.
10671 Generate output for a 68020. This is the default
10672 when the compiler is configured for 68020-based systems.
10673 It is equivalent to @option{-march=68020}.
10677 Generate output for a 68030. This is the default when the compiler is
10678 configured for 68030-based systems. It is equivalent to
10679 @option{-march=68030}.
10683 Generate output for a 68040. This is the default when the compiler is
10684 configured for 68040-based systems. It is equivalent to
10685 @option{-march=68040}.
10687 This option inhibits the use of 68881/68882 instructions that have to be
10688 emulated by software on the 68040. Use this option if your 68040 does not
10689 have code to emulate those instructions.
10693 Generate output for a 68060. This is the default when the compiler is
10694 configured for 68060-based systems. It is equivalent to
10695 @option{-march=68060}.
10697 This option inhibits the use of 68020 and 68881/68882 instructions that
10698 have to be emulated by software on the 68060. Use this option if your 68060
10699 does not have code to emulate those instructions.
10703 Generate output for a CPU32. This is the default
10704 when the compiler is configured for CPU32-based systems.
10705 It is equivalent to @option{-march=cpu32}.
10707 Use this option for microcontrollers with a
10708 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10709 68336, 68340, 68341, 68349 and 68360.
10713 Generate output for a 520X ColdFire CPU. This is the default
10714 when the compiler is configured for 520X-based systems.
10715 It is equivalent to @option{-mcpu=5206}, and is now deprecated
10716 in favor of that option.
10718 Use this option for microcontroller with a 5200 core, including
10719 the MCF5202, MCF5203, MCF5204 and MCF5206.
10723 Generate output for a 5206e ColdFire CPU. The option is now
10724 deprecated in favor of the equivalent @option{-mcpu=5206e}.
10728 Generate output for a member of the ColdFire 528X family.
10729 The option is now deprecated in favor of the equivalent
10730 @option{-mcpu=528x}.
10734 Generate output for a ColdFire 5307 CPU. The option is now deprecated
10735 in favor of the equivalent @option{-mcpu=5307}.
10739 Generate output for a ColdFire 5407 CPU. The option is now deprecated
10740 in favor of the equivalent @option{-mcpu=5407}.
10744 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
10745 This includes use of hardware floating point instructions.
10746 The option is equivalent to @option{-mcpu=547x}, and is now
10747 deprecated in favor of that option.
10751 Generate output for a 68040, without using any of the new instructions.
10752 This results in code which can run relatively efficiently on either a
10753 68020/68881 or a 68030 or a 68040. The generated code does use the
10754 68881 instructions that are emulated on the 68040.
10756 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
10760 Generate output for a 68060, without using any of the new instructions.
10761 This results in code which can run relatively efficiently on either a
10762 68020/68881 or a 68030 or a 68040. The generated code does use the
10763 68881 instructions that are emulated on the 68060.
10765 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
10769 @opindex mhard-float
10771 Generate floating-point instructions. This is the default for 68020
10772 and above, and for ColdFire devices that have an FPU. It defines the
10773 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
10774 on ColdFire targets.
10777 @opindex msoft-float
10778 Do not generate floating-point instructions; use library calls instead.
10779 This is the default for 68000, 68010, and 68832 targets. It is also
10780 the default for ColdFire devices that have no FPU.
10786 Generate (do not generate) ColdFire hardware divide and remainder
10787 instructions. If @option{-march} is used without @option{-mcpu},
10788 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
10789 architectures. Otherwise, the default is taken from the target CPU
10790 (either the default CPU, or the one specified by @option{-mcpu}). For
10791 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
10792 @option{-mcpu=5206e}.
10794 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
10798 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10799 Additionally, parameters passed on the stack are also aligned to a
10800 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10803 @opindex -mno-short
10804 Do not consider type @code{int} to be 16 bits wide. This is the default.
10807 @itemx -mno-bitfield
10808 @opindex mnobitfield
10809 @opindex mno-bitfield
10810 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10811 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10815 Do use the bit-field instructions. The @option{-m68020} option implies
10816 @option{-mbitfield}. This is the default if you use a configuration
10817 designed for a 68020.
10821 Use a different function-calling convention, in which functions
10822 that take a fixed number of arguments return with the @code{rtd}
10823 instruction, which pops their arguments while returning. This
10824 saves one instruction in the caller since there is no need to pop
10825 the arguments there.
10827 This calling convention is incompatible with the one normally
10828 used on Unix, so you cannot use it if you need to call libraries
10829 compiled with the Unix compiler.
10831 Also, you must provide function prototypes for all functions that
10832 take variable numbers of arguments (including @code{printf});
10833 otherwise incorrect code will be generated for calls to those
10836 In addition, seriously incorrect code will result if you call a
10837 function with too many arguments. (Normally, extra arguments are
10838 harmlessly ignored.)
10840 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10841 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10845 Do not use the calling conventions selected by @option{-mrtd}.
10846 This is the default.
10849 @itemx -mno-align-int
10850 @opindex malign-int
10851 @opindex mno-align-int
10852 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10853 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10854 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10855 Aligning variables on 32-bit boundaries produces code that runs somewhat
10856 faster on processors with 32-bit busses at the expense of more memory.
10858 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10859 align structures containing the above types differently than
10860 most published application binary interface specifications for the m68k.
10864 Use the pc-relative addressing mode of the 68000 directly, instead of
10865 using a global offset table. At present, this option implies @option{-fpic},
10866 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10867 not presently supported with @option{-mpcrel}, though this could be supported for
10868 68020 and higher processors.
10870 @item -mno-strict-align
10871 @itemx -mstrict-align
10872 @opindex mno-strict-align
10873 @opindex mstrict-align
10874 Do not (do) assume that unaligned memory references will be handled by
10878 Generate code that allows the data segment to be located in a different
10879 area of memory from the text segment. This allows for execute in place in
10880 an environment without virtual memory management. This option implies
10883 @item -mno-sep-data
10884 Generate code that assumes that the data segment follows the text segment.
10885 This is the default.
10887 @item -mid-shared-library
10888 Generate code that supports shared libraries via the library ID method.
10889 This allows for execute in place and shared libraries in an environment
10890 without virtual memory management. This option implies @option{-fPIC}.
10892 @item -mno-id-shared-library
10893 Generate code that doesn't assume ID based shared libraries are being used.
10894 This is the default.
10896 @item -mshared-library-id=n
10897 Specified the identification number of the ID based shared library being
10898 compiled. Specifying a value of 0 will generate more compact code, specifying
10899 other values will force the allocation of that number to the current
10900 library but is no more space or time efficient than omitting this option.
10904 @node M68hc1x Options
10905 @subsection M68hc1x Options
10906 @cindex M68hc1x options
10908 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10909 microcontrollers. The default values for these options depends on
10910 which style of microcontroller was selected when the compiler was configured;
10911 the defaults for the most common choices are given below.
10918 Generate output for a 68HC11. This is the default
10919 when the compiler is configured for 68HC11-based systems.
10925 Generate output for a 68HC12. This is the default
10926 when the compiler is configured for 68HC12-based systems.
10932 Generate output for a 68HCS12.
10934 @item -mauto-incdec
10935 @opindex mauto-incdec
10936 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10943 Enable the use of 68HC12 min and max instructions.
10946 @itemx -mno-long-calls
10947 @opindex mlong-calls
10948 @opindex mno-long-calls
10949 Treat all calls as being far away (near). If calls are assumed to be
10950 far away, the compiler will use the @code{call} instruction to
10951 call a function and the @code{rtc} instruction for returning.
10955 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10957 @item -msoft-reg-count=@var{count}
10958 @opindex msoft-reg-count
10959 Specify the number of pseudo-soft registers which are used for the
10960 code generation. The maximum number is 32. Using more pseudo-soft
10961 register may or may not result in better code depending on the program.
10962 The default is 4 for 68HC11 and 2 for 68HC12.
10966 @node MCore Options
10967 @subsection MCore Options
10968 @cindex MCore options
10970 These are the @samp{-m} options defined for the Motorola M*Core
10976 @itemx -mno-hardlit
10978 @opindex mno-hardlit
10979 Inline constants into the code stream if it can be done in two
10980 instructions or less.
10986 Use the divide instruction. (Enabled by default).
10988 @item -mrelax-immediate
10989 @itemx -mno-relax-immediate
10990 @opindex mrelax-immediate
10991 @opindex mno-relax-immediate
10992 Allow arbitrary sized immediates in bit operations.
10994 @item -mwide-bitfields
10995 @itemx -mno-wide-bitfields
10996 @opindex mwide-bitfields
10997 @opindex mno-wide-bitfields
10998 Always treat bit-fields as int-sized.
11000 @item -m4byte-functions
11001 @itemx -mno-4byte-functions
11002 @opindex m4byte-functions
11003 @opindex mno-4byte-functions
11004 Force all functions to be aligned to a four byte boundary.
11006 @item -mcallgraph-data
11007 @itemx -mno-callgraph-data
11008 @opindex mcallgraph-data
11009 @opindex mno-callgraph-data
11010 Emit callgraph information.
11013 @itemx -mno-slow-bytes
11014 @opindex mslow-bytes
11015 @opindex mno-slow-bytes
11016 Prefer word access when reading byte quantities.
11018 @item -mlittle-endian
11019 @itemx -mbig-endian
11020 @opindex mlittle-endian
11021 @opindex mbig-endian
11022 Generate code for a little endian target.
11028 Generate code for the 210 processor.
11032 @subsection MIPS Options
11033 @cindex MIPS options
11039 Generate big-endian code.
11043 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11046 @item -march=@var{arch}
11048 Generate code that will run on @var{arch}, which can be the name of a
11049 generic MIPS ISA, or the name of a particular processor.
11051 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11052 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11053 The processor names are:
11054 @samp{4kc}, @samp{4km}, @samp{4kp},
11055 @samp{4kec}, @samp{4kem}, @samp{4kep},
11056 @samp{5kc}, @samp{5kf},
11058 @samp{24kc}, @samp{24kf}, @samp{24kx},
11059 @samp{24kec}, @samp{24kef}, @samp{24kex},
11060 @samp{34kc}, @samp{34kf}, @samp{34kx},
11063 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11064 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11065 @samp{rm7000}, @samp{rm9000},
11068 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11069 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11070 The special value @samp{from-abi} selects the
11071 most compatible architecture for the selected ABI (that is,
11072 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11074 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11075 (for example, @samp{-march=r2k}). Prefixes are optional, and
11076 @samp{vr} may be written @samp{r}.
11078 GCC defines two macros based on the value of this option. The first
11079 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11080 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11081 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11082 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11083 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11085 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11086 above. In other words, it will have the full prefix and will not
11087 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11088 the macro names the resolved architecture (either @samp{"mips1"} or
11089 @samp{"mips3"}). It names the default architecture when no
11090 @option{-march} option is given.
11092 @item -mtune=@var{arch}
11094 Optimize for @var{arch}. Among other things, this option controls
11095 the way instructions are scheduled, and the perceived cost of arithmetic
11096 operations. The list of @var{arch} values is the same as for
11099 When this option is not used, GCC will optimize for the processor
11100 specified by @option{-march}. By using @option{-march} and
11101 @option{-mtune} together, it is possible to generate code that will
11102 run on a family of processors, but optimize the code for one
11103 particular member of that family.
11105 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11106 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11107 @samp{-march} ones described above.
11111 Equivalent to @samp{-march=mips1}.
11115 Equivalent to @samp{-march=mips2}.
11119 Equivalent to @samp{-march=mips3}.
11123 Equivalent to @samp{-march=mips4}.
11127 Equivalent to @samp{-march=mips32}.
11131 Equivalent to @samp{-march=mips32r2}.
11135 Equivalent to @samp{-march=mips64}.
11140 @opindex mno-mips16
11141 Generate (do not generate) MIPS16 code. If GCC is targetting a
11142 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11154 Generate code for the given ABI@.
11156 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11157 generates 64-bit code when you select a 64-bit architecture, but you
11158 can use @option{-mgp32} to get 32-bit code instead.
11160 For information about the O64 ABI, see
11161 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11163 GCC supports a variant of the o32 ABI in which floating-point registers
11164 are 64 rather than 32 bits wide. You can select this combination with
11165 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11166 and @samp{mfhc1} instructions and is therefore only supported for
11167 MIPS32R2 processors.
11169 The register assignments for arguments and return values remain the
11170 same, but each scalar value is passed in a single 64-bit register
11171 rather than a pair of 32-bit registers. For example, scalar
11172 floating-point values are returned in @samp{$f0} only, not a
11173 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11174 remains the same, but all 64 bits are saved.
11177 @itemx -mno-abicalls
11179 @opindex mno-abicalls
11180 Generate (do not generate) code that is suitable for SVR4-style
11181 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11186 Generate (do not generate) code that is fully position-independent,
11187 and that can therefore be linked into shared libraries. This option
11188 only affects @option{-mabicalls}.
11190 All @option{-mabicalls} code has traditionally been position-independent,
11191 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11192 as an extension, the GNU toolchain allows executables to use absolute
11193 accesses for locally-binding symbols. It can also use shorter GP
11194 initialization sequences and generate direct calls to locally-defined
11195 functions. This mode is selected by @option{-mno-shared}.
11197 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11198 objects that can only be linked by the GNU linker. However, the option
11199 does not affect the ABI of the final executable; it only affects the ABI
11200 of relocatable objects. Using @option{-mno-shared} will generally make
11201 executables both smaller and quicker.
11203 @option{-mshared} is the default.
11209 Lift (do not lift) the usual restrictions on the size of the global
11212 GCC normally uses a single instruction to load values from the GOT@.
11213 While this is relatively efficient, it will only work if the GOT
11214 is smaller than about 64k. Anything larger will cause the linker
11215 to report an error such as:
11217 @cindex relocation truncated to fit (MIPS)
11219 relocation truncated to fit: R_MIPS_GOT16 foobar
11222 If this happens, you should recompile your code with @option{-mxgot}.
11223 It should then work with very large GOTs, although it will also be
11224 less efficient, since it will take three instructions to fetch the
11225 value of a global symbol.
11227 Note that some linkers can create multiple GOTs. If you have such a
11228 linker, you should only need to use @option{-mxgot} when a single object
11229 file accesses more than 64k's worth of GOT entries. Very few do.
11231 These options have no effect unless GCC is generating position
11236 Assume that general-purpose registers are 32 bits wide.
11240 Assume that general-purpose registers are 64 bits wide.
11244 Assume that floating-point registers are 32 bits wide.
11248 Assume that floating-point registers are 64 bits wide.
11251 @opindex mhard-float
11252 Use floating-point coprocessor instructions.
11255 @opindex msoft-float
11256 Do not use floating-point coprocessor instructions. Implement
11257 floating-point calculations using library calls instead.
11259 @item -msingle-float
11260 @opindex msingle-float
11261 Assume that the floating-point coprocessor only supports single-precision
11264 @itemx -mdouble-float
11265 @opindex mdouble-float
11266 Assume that the floating-point coprocessor supports double-precision
11267 operations. This is the default.
11273 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
11275 @itemx -mpaired-single
11276 @itemx -mno-paired-single
11277 @opindex mpaired-single
11278 @opindex mno-paired-single
11279 Use (do not use) paired-single floating-point instructions.
11280 @xref{MIPS Paired-Single Support}. This option can only be used
11281 when generating 64-bit code and requires hardware floating-point
11282 support to be enabled.
11287 @opindex mno-mips3d
11288 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
11289 The option @option{-mips3d} implies @option{-mpaired-single}.
11293 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
11294 an explanation of the default and the way that the pointer size is
11299 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
11301 The default size of @code{int}s, @code{long}s and pointers depends on
11302 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
11303 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
11304 32-bit @code{long}s. Pointers are the same size as @code{long}s,
11305 or the same size as integer registers, whichever is smaller.
11311 Assume (do not assume) that all symbols have 32-bit values, regardless
11312 of the selected ABI@. This option is useful in combination with
11313 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
11314 to generate shorter and faster references to symbolic addresses.
11318 @cindex smaller data references (MIPS)
11319 @cindex gp-relative references (MIPS)
11320 Put global and static items less than or equal to @var{num} bytes into
11321 the small data or bss section instead of the normal data or bss section.
11322 This allows the data to be accessed using a single instruction.
11324 All modules should be compiled with the same @option{-G @var{num}}
11327 @item -membedded-data
11328 @itemx -mno-embedded-data
11329 @opindex membedded-data
11330 @opindex mno-embedded-data
11331 Allocate variables to the read-only data section first if possible, then
11332 next in the small data section if possible, otherwise in data. This gives
11333 slightly slower code than the default, but reduces the amount of RAM required
11334 when executing, and thus may be preferred for some embedded systems.
11336 @item -muninit-const-in-rodata
11337 @itemx -mno-uninit-const-in-rodata
11338 @opindex muninit-const-in-rodata
11339 @opindex mno-uninit-const-in-rodata
11340 Put uninitialized @code{const} variables in the read-only data section.
11341 This option is only meaningful in conjunction with @option{-membedded-data}.
11343 @item -msplit-addresses
11344 @itemx -mno-split-addresses
11345 @opindex msplit-addresses
11346 @opindex mno-split-addresses
11347 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
11348 relocation operators. This option has been superseded by
11349 @option{-mexplicit-relocs} but is retained for backwards compatibility.
11351 @item -mexplicit-relocs
11352 @itemx -mno-explicit-relocs
11353 @opindex mexplicit-relocs
11354 @opindex mno-explicit-relocs
11355 Use (do not use) assembler relocation operators when dealing with symbolic
11356 addresses. The alternative, selected by @option{-mno-explicit-relocs},
11357 is to use assembler macros instead.
11359 @option{-mexplicit-relocs} is the default if GCC was configured
11360 to use an assembler that supports relocation operators.
11362 @item -mcheck-zero-division
11363 @itemx -mno-check-zero-division
11364 @opindex mcheck-zero-division
11365 @opindex mno-check-zero-division
11366 Trap (do not trap) on integer division by zero.
11368 The default is @option{-mcheck-zero-division}.
11370 @item -mdivide-traps
11371 @itemx -mdivide-breaks
11372 @opindex mdivide-traps
11373 @opindex mdivide-breaks
11374 MIPS systems check for division by zero by generating either a
11375 conditional trap or a break instruction. Using traps results in
11376 smaller code, but is only supported on MIPS II and later. Also, some
11377 versions of the Linux kernel have a bug that prevents trap from
11378 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
11379 allow conditional traps on architectures that support them and
11380 @option{-mdivide-breaks} to force the use of breaks.
11382 The default is usually @option{-mdivide-traps}, but this can be
11383 overridden at configure time using @option{--with-divide=breaks}.
11384 Divide-by-zero checks can be completely disabled using
11385 @option{-mno-check-zero-division}.
11390 @opindex mno-memcpy
11391 Force (do not force) the use of @code{memcpy()} for non-trivial block
11392 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
11393 most constant-sized copies.
11396 @itemx -mno-long-calls
11397 @opindex mlong-calls
11398 @opindex mno-long-calls
11399 Disable (do not disable) use of the @code{jal} instruction. Calling
11400 functions using @code{jal} is more efficient but requires the caller
11401 and callee to be in the same 256 megabyte segment.
11403 This option has no effect on abicalls code. The default is
11404 @option{-mno-long-calls}.
11410 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
11411 instructions, as provided by the R4650 ISA@.
11414 @itemx -mno-fused-madd
11415 @opindex mfused-madd
11416 @opindex mno-fused-madd
11417 Enable (disable) use of the floating point multiply-accumulate
11418 instructions, when they are available. The default is
11419 @option{-mfused-madd}.
11421 When multiply-accumulate instructions are used, the intermediate
11422 product is calculated to infinite precision and is not subject to
11423 the FCSR Flush to Zero bit. This may be undesirable in some
11428 Tell the MIPS assembler to not run its preprocessor over user
11429 assembler files (with a @samp{.s} suffix) when assembling them.
11432 @itemx -mno-fix-r4000
11433 @opindex mfix-r4000
11434 @opindex mno-fix-r4000
11435 Work around certain R4000 CPU errata:
11438 A double-word or a variable shift may give an incorrect result if executed
11439 immediately after starting an integer division.
11441 A double-word or a variable shift may give an incorrect result if executed
11442 while an integer multiplication is in progress.
11444 An integer division may give an incorrect result if started in a delay slot
11445 of a taken branch or a jump.
11449 @itemx -mno-fix-r4400
11450 @opindex mfix-r4400
11451 @opindex mno-fix-r4400
11452 Work around certain R4400 CPU errata:
11455 A double-word or a variable shift may give an incorrect result if executed
11456 immediately after starting an integer division.
11460 @itemx -mno-fix-vr4120
11461 @opindex mfix-vr4120
11462 Work around certain VR4120 errata:
11465 @code{dmultu} does not always produce the correct result.
11467 @code{div} and @code{ddiv} do not always produce the correct result if one
11468 of the operands is negative.
11470 The workarounds for the division errata rely on special functions in
11471 @file{libgcc.a}. At present, these functions are only provided by
11472 the @code{mips64vr*-elf} configurations.
11474 Other VR4120 errata require a nop to be inserted between certain pairs of
11475 instructions. These errata are handled by the assembler, not by GCC itself.
11478 @opindex mfix-vr4130
11479 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
11480 workarounds are implemented by the assembler rather than by GCC,
11481 although GCC will avoid using @code{mflo} and @code{mfhi} if the
11482 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
11483 instructions are available instead.
11486 @itemx -mno-fix-sb1
11488 Work around certain SB-1 CPU core errata.
11489 (This flag currently works around the SB-1 revision 2
11490 ``F1'' and ``F2'' floating point errata.)
11492 @item -mflush-func=@var{func}
11493 @itemx -mno-flush-func
11494 @opindex mflush-func
11495 Specifies the function to call to flush the I and D caches, or to not
11496 call any such function. If called, the function must take the same
11497 arguments as the common @code{_flush_func()}, that is, the address of the
11498 memory range for which the cache is being flushed, the size of the
11499 memory range, and the number 3 (to flush both caches). The default
11500 depends on the target GCC was configured for, but commonly is either
11501 @samp{_flush_func} or @samp{__cpu_flush}.
11503 @item -mbranch-likely
11504 @itemx -mno-branch-likely
11505 @opindex mbranch-likely
11506 @opindex mno-branch-likely
11507 Enable or disable use of Branch Likely instructions, regardless of the
11508 default for the selected architecture. By default, Branch Likely
11509 instructions may be generated if they are supported by the selected
11510 architecture. An exception is for the MIPS32 and MIPS64 architectures
11511 and processors which implement those architectures; for those, Branch
11512 Likely instructions will not be generated by default because the MIPS32
11513 and MIPS64 architectures specifically deprecate their use.
11515 @item -mfp-exceptions
11516 @itemx -mno-fp-exceptions
11517 @opindex mfp-exceptions
11518 Specifies whether FP exceptions are enabled. This affects how we schedule
11519 FP instructions for some processors. The default is that FP exceptions are
11522 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11523 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11526 @item -mvr4130-align
11527 @itemx -mno-vr4130-align
11528 @opindex mvr4130-align
11529 The VR4130 pipeline is two-way superscalar, but can only issue two
11530 instructions together if the first one is 8-byte aligned. When this
11531 option is enabled, GCC will align pairs of instructions that it
11532 thinks should execute in parallel.
11534 This option only has an effect when optimizing for the VR4130.
11535 It normally makes code faster, but at the expense of making it bigger.
11536 It is enabled by default at optimization level @option{-O3}.
11540 @subsection MMIX Options
11541 @cindex MMIX Options
11543 These options are defined for the MMIX:
11547 @itemx -mno-libfuncs
11549 @opindex mno-libfuncs
11550 Specify that intrinsic library functions are being compiled, passing all
11551 values in registers, no matter the size.
11554 @itemx -mno-epsilon
11556 @opindex mno-epsilon
11557 Generate floating-point comparison instructions that compare with respect
11558 to the @code{rE} epsilon register.
11560 @item -mabi=mmixware
11562 @opindex mabi-mmixware
11564 Generate code that passes function parameters and return values that (in
11565 the called function) are seen as registers @code{$0} and up, as opposed to
11566 the GNU ABI which uses global registers @code{$231} and up.
11568 @item -mzero-extend
11569 @itemx -mno-zero-extend
11570 @opindex mzero-extend
11571 @opindex mno-zero-extend
11572 When reading data from memory in sizes shorter than 64 bits, use (do not
11573 use) zero-extending load instructions by default, rather than
11574 sign-extending ones.
11577 @itemx -mno-knuthdiv
11579 @opindex mno-knuthdiv
11580 Make the result of a division yielding a remainder have the same sign as
11581 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
11582 remainder follows the sign of the dividend. Both methods are
11583 arithmetically valid, the latter being almost exclusively used.
11585 @item -mtoplevel-symbols
11586 @itemx -mno-toplevel-symbols
11587 @opindex mtoplevel-symbols
11588 @opindex mno-toplevel-symbols
11589 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
11590 code can be used with the @code{PREFIX} assembly directive.
11594 Generate an executable in the ELF format, rather than the default
11595 @samp{mmo} format used by the @command{mmix} simulator.
11597 @item -mbranch-predict
11598 @itemx -mno-branch-predict
11599 @opindex mbranch-predict
11600 @opindex mno-branch-predict
11601 Use (do not use) the probable-branch instructions, when static branch
11602 prediction indicates a probable branch.
11604 @item -mbase-addresses
11605 @itemx -mno-base-addresses
11606 @opindex mbase-addresses
11607 @opindex mno-base-addresses
11608 Generate (do not generate) code that uses @emph{base addresses}. Using a
11609 base address automatically generates a request (handled by the assembler
11610 and the linker) for a constant to be set up in a global register. The
11611 register is used for one or more base address requests within the range 0
11612 to 255 from the value held in the register. The generally leads to short
11613 and fast code, but the number of different data items that can be
11614 addressed is limited. This means that a program that uses lots of static
11615 data may require @option{-mno-base-addresses}.
11617 @item -msingle-exit
11618 @itemx -mno-single-exit
11619 @opindex msingle-exit
11620 @opindex mno-single-exit
11621 Force (do not force) generated code to have a single exit point in each
11625 @node MN10300 Options
11626 @subsection MN10300 Options
11627 @cindex MN10300 options
11629 These @option{-m} options are defined for Matsushita MN10300 architectures:
11634 Generate code to avoid bugs in the multiply instructions for the MN10300
11635 processors. This is the default.
11637 @item -mno-mult-bug
11638 @opindex mno-mult-bug
11639 Do not generate code to avoid bugs in the multiply instructions for the
11640 MN10300 processors.
11644 Generate code which uses features specific to the AM33 processor.
11648 Do not generate code which uses features specific to the AM33 processor. This
11651 @item -mreturn-pointer-on-d0
11652 @opindex mreturn-pointer-on-d0
11653 When generating a function which returns a pointer, return the pointer
11654 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11655 only in a0, and attempts to call such functions without a prototype
11656 would result in errors. Note that this option is on by default; use
11657 @option{-mno-return-pointer-on-d0} to disable it.
11661 Do not link in the C run-time initialization object file.
11665 Indicate to the linker that it should perform a relaxation optimization pass
11666 to shorten branches, calls and absolute memory addresses. This option only
11667 has an effect when used on the command line for the final link step.
11669 This option makes symbolic debugging impossible.
11673 @subsection MT Options
11676 These @option{-m} options are defined for Morpho MT architectures:
11680 @item -march=@var{cpu-type}
11682 Generate code that will run on @var{cpu-type}, which is the name of a system
11683 representing a certain processor type. Possible values for
11684 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11685 @samp{ms1-16-003} and @samp{ms2}.
11687 When this option is not used, the default is @option{-march=ms1-16-002}.
11691 Use byte loads and stores when generating code.
11695 Do not use byte loads and stores when generating code.
11699 Use simulator runtime
11703 Do not link in the C run-time initialization object file
11704 @file{crti.o}. Other run-time initialization and termination files
11705 such as @file{startup.o} and @file{exit.o} are still included on the
11706 linker command line.
11710 @node PDP-11 Options
11711 @subsection PDP-11 Options
11712 @cindex PDP-11 Options
11714 These options are defined for the PDP-11:
11719 Use hardware FPP floating point. This is the default. (FIS floating
11720 point on the PDP-11/40 is not supported.)
11723 @opindex msoft-float
11724 Do not use hardware floating point.
11728 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11732 Return floating-point results in memory. This is the default.
11736 Generate code for a PDP-11/40.
11740 Generate code for a PDP-11/45. This is the default.
11744 Generate code for a PDP-11/10.
11746 @item -mbcopy-builtin
11747 @opindex bcopy-builtin
11748 Use inline @code{movmemhi} patterns for copying memory. This is the
11753 Do not use inline @code{movmemhi} patterns for copying memory.
11759 Use 16-bit @code{int}. This is the default.
11765 Use 32-bit @code{int}.
11768 @itemx -mno-float32
11770 @opindex mno-float32
11771 Use 64-bit @code{float}. This is the default.
11774 @itemx -mno-float64
11776 @opindex mno-float64
11777 Use 32-bit @code{float}.
11781 Use @code{abshi2} pattern. This is the default.
11785 Do not use @code{abshi2} pattern.
11787 @item -mbranch-expensive
11788 @opindex mbranch-expensive
11789 Pretend that branches are expensive. This is for experimenting with
11790 code generation only.
11792 @item -mbranch-cheap
11793 @opindex mbranch-cheap
11794 Do not pretend that branches are expensive. This is the default.
11798 Generate code for a system with split I&D@.
11802 Generate code for a system without split I&D@. This is the default.
11806 Use Unix assembler syntax. This is the default when configured for
11807 @samp{pdp11-*-bsd}.
11811 Use DEC assembler syntax. This is the default when configured for any
11812 PDP-11 target other than @samp{pdp11-*-bsd}.
11815 @node PowerPC Options
11816 @subsection PowerPC Options
11817 @cindex PowerPC options
11819 These are listed under @xref{RS/6000 and PowerPC Options}.
11821 @node RS/6000 and PowerPC Options
11822 @subsection IBM RS/6000 and PowerPC Options
11823 @cindex RS/6000 and PowerPC Options
11824 @cindex IBM RS/6000 and PowerPC Options
11826 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11833 @itemx -mno-powerpc
11834 @itemx -mpowerpc-gpopt
11835 @itemx -mno-powerpc-gpopt
11836 @itemx -mpowerpc-gfxopt
11837 @itemx -mno-powerpc-gfxopt
11839 @itemx -mno-powerpc64
11843 @itemx -mno-popcntb
11855 @opindex mno-power2
11857 @opindex mno-powerpc
11858 @opindex mpowerpc-gpopt
11859 @opindex mno-powerpc-gpopt
11860 @opindex mpowerpc-gfxopt
11861 @opindex mno-powerpc-gfxopt
11862 @opindex mpowerpc64
11863 @opindex mno-powerpc64
11867 @opindex mno-popcntb
11873 @opindex mno-mfpgpr
11876 GCC supports two related instruction set architectures for the
11877 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11878 instructions supported by the @samp{rios} chip set used in the original
11879 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11880 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11881 the IBM 4xx, 6xx, and follow-on microprocessors.
11883 Neither architecture is a subset of the other. However there is a
11884 large common subset of instructions supported by both. An MQ
11885 register is included in processors supporting the POWER architecture.
11887 You use these options to specify which instructions are available on the
11888 processor you are using. The default value of these options is
11889 determined when configuring GCC@. Specifying the
11890 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11891 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11892 rather than the options listed above.
11894 The @option{-mpower} option allows GCC to generate instructions that
11895 are found only in the POWER architecture and to use the MQ register.
11896 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11897 to generate instructions that are present in the POWER2 architecture but
11898 not the original POWER architecture.
11900 The @option{-mpowerpc} option allows GCC to generate instructions that
11901 are found only in the 32-bit subset of the PowerPC architecture.
11902 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11903 GCC to use the optional PowerPC architecture instructions in the
11904 General Purpose group, including floating-point square root. Specifying
11905 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11906 use the optional PowerPC architecture instructions in the Graphics
11907 group, including floating-point select.
11909 The @option{-mmfcrf} option allows GCC to generate the move from
11910 condition register field instruction implemented on the POWER4
11911 processor and other processors that support the PowerPC V2.01
11913 The @option{-mpopcntb} option allows GCC to generate the popcount and
11914 double precision FP reciprocal estimate instruction implemented on the
11915 POWER5 processor and other processors that support the PowerPC V2.02
11917 The @option{-mfprnd} option allows GCC to generate the FP round to
11918 integer instructions implemented on the POWER5+ processor and other
11919 processors that support the PowerPC V2.03 architecture.
11920 The @option{-mcmpb} option allows GCC to generate the compare bytes
11921 instruction implemented on the POWER6 processor and other processors
11922 that support the PowerPC V2.05 architecture.
11923 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
11924 general purpose register instructions implemented on the POWER6X
11925 processor and other processors that support the extended PowerPC V2.05
11927 The @option{-mdfp} option allows GCC to generate the decimal floating
11928 point instructions implemented on some POWER processors.
11930 The @option{-mpowerpc64} option allows GCC to generate the additional
11931 64-bit instructions that are found in the full PowerPC64 architecture
11932 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11933 @option{-mno-powerpc64}.
11935 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11936 will use only the instructions in the common subset of both
11937 architectures plus some special AIX common-mode calls, and will not use
11938 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11939 permits GCC to use any instruction from either architecture and to
11940 allow use of the MQ register; specify this for the Motorola MPC601.
11942 @item -mnew-mnemonics
11943 @itemx -mold-mnemonics
11944 @opindex mnew-mnemonics
11945 @opindex mold-mnemonics
11946 Select which mnemonics to use in the generated assembler code. With
11947 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11948 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11949 assembler mnemonics defined for the POWER architecture. Instructions
11950 defined in only one architecture have only one mnemonic; GCC uses that
11951 mnemonic irrespective of which of these options is specified.
11953 GCC defaults to the mnemonics appropriate for the architecture in
11954 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11955 value of these option. Unless you are building a cross-compiler, you
11956 should normally not specify either @option{-mnew-mnemonics} or
11957 @option{-mold-mnemonics}, but should instead accept the default.
11959 @item -mcpu=@var{cpu_type}
11961 Set architecture type, register usage, choice of mnemonics, and
11962 instruction scheduling parameters for machine type @var{cpu_type}.
11963 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11964 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11965 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11966 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11967 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11968 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11969 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11970 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11971 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
11972 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11974 @option{-mcpu=common} selects a completely generic processor. Code
11975 generated under this option will run on any POWER or PowerPC processor.
11976 GCC will use only the instructions in the common subset of both
11977 architectures, and will not use the MQ register. GCC assumes a generic
11978 processor model for scheduling purposes.
11980 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11981 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11982 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11983 types, with an appropriate, generic processor model assumed for
11984 scheduling purposes.
11986 The other options specify a specific processor. Code generated under
11987 those options will run best on that processor, and may not run at all on
11990 The @option{-mcpu} options automatically enable or disable the
11993 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
11994 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
11995 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
11997 The particular options set for any particular CPU will vary between
11998 compiler versions, depending on what setting seems to produce optimal
11999 code for that CPU; it doesn't necessarily reflect the actual hardware's
12000 capabilities. If you wish to set an individual option to a particular
12001 value, you may specify it after the @option{-mcpu} option, like
12002 @samp{-mcpu=970 -mno-altivec}.
12004 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
12005 not enabled or disabled by the @option{-mcpu} option at present because
12006 AIX does not have full support for these options. You may still
12007 enable or disable them individually if you're sure it'll work in your
12010 @item -mtune=@var{cpu_type}
12012 Set the instruction scheduling parameters for machine type
12013 @var{cpu_type}, but do not set the architecture type, register usage, or
12014 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
12015 values for @var{cpu_type} are used for @option{-mtune} as for
12016 @option{-mcpu}. If both are specified, the code generated will use the
12017 architecture, registers, and mnemonics set by @option{-mcpu}, but the
12018 scheduling parameters set by @option{-mtune}.
12024 Generate code to compute division as reciprocal estimate and iterative
12025 refinement, creating opportunities for increased throughput. This
12026 feature requires: optional PowerPC Graphics instruction set for single
12027 precision and FRE instruction for double precision, assuming divides
12028 cannot generate user-visible traps, and the domain values not include
12029 Infinities, denormals or zero denominator.
12032 @itemx -mno-altivec
12034 @opindex mno-altivec
12035 Generate code that uses (does not use) AltiVec instructions, and also
12036 enable the use of built-in functions that allow more direct access to
12037 the AltiVec instruction set. You may also need to set
12038 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
12044 @opindex mno-vrsave
12045 Generate VRSAVE instructions when generating AltiVec code.
12048 @opindex msecure-plt
12049 Generate code that allows ld and ld.so to build executables and shared
12050 libraries with non-exec .plt and .got sections. This is a PowerPC
12051 32-bit SYSV ABI option.
12055 Generate code that uses a BSS .plt section that ld.so fills in, and
12056 requires .plt and .got sections that are both writable and executable.
12057 This is a PowerPC 32-bit SYSV ABI option.
12063 This switch enables or disables the generation of ISEL instructions.
12065 @item -misel=@var{yes/no}
12066 This switch has been deprecated. Use @option{-misel} and
12067 @option{-mno-isel} instead.
12073 This switch enables or disables the generation of SPE simd
12076 @item -mspe=@var{yes/no}
12077 This option has been deprecated. Use @option{-mspe} and
12078 @option{-mno-spe} instead.
12080 @item -mfloat-gprs=@var{yes/single/double/no}
12081 @itemx -mfloat-gprs
12082 @opindex mfloat-gprs
12083 This switch enables or disables the generation of floating point
12084 operations on the general purpose registers for architectures that
12087 The argument @var{yes} or @var{single} enables the use of
12088 single-precision floating point operations.
12090 The argument @var{double} enables the use of single and
12091 double-precision floating point operations.
12093 The argument @var{no} disables floating point operations on the
12094 general purpose registers.
12096 This option is currently only available on the MPC854x.
12102 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12103 targets (including GNU/Linux). The 32-bit environment sets int, long
12104 and pointer to 32 bits and generates code that runs on any PowerPC
12105 variant. The 64-bit environment sets int to 32 bits and long and
12106 pointer to 64 bits, and generates code for PowerPC64, as for
12107 @option{-mpowerpc64}.
12110 @itemx -mno-fp-in-toc
12111 @itemx -mno-sum-in-toc
12112 @itemx -mminimal-toc
12114 @opindex mno-fp-in-toc
12115 @opindex mno-sum-in-toc
12116 @opindex mminimal-toc
12117 Modify generation of the TOC (Table Of Contents), which is created for
12118 every executable file. The @option{-mfull-toc} option is selected by
12119 default. In that case, GCC will allocate at least one TOC entry for
12120 each unique non-automatic variable reference in your program. GCC
12121 will also place floating-point constants in the TOC@. However, only
12122 16,384 entries are available in the TOC@.
12124 If you receive a linker error message that saying you have overflowed
12125 the available TOC space, you can reduce the amount of TOC space used
12126 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
12127 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
12128 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
12129 generate code to calculate the sum of an address and a constant at
12130 run-time instead of putting that sum into the TOC@. You may specify one
12131 or both of these options. Each causes GCC to produce very slightly
12132 slower and larger code at the expense of conserving TOC space.
12134 If you still run out of space in the TOC even when you specify both of
12135 these options, specify @option{-mminimal-toc} instead. This option causes
12136 GCC to make only one TOC entry for every file. When you specify this
12137 option, GCC will produce code that is slower and larger but which
12138 uses extremely little TOC space. You may wish to use this option
12139 only on files that contain less frequently executed code.
12145 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
12146 @code{long} type, and the infrastructure needed to support them.
12147 Specifying @option{-maix64} implies @option{-mpowerpc64} and
12148 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
12149 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
12152 @itemx -mno-xl-compat
12153 @opindex mxl-compat
12154 @opindex mno-xl-compat
12155 Produce code that conforms more closely to IBM XL compiler semantics
12156 when using AIX-compatible ABI. Pass floating-point arguments to
12157 prototyped functions beyond the register save area (RSA) on the stack
12158 in addition to argument FPRs. Do not assume that most significant
12159 double in 128-bit long double value is properly rounded when comparing
12160 values and converting to double. Use XL symbol names for long double
12163 The AIX calling convention was extended but not initially documented to
12164 handle an obscure K&R C case of calling a function that takes the
12165 address of its arguments with fewer arguments than declared. IBM XL
12166 compilers access floating point arguments which do not fit in the
12167 RSA from the stack when a subroutine is compiled without
12168 optimization. Because always storing floating-point arguments on the
12169 stack is inefficient and rarely needed, this option is not enabled by
12170 default and only is necessary when calling subroutines compiled by IBM
12171 XL compilers without optimization.
12175 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
12176 application written to use message passing with special startup code to
12177 enable the application to run. The system must have PE installed in the
12178 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
12179 must be overridden with the @option{-specs=} option to specify the
12180 appropriate directory location. The Parallel Environment does not
12181 support threads, so the @option{-mpe} option and the @option{-pthread}
12182 option are incompatible.
12184 @item -malign-natural
12185 @itemx -malign-power
12186 @opindex malign-natural
12187 @opindex malign-power
12188 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12189 @option{-malign-natural} overrides the ABI-defined alignment of larger
12190 types, such as floating-point doubles, on their natural size-based boundary.
12191 The option @option{-malign-power} instructs GCC to follow the ABI-specified
12192 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
12194 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
12198 @itemx -mhard-float
12199 @opindex msoft-float
12200 @opindex mhard-float
12201 Generate code that does not use (uses) the floating-point register set.
12202 Software floating point emulation is provided if you use the
12203 @option{-msoft-float} option, and pass the option to GCC when linking.
12206 @itemx -mno-multiple
12208 @opindex mno-multiple
12209 Generate code that uses (does not use) the load multiple word
12210 instructions and the store multiple word instructions. These
12211 instructions are generated by default on POWER systems, and not
12212 generated on PowerPC systems. Do not use @option{-mmultiple} on little
12213 endian PowerPC systems, since those instructions do not work when the
12214 processor is in little endian mode. The exceptions are PPC740 and
12215 PPC750 which permit the instructions usage in little endian mode.
12220 @opindex mno-string
12221 Generate code that uses (does not use) the load string instructions
12222 and the store string word instructions to save multiple registers and
12223 do small block moves. These instructions are generated by default on
12224 POWER systems, and not generated on PowerPC systems. Do not use
12225 @option{-mstring} on little endian PowerPC systems, since those
12226 instructions do not work when the processor is in little endian mode.
12227 The exceptions are PPC740 and PPC750 which permit the instructions
12228 usage in little endian mode.
12233 @opindex mno-update
12234 Generate code that uses (does not use) the load or store instructions
12235 that update the base register to the address of the calculated memory
12236 location. These instructions are generated by default. If you use
12237 @option{-mno-update}, there is a small window between the time that the
12238 stack pointer is updated and the address of the previous frame is
12239 stored, which means code that walks the stack frame across interrupts or
12240 signals may get corrupted data.
12243 @itemx -mno-fused-madd
12244 @opindex mfused-madd
12245 @opindex mno-fused-madd
12246 Generate code that uses (does not use) the floating point multiply and
12247 accumulate instructions. These instructions are generated by default if
12248 hardware floating is used.
12254 Generate code that uses (does not use) the half-word multiply and
12255 multiply-accumulate instructions on the IBM 405 and 440 processors.
12256 These instructions are generated by default when targetting those
12263 Generate code that uses (does not use) the string-search @samp{dlmzb}
12264 instruction on the IBM 405 and 440 processors. This instruction is
12265 generated by default when targetting those processors.
12267 @item -mno-bit-align
12269 @opindex mno-bit-align
12270 @opindex mbit-align
12271 On System V.4 and embedded PowerPC systems do not (do) force structures
12272 and unions that contain bit-fields to be aligned to the base type of the
12275 For example, by default a structure containing nothing but 8
12276 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
12277 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
12278 the structure would be aligned to a 1 byte boundary and be one byte in
12281 @item -mno-strict-align
12282 @itemx -mstrict-align
12283 @opindex mno-strict-align
12284 @opindex mstrict-align
12285 On System V.4 and embedded PowerPC systems do not (do) assume that
12286 unaligned memory references will be handled by the system.
12288 @item -mrelocatable
12289 @itemx -mno-relocatable
12290 @opindex mrelocatable
12291 @opindex mno-relocatable
12292 On embedded PowerPC systems generate code that allows (does not allow)
12293 the program to be relocated to a different address at runtime. If you
12294 use @option{-mrelocatable} on any module, all objects linked together must
12295 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
12297 @item -mrelocatable-lib
12298 @itemx -mno-relocatable-lib
12299 @opindex mrelocatable-lib
12300 @opindex mno-relocatable-lib
12301 On embedded PowerPC systems generate code that allows (does not allow)
12302 the program to be relocated to a different address at runtime. Modules
12303 compiled with @option{-mrelocatable-lib} can be linked with either modules
12304 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
12305 with modules compiled with the @option{-mrelocatable} options.
12311 On System V.4 and embedded PowerPC systems do not (do) assume that
12312 register 2 contains a pointer to a global area pointing to the addresses
12313 used in the program.
12316 @itemx -mlittle-endian
12318 @opindex mlittle-endian
12319 On System V.4 and embedded PowerPC systems compile code for the
12320 processor in little endian mode. The @option{-mlittle-endian} option is
12321 the same as @option{-mlittle}.
12324 @itemx -mbig-endian
12326 @opindex mbig-endian
12327 On System V.4 and embedded PowerPC systems compile code for the
12328 processor in big endian mode. The @option{-mbig-endian} option is
12329 the same as @option{-mbig}.
12331 @item -mdynamic-no-pic
12332 @opindex mdynamic-no-pic
12333 On Darwin and Mac OS X systems, compile code so that it is not
12334 relocatable, but that its external references are relocatable. The
12335 resulting code is suitable for applications, but not shared
12338 @item -mprioritize-restricted-insns=@var{priority}
12339 @opindex mprioritize-restricted-insns
12340 This option controls the priority that is assigned to
12341 dispatch-slot restricted instructions during the second scheduling
12342 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
12343 @var{no/highest/second-highest} priority to dispatch slot restricted
12346 @item -msched-costly-dep=@var{dependence_type}
12347 @opindex msched-costly-dep
12348 This option controls which dependences are considered costly
12349 by the target during instruction scheduling. The argument
12350 @var{dependence_type} takes one of the following values:
12351 @var{no}: no dependence is costly,
12352 @var{all}: all dependences are costly,
12353 @var{true_store_to_load}: a true dependence from store to load is costly,
12354 @var{store_to_load}: any dependence from store to load is costly,
12355 @var{number}: any dependence which latency >= @var{number} is costly.
12357 @item -minsert-sched-nops=@var{scheme}
12358 @opindex minsert-sched-nops
12359 This option controls which nop insertion scheme will be used during
12360 the second scheduling pass. The argument @var{scheme} takes one of the
12362 @var{no}: Don't insert nops.
12363 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
12364 according to the scheduler's grouping.
12365 @var{regroup_exact}: Insert nops to force costly dependent insns into
12366 separate groups. Insert exactly as many nops as needed to force an insn
12367 to a new group, according to the estimated processor grouping.
12368 @var{number}: Insert nops to force costly dependent insns into
12369 separate groups. Insert @var{number} nops to force an insn to a new group.
12372 @opindex mcall-sysv
12373 On System V.4 and embedded PowerPC systems compile code using calling
12374 conventions that adheres to the March 1995 draft of the System V
12375 Application Binary Interface, PowerPC processor supplement. This is the
12376 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
12378 @item -mcall-sysv-eabi
12379 @opindex mcall-sysv-eabi
12380 Specify both @option{-mcall-sysv} and @option{-meabi} options.
12382 @item -mcall-sysv-noeabi
12383 @opindex mcall-sysv-noeabi
12384 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
12386 @item -mcall-solaris
12387 @opindex mcall-solaris
12388 On System V.4 and embedded PowerPC systems compile code for the Solaris
12392 @opindex mcall-linux
12393 On System V.4 and embedded PowerPC systems compile code for the
12394 Linux-based GNU system.
12398 On System V.4 and embedded PowerPC systems compile code for the
12399 Hurd-based GNU system.
12401 @item -mcall-netbsd
12402 @opindex mcall-netbsd
12403 On System V.4 and embedded PowerPC systems compile code for the
12404 NetBSD operating system.
12406 @item -maix-struct-return
12407 @opindex maix-struct-return
12408 Return all structures in memory (as specified by the AIX ABI)@.
12410 @item -msvr4-struct-return
12411 @opindex msvr4-struct-return
12412 Return structures smaller than 8 bytes in registers (as specified by the
12415 @item -mabi=@var{abi-type}
12417 Extend the current ABI with a particular extension, or remove such extension.
12418 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
12419 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
12423 Extend the current ABI with SPE ABI extensions. This does not change
12424 the default ABI, instead it adds the SPE ABI extensions to the current
12428 @opindex mabi=no-spe
12429 Disable Booke SPE ABI extensions for the current ABI@.
12431 @item -mabi=ibmlongdouble
12432 @opindex mabi=ibmlongdouble
12433 Change the current ABI to use IBM extended precision long double.
12434 This is a PowerPC 32-bit SYSV ABI option.
12436 @item -mabi=ieeelongdouble
12437 @opindex mabi=ieeelongdouble
12438 Change the current ABI to use IEEE extended precision long double.
12439 This is a PowerPC 32-bit Linux ABI option.
12442 @itemx -mno-prototype
12443 @opindex mprototype
12444 @opindex mno-prototype
12445 On System V.4 and embedded PowerPC systems assume that all calls to
12446 variable argument functions are properly prototyped. Otherwise, the
12447 compiler must insert an instruction before every non prototyped call to
12448 set or clear bit 6 of the condition code register (@var{CR}) to
12449 indicate whether floating point values were passed in the floating point
12450 registers in case the function takes a variable arguments. With
12451 @option{-mprototype}, only calls to prototyped variable argument functions
12452 will set or clear the bit.
12456 On embedded PowerPC systems, assume that the startup module is called
12457 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
12458 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
12463 On embedded PowerPC systems, assume that the startup module is called
12464 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
12469 On embedded PowerPC systems, assume that the startup module is called
12470 @file{crt0.o} and the standard C libraries are @file{libads.a} and
12473 @item -myellowknife
12474 @opindex myellowknife
12475 On embedded PowerPC systems, assume that the startup module is called
12476 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
12481 On System V.4 and embedded PowerPC systems, specify that you are
12482 compiling for a VxWorks system.
12486 Specify that you are compiling for the WindISS simulation environment.
12490 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
12491 header to indicate that @samp{eabi} extended relocations are used.
12497 On System V.4 and embedded PowerPC systems do (do not) adhere to the
12498 Embedded Applications Binary Interface (eabi) which is a set of
12499 modifications to the System V.4 specifications. Selecting @option{-meabi}
12500 means that the stack is aligned to an 8 byte boundary, a function
12501 @code{__eabi} is called to from @code{main} to set up the eabi
12502 environment, and the @option{-msdata} option can use both @code{r2} and
12503 @code{r13} to point to two separate small data areas. Selecting
12504 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
12505 do not call an initialization function from @code{main}, and the
12506 @option{-msdata} option will only use @code{r13} to point to a single
12507 small data area. The @option{-meabi} option is on by default if you
12508 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
12511 @opindex msdata=eabi
12512 On System V.4 and embedded PowerPC systems, put small initialized
12513 @code{const} global and static data in the @samp{.sdata2} section, which
12514 is pointed to by register @code{r2}. Put small initialized
12515 non-@code{const} global and static data in the @samp{.sdata} section,
12516 which is pointed to by register @code{r13}. Put small uninitialized
12517 global and static data in the @samp{.sbss} section, which is adjacent to
12518 the @samp{.sdata} section. The @option{-msdata=eabi} option is
12519 incompatible with the @option{-mrelocatable} option. The
12520 @option{-msdata=eabi} option also sets the @option{-memb} option.
12523 @opindex msdata=sysv
12524 On System V.4 and embedded PowerPC systems, put small global and static
12525 data in the @samp{.sdata} section, which is pointed to by register
12526 @code{r13}. Put small uninitialized global and static data in the
12527 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
12528 The @option{-msdata=sysv} option is incompatible with the
12529 @option{-mrelocatable} option.
12531 @item -msdata=default
12533 @opindex msdata=default
12535 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12536 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12537 same as @option{-msdata=sysv}.
12540 @opindex msdata-data
12541 On System V.4 and embedded PowerPC systems, put small global
12542 data in the @samp{.sdata} section. Put small uninitialized global
12543 data in the @samp{.sbss} section. Do not use register @code{r13}
12544 to address small data however. This is the default behavior unless
12545 other @option{-msdata} options are used.
12549 @opindex msdata=none
12551 On embedded PowerPC systems, put all initialized global and static data
12552 in the @samp{.data} section, and all uninitialized data in the
12553 @samp{.bss} section.
12557 @cindex smaller data references (PowerPC)
12558 @cindex .sdata/.sdata2 references (PowerPC)
12559 On embedded PowerPC systems, put global and static items less than or
12560 equal to @var{num} bytes into the small data or bss sections instead of
12561 the normal data or bss section. By default, @var{num} is 8. The
12562 @option{-G @var{num}} switch is also passed to the linker.
12563 All modules should be compiled with the same @option{-G @var{num}} value.
12566 @itemx -mno-regnames
12568 @opindex mno-regnames
12569 On System V.4 and embedded PowerPC systems do (do not) emit register
12570 names in the assembly language output using symbolic forms.
12573 @itemx -mno-longcall
12575 @opindex mno-longcall
12576 By default assume that all calls are far away so that a longer more
12577 expensive calling sequence is required. This is required for calls
12578 further than 32 megabytes (33,554,432 bytes) from the current location.
12579 A short call will be generated if the compiler knows
12580 the call cannot be that far away. This setting can be overridden by
12581 the @code{shortcall} function attribute, or by @code{#pragma
12584 Some linkers are capable of detecting out-of-range calls and generating
12585 glue code on the fly. On these systems, long calls are unnecessary and
12586 generate slower code. As of this writing, the AIX linker can do this,
12587 as can the GNU linker for PowerPC/64. It is planned to add this feature
12588 to the GNU linker for 32-bit PowerPC systems as well.
12590 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
12591 callee, L42'', plus a ``branch island'' (glue code). The two target
12592 addresses represent the callee and the ``branch island''. The
12593 Darwin/PPC linker will prefer the first address and generate a ``bl
12594 callee'' if the PPC ``bl'' instruction will reach the callee directly;
12595 otherwise, the linker will generate ``bl L42'' to call the ``branch
12596 island''. The ``branch island'' is appended to the body of the
12597 calling function; it computes the full 32-bit address of the callee
12600 On Mach-O (Darwin) systems, this option directs the compiler emit to
12601 the glue for every direct call, and the Darwin linker decides whether
12602 to use or discard it.
12604 In the future, we may cause GCC to ignore all longcall specifications
12605 when the linker is known to generate glue.
12609 Adds support for multithreading with the @dfn{pthreads} library.
12610 This option sets flags for both the preprocessor and linker.
12614 @node S/390 and zSeries Options
12615 @subsection S/390 and zSeries Options
12616 @cindex S/390 and zSeries Options
12618 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
12622 @itemx -msoft-float
12623 @opindex mhard-float
12624 @opindex msoft-float
12625 Use (do not use) the hardware floating-point instructions and registers
12626 for floating-point operations. When @option{-msoft-float} is specified,
12627 functions in @file{libgcc.a} will be used to perform floating-point
12628 operations. When @option{-mhard-float} is specified, the compiler
12629 generates IEEE floating-point instructions. This is the default.
12631 @item -mlong-double-64
12632 @itemx -mlong-double-128
12633 @opindex mlong-double-64
12634 @opindex mlong-double-128
12635 These switches control the size of @code{long double} type. A size
12636 of 64bit makes the @code{long double} type equivalent to the @code{double}
12637 type. This is the default.
12640 @itemx -mno-backchain
12641 @opindex mbackchain
12642 @opindex mno-backchain
12643 Store (do not store) the address of the caller's frame as backchain pointer
12644 into the callee's stack frame.
12645 A backchain may be needed to allow debugging using tools that do not understand
12646 DWARF-2 call frame information.
12647 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12648 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12649 the backchain is placed into the topmost word of the 96/160 byte register
12652 In general, code compiled with @option{-mbackchain} is call-compatible with
12653 code compiled with @option{-mmo-backchain}; however, use of the backchain
12654 for debugging purposes usually requires that the whole binary is built with
12655 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12656 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12657 to build a linux kernel use @option{-msoft-float}.
12659 The default is to not maintain the backchain.
12661 @item -mpacked-stack
12662 @item -mno-packed-stack
12663 @opindex mpacked-stack
12664 @opindex mno-packed-stack
12665 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12666 specified, the compiler uses the all fields of the 96/160 byte register save
12667 area only for their default purpose; unused fields still take up stack space.
12668 When @option{-mpacked-stack} is specified, register save slots are densely
12669 packed at the top of the register save area; unused space is reused for other
12670 purposes, allowing for more efficient use of the available stack space.
12671 However, when @option{-mbackchain} is also in effect, the topmost word of
12672 the save area is always used to store the backchain, and the return address
12673 register is always saved two words below the backchain.
12675 As long as the stack frame backchain is not used, code generated with
12676 @option{-mpacked-stack} is call-compatible with code generated with
12677 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12678 S/390 or zSeries generated code that uses the stack frame backchain at run
12679 time, not just for debugging purposes. Such code is not call-compatible
12680 with code compiled with @option{-mpacked-stack}. Also, note that the
12681 combination of @option{-mbackchain},
12682 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12683 to build a linux kernel use @option{-msoft-float}.
12685 The default is to not use the packed stack layout.
12688 @itemx -mno-small-exec
12689 @opindex msmall-exec
12690 @opindex mno-small-exec
12691 Generate (or do not generate) code using the @code{bras} instruction
12692 to do subroutine calls.
12693 This only works reliably if the total executable size does not
12694 exceed 64k. The default is to use the @code{basr} instruction instead,
12695 which does not have this limitation.
12701 When @option{-m31} is specified, generate code compliant to the
12702 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12703 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12704 particular to generate 64-bit instructions. For the @samp{s390}
12705 targets, the default is @option{-m31}, while the @samp{s390x}
12706 targets default to @option{-m64}.
12712 When @option{-mzarch} is specified, generate code using the
12713 instructions available on z/Architecture.
12714 When @option{-mesa} is specified, generate code using the
12715 instructions available on ESA/390. Note that @option{-mesa} is
12716 not possible with @option{-m64}.
12717 When generating code compliant to the GNU/Linux for S/390 ABI,
12718 the default is @option{-mesa}. When generating code compliant
12719 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12725 Generate (or do not generate) code using the @code{mvcle} instruction
12726 to perform block moves. When @option{-mno-mvcle} is specified,
12727 use a @code{mvc} loop instead. This is the default unless optimizing for
12734 Print (or do not print) additional debug information when compiling.
12735 The default is to not print debug information.
12737 @item -march=@var{cpu-type}
12739 Generate code that will run on @var{cpu-type}, which is the name of a system
12740 representing a certain processor type. Possible values for
12741 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12742 When generating code using the instructions available on z/Architecture,
12743 the default is @option{-march=z900}. Otherwise, the default is
12744 @option{-march=g5}.
12746 @item -mtune=@var{cpu-type}
12748 Tune to @var{cpu-type} everything applicable about the generated code,
12749 except for the ABI and the set of available instructions.
12750 The list of @var{cpu-type} values is the same as for @option{-march}.
12751 The default is the value used for @option{-march}.
12754 @itemx -mno-tpf-trace
12755 @opindex mtpf-trace
12756 @opindex mno-tpf-trace
12757 Generate code that adds (does not add) in TPF OS specific branches to trace
12758 routines in the operating system. This option is off by default, even
12759 when compiling for the TPF OS@.
12762 @itemx -mno-fused-madd
12763 @opindex mfused-madd
12764 @opindex mno-fused-madd
12765 Generate code that uses (does not use) the floating point multiply and
12766 accumulate instructions. These instructions are generated by default if
12767 hardware floating point is used.
12769 @item -mwarn-framesize=@var{framesize}
12770 @opindex mwarn-framesize
12771 Emit a warning if the current function exceeds the given frame size. Because
12772 this is a compile time check it doesn't need to be a real problem when the program
12773 runs. It is intended to identify functions which most probably cause
12774 a stack overflow. It is useful to be used in an environment with limited stack
12775 size e.g.@: the linux kernel.
12777 @item -mwarn-dynamicstack
12778 @opindex mwarn-dynamicstack
12779 Emit a warning if the function calls alloca or uses dynamically
12780 sized arrays. This is generally a bad idea with a limited stack size.
12782 @item -mstack-guard=@var{stack-guard}
12783 @item -mstack-size=@var{stack-size}
12784 @opindex mstack-guard
12785 @opindex mstack-size
12786 These arguments always have to be used in conjunction. If they are present the s390
12787 back end emits additional instructions in the function prologue which trigger a trap
12788 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12789 (remember that the stack on s390 grows downward). These options are intended to
12790 be used to help debugging stack overflow problems. The additionally emitted code
12791 causes only little overhead and hence can also be used in production like systems
12792 without greater performance degradation. The given values have to be exact
12793 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12795 In order to be efficient the extra code makes the assumption that the stack starts
12796 at an address aligned to the value given by @var{stack-size}.
12799 @node Score Options
12800 @subsection Score Options
12801 @cindex Score Options
12803 These options are defined for Score implementations:
12808 Compile code for little endian mode.
12812 Compile code for big endian mode. This is the default.
12816 Enable the use of multiply-accumulate instructions. Disabled by default.
12820 Specify the SCORE5U of the target architecture.
12824 Specify the SCORE7 of the target architecture. This is the default.
12828 @subsection SH Options
12830 These @samp{-m} options are defined for the SH implementations:
12835 Generate code for the SH1.
12839 Generate code for the SH2.
12842 Generate code for the SH2e.
12846 Generate code for the SH3.
12850 Generate code for the SH3e.
12854 Generate code for the SH4 without a floating-point unit.
12856 @item -m4-single-only
12857 @opindex m4-single-only
12858 Generate code for the SH4 with a floating-point unit that only
12859 supports single-precision arithmetic.
12863 Generate code for the SH4 assuming the floating-point unit is in
12864 single-precision mode by default.
12868 Generate code for the SH4.
12872 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12873 floating-point unit is not used.
12875 @item -m4a-single-only
12876 @opindex m4a-single-only
12877 Generate code for the SH4a, in such a way that no double-precision
12878 floating point operations are used.
12881 @opindex m4a-single
12882 Generate code for the SH4a assuming the floating-point unit is in
12883 single-precision mode by default.
12887 Generate code for the SH4a.
12891 Same as @option{-m4a-nofpu}, except that it implicitly passes
12892 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12893 instructions at the moment.
12897 Compile code for the processor in big endian mode.
12901 Compile code for the processor in little endian mode.
12905 Align doubles at 64-bit boundaries. Note that this changes the calling
12906 conventions, and thus some functions from the standard C library will
12907 not work unless you recompile it first with @option{-mdalign}.
12911 Shorten some address references at link time, when possible; uses the
12912 linker option @option{-relax}.
12916 Use 32-bit offsets in @code{switch} tables. The default is to use
12921 Enable the use of the instruction @code{fmovd}.
12925 Comply with the calling conventions defined by Renesas.
12929 Comply with the calling conventions defined by Renesas.
12933 Comply with the calling conventions defined for GCC before the Renesas
12934 conventions were available. This option is the default for all
12935 targets of the SH toolchain except for @samp{sh-symbianelf}.
12938 @opindex mnomacsave
12939 Mark the @code{MAC} register as call-clobbered, even if
12940 @option{-mhitachi} is given.
12944 Increase IEEE-compliance of floating-point code.
12945 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12946 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12947 comparisons of NANs / infinities incurs extra overhead in every
12948 floating point comparison, therefore the default is set to
12949 @option{-ffinite-math-only}.
12951 @item -minline-ic_invalidate
12952 @opindex minline-ic_invalidate
12953 Inline code to invalidate instruction cache entries after setting up
12954 nested function trampolines.
12955 This option has no effect if -musermode is in effect and the selected
12956 code generation option (e.g. -m4) does not allow the use of the icbi
12958 If the selected code generation option does not allow the use of the icbi
12959 instruction, and -musermode is not in effect, the inlined code will
12960 manipulate the instruction cache address array directly with an associative
12961 write. This not only requires privileged mode, but it will also
12962 fail if the cache line had been mapped via the TLB and has become unmapped.
12966 Dump instruction size and location in the assembly code.
12969 @opindex mpadstruct
12970 This option is deprecated. It pads structures to multiple of 4 bytes,
12971 which is incompatible with the SH ABI@.
12975 Optimize for space instead of speed. Implied by @option{-Os}.
12978 @opindex mprefergot
12979 When generating position-independent code, emit function calls using
12980 the Global Offset Table instead of the Procedure Linkage Table.
12984 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
12985 if the inlined code would not work in user mode.
12986 This is the default when the target is @code{sh-*-linux*}.
12988 @item -multcost=@var{number}
12989 @opindex multcost=@var{number}
12990 Set the cost to assume for a multiply insn.
12992 @item -mdiv=@var{strategy}
12993 @opindex mdiv=@var{strategy}
12994 Set the division strategy to use for SHmedia code. @var{strategy} must be
12995 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12996 inv:call2, inv:fp .
12997 "fp" performs the operation in floating point. This has a very high latency,
12998 but needs only a few instructions, so it might be a good choice if
12999 your code has enough easily exploitable ILP to allow the compiler to
13000 schedule the floating point instructions together with other instructions.
13001 Division by zero causes a floating point exception.
13002 "inv" uses integer operations to calculate the inverse of the divisor,
13003 and then multiplies the dividend with the inverse. This strategy allows
13004 cse and hoisting of the inverse calculation. Division by zero calculates
13005 an unspecified result, but does not trap.
13006 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
13007 have been found, or if the entire operation has been hoisted to the same
13008 place, the last stages of the inverse calculation are intertwined with the
13009 final multiply to reduce the overall latency, at the expense of using a few
13010 more instructions, and thus offering fewer scheduling opportunities with
13012 "call" calls a library function that usually implements the inv:minlat
13014 This gives high code density for m5-*media-nofpu compilations.
13015 "call2" uses a different entry point of the same library function, where it
13016 assumes that a pointer to a lookup table has already been set up, which
13017 exposes the pointer load to cse / code hoisting optimizations.
13018 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
13019 code generation, but if the code stays unoptimized, revert to the "call",
13020 "call2", or "fp" strategies, respectively. Note that the
13021 potentially-trapping side effect of division by zero is carried by a
13022 separate instruction, so it is possible that all the integer instructions
13023 are hoisted out, but the marker for the side effect stays where it is.
13024 A recombination to fp operations or a call is not possible in that case.
13025 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
13026 that the inverse calculation was nor separated from the multiply, they speed
13027 up division where the dividend fits into 20 bits (plus sign where applicable),
13028 by inserting a test to skip a number of operations in this case; this test
13029 slows down the case of larger dividends. inv20u assumes the case of a such
13030 a small dividend to be unlikely, and inv20l assumes it to be likely.
13032 @item -mdivsi3_libfunc=@var{name}
13033 @opindex mdivsi3_libfunc=@var{name}
13034 Set the name of the library function used for 32 bit signed division to
13035 @var{name}. This only affect the name used in the call and inv:call
13036 division strategies, and the compiler will still expect the same
13037 sets of input/output/clobbered registers as if this option was not present.
13039 @item -madjust-unroll
13040 @opindex madjust-unroll
13041 Throttle unrolling to avoid thrashing target registers.
13042 This option only has an effect if the gcc code base supports the
13043 TARGET_ADJUST_UNROLL_MAX target hook.
13045 @item -mindexed-addressing
13046 @opindex mindexed-addressing
13047 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
13048 This is only safe if the hardware and/or OS implement 32 bit wrap-around
13049 semantics for the indexed addressing mode. The architecture allows the
13050 implementation of processors with 64 bit MMU, which the OS could use to
13051 get 32 bit addressing, but since no current hardware implementation supports
13052 this or any other way to make the indexed addressing mode safe to use in
13053 the 32 bit ABI, the default is -mno-indexed-addressing.
13055 @item -mgettrcost=@var{number}
13056 @opindex mgettrcost=@var{number}
13057 Set the cost assumed for the gettr instruction to @var{number}.
13058 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
13062 Assume pt* instructions won't trap. This will generally generate better
13063 scheduled code, but is unsafe on current hardware. The current architecture
13064 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
13065 This has the unintentional effect of making it unsafe to schedule ptabs /
13066 ptrel before a branch, or hoist it out of a loop. For example,
13067 __do_global_ctors, a part of libgcc that runs constructors at program
13068 startup, calls functions in a list which is delimited by @minus{}1. With the
13069 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
13070 That means that all the constructors will be run a bit quicker, but when
13071 the loop comes to the end of the list, the program crashes because ptabs
13072 loads @minus{}1 into a target register. Since this option is unsafe for any
13073 hardware implementing the current architecture specification, the default
13074 is -mno-pt-fixed. Unless the user specifies a specific cost with
13075 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
13076 this deters register allocation using target registers for storing
13079 @item -minvalid-symbols
13080 @opindex minvalid-symbols
13081 Assume symbols might be invalid. Ordinary function symbols generated by
13082 the compiler will always be valid to load with movi/shori/ptabs or
13083 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
13084 to generate symbols that will cause ptabs / ptrel to trap.
13085 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
13086 It will then prevent cross-basic-block cse, hoisting and most scheduling
13087 of symbol loads. The default is @option{-mno-invalid-symbols}.
13090 @node SPARC Options
13091 @subsection SPARC Options
13092 @cindex SPARC options
13094 These @samp{-m} options are supported on the SPARC:
13097 @item -mno-app-regs
13099 @opindex mno-app-regs
13101 Specify @option{-mapp-regs} to generate output using the global registers
13102 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
13105 To be fully SVR4 ABI compliant at the cost of some performance loss,
13106 specify @option{-mno-app-regs}. You should compile libraries and system
13107 software with this option.
13110 @itemx -mhard-float
13112 @opindex mhard-float
13113 Generate output containing floating point instructions. This is the
13117 @itemx -msoft-float
13119 @opindex msoft-float
13120 Generate output containing library calls for floating point.
13121 @strong{Warning:} the requisite libraries are not available for all SPARC
13122 targets. Normally the facilities of the machine's usual C compiler are
13123 used, but this cannot be done directly in cross-compilation. You must make
13124 your own arrangements to provide suitable library functions for
13125 cross-compilation. The embedded targets @samp{sparc-*-aout} and
13126 @samp{sparclite-*-*} do provide software floating point support.
13128 @option{-msoft-float} changes the calling convention in the output file;
13129 therefore, it is only useful if you compile @emph{all} of a program with
13130 this option. In particular, you need to compile @file{libgcc.a}, the
13131 library that comes with GCC, with @option{-msoft-float} in order for
13134 @item -mhard-quad-float
13135 @opindex mhard-quad-float
13136 Generate output containing quad-word (long double) floating point
13139 @item -msoft-quad-float
13140 @opindex msoft-quad-float
13141 Generate output containing library calls for quad-word (long double)
13142 floating point instructions. The functions called are those specified
13143 in the SPARC ABI@. This is the default.
13145 As of this writing, there are no SPARC implementations that have hardware
13146 support for the quad-word floating point instructions. They all invoke
13147 a trap handler for one of these instructions, and then the trap handler
13148 emulates the effect of the instruction. Because of the trap handler overhead,
13149 this is much slower than calling the ABI library routines. Thus the
13150 @option{-msoft-quad-float} option is the default.
13152 @item -mno-unaligned-doubles
13153 @itemx -munaligned-doubles
13154 @opindex mno-unaligned-doubles
13155 @opindex munaligned-doubles
13156 Assume that doubles have 8 byte alignment. This is the default.
13158 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
13159 alignment only if they are contained in another type, or if they have an
13160 absolute address. Otherwise, it assumes they have 4 byte alignment.
13161 Specifying this option avoids some rare compatibility problems with code
13162 generated by other compilers. It is not the default because it results
13163 in a performance loss, especially for floating point code.
13165 @item -mno-faster-structs
13166 @itemx -mfaster-structs
13167 @opindex mno-faster-structs
13168 @opindex mfaster-structs
13169 With @option{-mfaster-structs}, the compiler assumes that structures
13170 should have 8 byte alignment. This enables the use of pairs of
13171 @code{ldd} and @code{std} instructions for copies in structure
13172 assignment, in place of twice as many @code{ld} and @code{st} pairs.
13173 However, the use of this changed alignment directly violates the SPARC
13174 ABI@. Thus, it's intended only for use on targets where the developer
13175 acknowledges that their resulting code will not be directly in line with
13176 the rules of the ABI@.
13178 @item -mimpure-text
13179 @opindex mimpure-text
13180 @option{-mimpure-text}, used in addition to @option{-shared}, tells
13181 the compiler to not pass @option{-z text} to the linker when linking a
13182 shared object. Using this option, you can link position-dependent
13183 code into a shared object.
13185 @option{-mimpure-text} suppresses the ``relocations remain against
13186 allocatable but non-writable sections'' linker error message.
13187 However, the necessary relocations will trigger copy-on-write, and the
13188 shared object is not actually shared across processes. Instead of
13189 using @option{-mimpure-text}, you should compile all source code with
13190 @option{-fpic} or @option{-fPIC}.
13192 This option is only available on SunOS and Solaris.
13194 @item -mcpu=@var{cpu_type}
13196 Set the instruction set, register set, and instruction scheduling parameters
13197 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
13198 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
13199 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
13200 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
13201 @samp{ultrasparc3}, and @samp{niagara}.
13203 Default instruction scheduling parameters are used for values that select
13204 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
13205 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
13207 Here is a list of each supported architecture and their supported
13212 v8: supersparc, hypersparc
13213 sparclite: f930, f934, sparclite86x
13215 v9: ultrasparc, ultrasparc3, niagara
13218 By default (unless configured otherwise), GCC generates code for the V7
13219 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
13220 additionally optimizes it for the Cypress CY7C602 chip, as used in the
13221 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
13222 SPARCStation 1, 2, IPX etc.
13224 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
13225 architecture. The only difference from V7 code is that the compiler emits
13226 the integer multiply and integer divide instructions which exist in SPARC-V8
13227 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
13228 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
13231 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
13232 the SPARC architecture. This adds the integer multiply, integer divide step
13233 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
13234 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
13235 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
13236 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
13237 MB86934 chip, which is the more recent SPARClite with FPU@.
13239 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
13240 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
13241 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
13242 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
13243 optimizes it for the TEMIC SPARClet chip.
13245 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
13246 architecture. This adds 64-bit integer and floating-point move instructions,
13247 3 additional floating-point condition code registers and conditional move
13248 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
13249 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
13250 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
13251 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
13252 @option{-mcpu=niagara}, the compiler additionally optimizes it for
13253 Sun UltraSPARC T1 chips.
13255 @item -mtune=@var{cpu_type}
13257 Set the instruction scheduling parameters for machine type
13258 @var{cpu_type}, but do not set the instruction set or register set that the
13259 option @option{-mcpu=@var{cpu_type}} would.
13261 The same values for @option{-mcpu=@var{cpu_type}} can be used for
13262 @option{-mtune=@var{cpu_type}}, but the only useful values are those
13263 that select a particular cpu implementation. Those are @samp{cypress},
13264 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
13265 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
13266 @samp{ultrasparc3}, and @samp{niagara}.
13271 @opindex mno-v8plus
13272 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
13273 difference from the V8 ABI is that the global and out registers are
13274 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
13275 mode for all SPARC-V9 processors.
13281 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
13282 Visual Instruction Set extensions. The default is @option{-mno-vis}.
13285 These @samp{-m} options are supported in addition to the above
13286 on SPARC-V9 processors in 64-bit environments:
13289 @item -mlittle-endian
13290 @opindex mlittle-endian
13291 Generate code for a processor running in little-endian mode. It is only
13292 available for a few configurations and most notably not on Solaris and Linux.
13298 Generate code for a 32-bit or 64-bit environment.
13299 The 32-bit environment sets int, long and pointer to 32 bits.
13300 The 64-bit environment sets int to 32 bits and long and pointer
13303 @item -mcmodel=medlow
13304 @opindex mcmodel=medlow
13305 Generate code for the Medium/Low code model: 64-bit addresses, programs
13306 must be linked in the low 32 bits of memory. Programs can be statically
13307 or dynamically linked.
13309 @item -mcmodel=medmid
13310 @opindex mcmodel=medmid
13311 Generate code for the Medium/Middle code model: 64-bit addresses, programs
13312 must be linked in the low 44 bits of memory, the text and data segments must
13313 be less than 2GB in size and the data segment must be located within 2GB of
13316 @item -mcmodel=medany
13317 @opindex mcmodel=medany
13318 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
13319 may be linked anywhere in memory, the text and data segments must be less
13320 than 2GB in size and the data segment must be located within 2GB of the
13323 @item -mcmodel=embmedany
13324 @opindex mcmodel=embmedany
13325 Generate code for the Medium/Anywhere code model for embedded systems:
13326 64-bit addresses, the text and data segments must be less than 2GB in
13327 size, both starting anywhere in memory (determined at link time). The
13328 global register %g4 points to the base of the data segment. Programs
13329 are statically linked and PIC is not supported.
13332 @itemx -mno-stack-bias
13333 @opindex mstack-bias
13334 @opindex mno-stack-bias
13335 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
13336 frame pointer if present, are offset by @minus{}2047 which must be added back
13337 when making stack frame references. This is the default in 64-bit mode.
13338 Otherwise, assume no such offset is present.
13341 These switches are supported in addition to the above on Solaris:
13346 Add support for multithreading using the Solaris threads library. This
13347 option sets flags for both the preprocessor and linker. This option does
13348 not affect the thread safety of object code produced by the compiler or
13349 that of libraries supplied with it.
13353 Add support for multithreading using the POSIX threads library. This
13354 option sets flags for both the preprocessor and linker. This option does
13355 not affect the thread safety of object code produced by the compiler or
13356 that of libraries supplied with it.
13360 This is a synonym for @option{-pthreads}.
13364 @subsection SPU Options
13365 @cindex SPU options
13367 These @samp{-m} options are supported on the SPU:
13371 @itemx -merror-reloc
13372 @opindex mwarn-reloc
13373 @opindex merror-reloc
13375 The loader for SPU does not handle dynamic relocations. By default, GCC
13376 will give an error when it generates code that requires a dynamic
13377 relocation. @option{-mno-error-reloc} disables the error,
13378 @option{-mwarn-reloc} will generate a warning instead.
13381 @itemx -munsafe-dma
13383 @opindex munsafe-dma
13385 Instructions which initiate or test completion of DMA must not be
13386 reordered with respect to loads and stores of the memory which is being
13387 accessed. Users typically address this problem using the volatile
13388 keyword, but that can lead to inefficient code in places where the
13389 memory is known to not change. Rather than mark the memory as volatile
13390 we treat the DMA instructions as potentially effecting all memory. With
13391 @option{-munsafe-dma} users must use the volatile keyword to protect
13394 @item -mbranch-hints
13395 @opindex mbranch-hints
13397 By default, GCC will generate a branch hint instruction to avoid
13398 pipeline stalls for always taken or probably taken branches. A hint
13399 will not be generated closer than 8 instructions away from its branch.
13400 There is little reason to disable them, except for debugging purposes,
13401 or to make an object a little bit smaller.
13405 @opindex msmall-mem
13406 @opindex mlarge-mem
13408 By default, GCC generates code assuming that addresses are never larger
13409 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
13410 a full 32 bit address.
13415 By default, GCC links against startup code that assumes the SPU-style
13416 main function interface (which has an unconventional parameter list).
13417 With @option{-mstdmain}, GCC will link your program against startup
13418 code that assumes a C99-style interface to @code{main}, including a
13419 local copy of @code{argv} strings.
13421 @item -mfixed-range=@var{register-range}
13422 @opindex mfixed-range
13423 Generate code treating the given register range as fixed registers.
13424 A fixed register is one that the register allocator can not use. This is
13425 useful when compiling kernel code. A register range is specified as
13426 two registers separated by a dash. Multiple register ranges can be
13427 specified separated by a comma.
13431 @node System V Options
13432 @subsection Options for System V
13434 These additional options are available on System V Release 4 for
13435 compatibility with other compilers on those systems:
13440 Create a shared object.
13441 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
13445 Identify the versions of each tool used by the compiler, in a
13446 @code{.ident} assembler directive in the output.
13450 Refrain from adding @code{.ident} directives to the output file (this is
13453 @item -YP,@var{dirs}
13455 Search the directories @var{dirs}, and no others, for libraries
13456 specified with @option{-l}.
13458 @item -Ym,@var{dir}
13460 Look in the directory @var{dir} to find the M4 preprocessor.
13461 The assembler uses this option.
13462 @c This is supposed to go with a -Yd for predefined M4 macro files, but
13463 @c the generic assembler that comes with Solaris takes just -Ym.
13466 @node TMS320C3x/C4x Options
13467 @subsection TMS320C3x/C4x Options
13468 @cindex TMS320C3x/C4x Options
13470 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
13474 @item -mcpu=@var{cpu_type}
13476 Set the instruction set, register set, and instruction scheduling
13477 parameters for machine type @var{cpu_type}. Supported values for
13478 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
13479 @samp{c44}. The default is @samp{c40} to generate code for the
13484 @itemx -msmall-memory
13486 @opindex mbig-memory
13488 @opindex msmall-memory
13490 Generates code for the big or small memory model. The small memory
13491 model assumed that all data fits into one 64K word page. At run-time
13492 the data page (DP) register must be set to point to the 64K page
13493 containing the .bss and .data program sections. The big memory model is
13494 the default and requires reloading of the DP register for every direct
13501 Allow (disallow) allocation of general integer operands into the block
13502 count register BK@.
13508 Enable (disable) generation of code using decrement and branch,
13509 DBcond(D), instructions. This is enabled by default for the C4x. To be
13510 on the safe side, this is disabled for the C3x, since the maximum
13511 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
13512 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
13513 that it can utilize the decrement and branch instruction, but will give
13514 up if there is more than one memory reference in the loop. Thus a loop
13515 where the loop counter is decremented can generate slightly more
13516 efficient code, in cases where the RPTB instruction cannot be utilized.
13518 @item -mdp-isr-reload
13520 @opindex mdp-isr-reload
13522 Force the DP register to be saved on entry to an interrupt service
13523 routine (ISR), reloaded to point to the data section, and restored on
13524 exit from the ISR@. This should not be required unless someone has
13525 violated the small memory model by modifying the DP register, say within
13532 For the C3x use the 24-bit MPYI instruction for integer multiplies
13533 instead of a library call to guarantee 32-bit results. Note that if one
13534 of the operands is a constant, then the multiplication will be performed
13535 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
13536 then squaring operations are performed inline instead of a library call.
13539 @itemx -mno-fast-fix
13541 @opindex mno-fast-fix
13542 The C3x/C4x FIX instruction to convert a floating point value to an
13543 integer value chooses the nearest integer less than or equal to the
13544 floating point value rather than to the nearest integer. Thus if the
13545 floating point number is negative, the result will be incorrectly
13546 truncated an additional code is necessary to detect and correct this
13547 case. This option can be used to disable generation of the additional
13548 code required to correct the result.
13554 Enable (disable) generation of repeat block sequences using the RPTB
13555 instruction for zero overhead looping. The RPTB construct is only used
13556 for innermost loops that do not call functions or jump across the loop
13557 boundaries. There is no advantage having nested RPTB loops due to the
13558 overhead required to save and restore the RC, RS, and RE registers.
13559 This is enabled by default with @option{-O2}.
13561 @item -mrpts=@var{count}
13565 Enable (disable) the use of the single instruction repeat instruction
13566 RPTS@. If a repeat block contains a single instruction, and the loop
13567 count can be guaranteed to be less than the value @var{count}, GCC will
13568 emit a RPTS instruction instead of a RPTB@. If no value is specified,
13569 then a RPTS will be emitted even if the loop count cannot be determined
13570 at compile time. Note that the repeated instruction following RPTS does
13571 not have to be reloaded from memory each iteration, thus freeing up the
13572 CPU buses for operands. However, since interrupts are blocked by this
13573 instruction, it is disabled by default.
13575 @item -mloop-unsigned
13576 @itemx -mno-loop-unsigned
13577 @opindex mloop-unsigned
13578 @opindex mno-loop-unsigned
13579 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
13580 is @math{2^{31} + 1} since these instructions test if the iteration count is
13581 negative to terminate the loop. If the iteration count is unsigned
13582 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
13583 exceeded. This switch allows an unsigned iteration count.
13587 Try to emit an assembler syntax that the TI assembler (asm30) is happy
13588 with. This also enforces compatibility with the API employed by the TI
13589 C3x C compiler. For example, long doubles are passed as structures
13590 rather than in floating point registers.
13596 Generate code that uses registers (stack) for passing arguments to functions.
13597 By default, arguments are passed in registers where possible rather
13598 than by pushing arguments on to the stack.
13600 @item -mparallel-insns
13601 @itemx -mno-parallel-insns
13602 @opindex mparallel-insns
13603 @opindex mno-parallel-insns
13604 Allow the generation of parallel instructions. This is enabled by
13605 default with @option{-O2}.
13607 @item -mparallel-mpy
13608 @itemx -mno-parallel-mpy
13609 @opindex mparallel-mpy
13610 @opindex mno-parallel-mpy
13611 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
13612 provided @option{-mparallel-insns} is also specified. These instructions have
13613 tight register constraints which can pessimize the code generation
13614 of large functions.
13619 @subsection V850 Options
13620 @cindex V850 Options
13622 These @samp{-m} options are defined for V850 implementations:
13626 @itemx -mno-long-calls
13627 @opindex mlong-calls
13628 @opindex mno-long-calls
13629 Treat all calls as being far away (near). If calls are assumed to be
13630 far away, the compiler will always load the functions address up into a
13631 register, and call indirect through the pointer.
13637 Do not optimize (do optimize) basic blocks that use the same index
13638 pointer 4 or more times to copy pointer into the @code{ep} register, and
13639 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
13640 option is on by default if you optimize.
13642 @item -mno-prolog-function
13643 @itemx -mprolog-function
13644 @opindex mno-prolog-function
13645 @opindex mprolog-function
13646 Do not use (do use) external functions to save and restore registers
13647 at the prologue and epilogue of a function. The external functions
13648 are slower, but use less code space if more than one function saves
13649 the same number of registers. The @option{-mprolog-function} option
13650 is on by default if you optimize.
13654 Try to make the code as small as possible. At present, this just turns
13655 on the @option{-mep} and @option{-mprolog-function} options.
13657 @item -mtda=@var{n}
13659 Put static or global variables whose size is @var{n} bytes or less into
13660 the tiny data area that register @code{ep} points to. The tiny data
13661 area can hold up to 256 bytes in total (128 bytes for byte references).
13663 @item -msda=@var{n}
13665 Put static or global variables whose size is @var{n} bytes or less into
13666 the small data area that register @code{gp} points to. The small data
13667 area can hold up to 64 kilobytes.
13669 @item -mzda=@var{n}
13671 Put static or global variables whose size is @var{n} bytes or less into
13672 the first 32 kilobytes of memory.
13676 Specify that the target processor is the V850.
13679 @opindex mbig-switch
13680 Generate code suitable for big switch tables. Use this option only if
13681 the assembler/linker complain about out of range branches within a switch
13686 This option will cause r2 and r5 to be used in the code generated by
13687 the compiler. This setting is the default.
13689 @item -mno-app-regs
13690 @opindex mno-app-regs
13691 This option will cause r2 and r5 to be treated as fixed registers.
13695 Specify that the target processor is the V850E1. The preprocessor
13696 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13697 this option is used.
13701 Specify that the target processor is the V850E@. The preprocessor
13702 constant @samp{__v850e__} will be defined if this option is used.
13704 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13705 are defined then a default target processor will be chosen and the
13706 relevant @samp{__v850*__} preprocessor constant will be defined.
13708 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13709 defined, regardless of which processor variant is the target.
13711 @item -mdisable-callt
13712 @opindex mdisable-callt
13713 This option will suppress generation of the CALLT instruction for the
13714 v850e and v850e1 flavors of the v850 architecture. The default is
13715 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13720 @subsection VAX Options
13721 @cindex VAX options
13723 These @samp{-m} options are defined for the VAX:
13728 Do not output certain jump instructions (@code{aobleq} and so on)
13729 that the Unix assembler for the VAX cannot handle across long
13734 Do output those jump instructions, on the assumption that you
13735 will assemble with the GNU assembler.
13739 Output code for g-format floating point numbers instead of d-format.
13742 @node x86-64 Options
13743 @subsection x86-64 Options
13744 @cindex x86-64 options
13746 These are listed under @xref{i386 and x86-64 Options}.
13748 @node Xstormy16 Options
13749 @subsection Xstormy16 Options
13750 @cindex Xstormy16 Options
13752 These options are defined for Xstormy16:
13757 Choose startup files and linker script suitable for the simulator.
13760 @node Xtensa Options
13761 @subsection Xtensa Options
13762 @cindex Xtensa Options
13764 These options are supported for Xtensa targets:
13768 @itemx -mno-const16
13770 @opindex mno-const16
13771 Enable or disable use of @code{CONST16} instructions for loading
13772 constant values. The @code{CONST16} instruction is currently not a
13773 standard option from Tensilica. When enabled, @code{CONST16}
13774 instructions are always used in place of the standard @code{L32R}
13775 instructions. The use of @code{CONST16} is enabled by default only if
13776 the @code{L32R} instruction is not available.
13779 @itemx -mno-fused-madd
13780 @opindex mfused-madd
13781 @opindex mno-fused-madd
13782 Enable or disable use of fused multiply/add and multiply/subtract
13783 instructions in the floating-point option. This has no effect if the
13784 floating-point option is not also enabled. Disabling fused multiply/add
13785 and multiply/subtract instructions forces the compiler to use separate
13786 instructions for the multiply and add/subtract operations. This may be
13787 desirable in some cases where strict IEEE 754-compliant results are
13788 required: the fused multiply add/subtract instructions do not round the
13789 intermediate result, thereby producing results with @emph{more} bits of
13790 precision than specified by the IEEE standard. Disabling fused multiply
13791 add/subtract instructions also ensures that the program output is not
13792 sensitive to the compiler's ability to combine multiply and add/subtract
13795 @item -mtext-section-literals
13796 @itemx -mno-text-section-literals
13797 @opindex mtext-section-literals
13798 @opindex mno-text-section-literals
13799 Control the treatment of literal pools. The default is
13800 @option{-mno-text-section-literals}, which places literals in a separate
13801 section in the output file. This allows the literal pool to be placed
13802 in a data RAM/ROM, and it also allows the linker to combine literal
13803 pools from separate object files to remove redundant literals and
13804 improve code size. With @option{-mtext-section-literals}, the literals
13805 are interspersed in the text section in order to keep them as close as
13806 possible to their references. This may be necessary for large assembly
13809 @item -mtarget-align
13810 @itemx -mno-target-align
13811 @opindex mtarget-align
13812 @opindex mno-target-align
13813 When this option is enabled, GCC instructs the assembler to
13814 automatically align instructions to reduce branch penalties at the
13815 expense of some code density. The assembler attempts to widen density
13816 instructions to align branch targets and the instructions following call
13817 instructions. If there are not enough preceding safe density
13818 instructions to align a target, no widening will be performed. The
13819 default is @option{-mtarget-align}. These options do not affect the
13820 treatment of auto-aligned instructions like @code{LOOP}, which the
13821 assembler will always align, either by widening density instructions or
13822 by inserting no-op instructions.
13825 @itemx -mno-longcalls
13826 @opindex mlongcalls
13827 @opindex mno-longcalls
13828 When this option is enabled, GCC instructs the assembler to translate
13829 direct calls to indirect calls unless it can determine that the target
13830 of a direct call is in the range allowed by the call instruction. This
13831 translation typically occurs for calls to functions in other source
13832 files. Specifically, the assembler translates a direct @code{CALL}
13833 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13834 The default is @option{-mno-longcalls}. This option should be used in
13835 programs where the call target can potentially be out of range. This
13836 option is implemented in the assembler, not the compiler, so the
13837 assembly code generated by GCC will still show direct call
13838 instructions---look at the disassembled object code to see the actual
13839 instructions. Note that the assembler will use an indirect call for
13840 every cross-file call, not just those that really will be out of range.
13843 @node zSeries Options
13844 @subsection zSeries Options
13845 @cindex zSeries options
13847 These are listed under @xref{S/390 and zSeries Options}.
13849 @node Code Gen Options
13850 @section Options for Code Generation Conventions
13851 @cindex code generation conventions
13852 @cindex options, code generation
13853 @cindex run-time options
13855 These machine-independent options control the interface conventions
13856 used in code generation.
13858 Most of them have both positive and negative forms; the negative form
13859 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13860 one of the forms is listed---the one which is not the default. You
13861 can figure out the other form by either removing @samp{no-} or adding
13865 @item -fbounds-check
13866 @opindex fbounds-check
13867 For front-ends that support it, generate additional code to check that
13868 indices used to access arrays are within the declared range. This is
13869 currently only supported by the Java and Fortran front-ends, where
13870 this option defaults to true and false respectively.
13874 This option generates traps for signed overflow on addition, subtraction,
13875 multiplication operations.
13879 This option instructs the compiler to assume that signed arithmetic
13880 overflow of addition, subtraction and multiplication wraps around
13881 using twos-complement representation. This flag enables some optimizations
13882 and disables others. This option is enabled by default for the Java
13883 front-end, as required by the Java language specification.
13886 @opindex fexceptions
13887 Enable exception handling. Generates extra code needed to propagate
13888 exceptions. For some targets, this implies GCC will generate frame
13889 unwind information for all functions, which can produce significant data
13890 size overhead, although it does not affect execution. If you do not
13891 specify this option, GCC will enable it by default for languages like
13892 C++ which normally require exception handling, and disable it for
13893 languages like C that do not normally require it. However, you may need
13894 to enable this option when compiling C code that needs to interoperate
13895 properly with exception handlers written in C++. You may also wish to
13896 disable this option if you are compiling older C++ programs that don't
13897 use exception handling.
13899 @item -fnon-call-exceptions
13900 @opindex fnon-call-exceptions
13901 Generate code that allows trapping instructions to throw exceptions.
13902 Note that this requires platform-specific runtime support that does
13903 not exist everywhere. Moreover, it only allows @emph{trapping}
13904 instructions to throw exceptions, i.e.@: memory references or floating
13905 point instructions. It does not allow exceptions to be thrown from
13906 arbitrary signal handlers such as @code{SIGALRM}.
13908 @item -funwind-tables
13909 @opindex funwind-tables
13910 Similar to @option{-fexceptions}, except that it will just generate any needed
13911 static data, but will not affect the generated code in any other way.
13912 You will normally not enable this option; instead, a language processor
13913 that needs this handling would enable it on your behalf.
13915 @item -fasynchronous-unwind-tables
13916 @opindex fasynchronous-unwind-tables
13917 Generate unwind table in dwarf2 format, if supported by target machine. The
13918 table is exact at each instruction boundary, so it can be used for stack
13919 unwinding from asynchronous events (such as debugger or garbage collector).
13921 @item -fpcc-struct-return
13922 @opindex fpcc-struct-return
13923 Return ``short'' @code{struct} and @code{union} values in memory like
13924 longer ones, rather than in registers. This convention is less
13925 efficient, but it has the advantage of allowing intercallability between
13926 GCC-compiled files and files compiled with other compilers, particularly
13927 the Portable C Compiler (pcc).
13929 The precise convention for returning structures in memory depends
13930 on the target configuration macros.
13932 Short structures and unions are those whose size and alignment match
13933 that of some integer type.
13935 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13936 switch is not binary compatible with code compiled with the
13937 @option{-freg-struct-return} switch.
13938 Use it to conform to a non-default application binary interface.
13940 @item -freg-struct-return
13941 @opindex freg-struct-return
13942 Return @code{struct} and @code{union} values in registers when possible.
13943 This is more efficient for small structures than
13944 @option{-fpcc-struct-return}.
13946 If you specify neither @option{-fpcc-struct-return} nor
13947 @option{-freg-struct-return}, GCC defaults to whichever convention is
13948 standard for the target. If there is no standard convention, GCC
13949 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13950 the principal compiler. In those cases, we can choose the standard, and
13951 we chose the more efficient register return alternative.
13953 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13954 switch is not binary compatible with code compiled with the
13955 @option{-fpcc-struct-return} switch.
13956 Use it to conform to a non-default application binary interface.
13958 @item -fshort-enums
13959 @opindex fshort-enums
13960 Allocate to an @code{enum} type only as many bytes as it needs for the
13961 declared range of possible values. Specifically, the @code{enum} type
13962 will be equivalent to the smallest integer type which has enough room.
13964 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13965 code that is not binary compatible with code generated without that switch.
13966 Use it to conform to a non-default application binary interface.
13968 @item -fshort-double
13969 @opindex fshort-double
13970 Use the same size for @code{double} as for @code{float}.
13972 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13973 code that is not binary compatible with code generated without that switch.
13974 Use it to conform to a non-default application binary interface.
13976 @item -fshort-wchar
13977 @opindex fshort-wchar
13978 Override the underlying type for @samp{wchar_t} to be @samp{short
13979 unsigned int} instead of the default for the target. This option is
13980 useful for building programs to run under WINE@.
13982 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13983 code that is not binary compatible with code generated without that switch.
13984 Use it to conform to a non-default application binary interface.
13987 @opindex fno-common
13988 In C, allocate even uninitialized global variables in the data section of the
13989 object file, rather than generating them as common blocks. This has the
13990 effect that if the same variable is declared (without @code{extern}) in
13991 two different compilations, you will get an error when you link them.
13992 The only reason this might be useful is if you wish to verify that the
13993 program will work on other systems which always work this way.
13997 Ignore the @samp{#ident} directive.
13999 @item -finhibit-size-directive
14000 @opindex finhibit-size-directive
14001 Don't output a @code{.size} assembler directive, or anything else that
14002 would cause trouble if the function is split in the middle, and the
14003 two halves are placed at locations far apart in memory. This option is
14004 used when compiling @file{crtstuff.c}; you should not need to use it
14007 @item -fverbose-asm
14008 @opindex fverbose-asm
14009 Put extra commentary information in the generated assembly code to
14010 make it more readable. This option is generally only of use to those
14011 who actually need to read the generated assembly code (perhaps while
14012 debugging the compiler itself).
14014 @option{-fno-verbose-asm}, the default, causes the
14015 extra information to be omitted and is useful when comparing two assembler
14018 @item -frecord-gcc-switches
14019 @opindex frecord-gcc-switches
14020 This switch causes the command line that was used to invoke the
14021 compiler to be recorded into the object file that is being created.
14022 This switch is only implemented on some targets and the exact format
14023 of the recording is target and binary file format dependent, but it
14024 usually takes the form of a section containing ASCII text. This
14025 switch is related to the @option{-fverbose-asm} switch, but that
14026 switch only records information in the assembler output file as
14027 comments, so it never reaches the object file.
14031 @cindex global offset table
14033 Generate position-independent code (PIC) suitable for use in a shared
14034 library, if supported for the target machine. Such code accesses all
14035 constant addresses through a global offset table (GOT)@. The dynamic
14036 loader resolves the GOT entries when the program starts (the dynamic
14037 loader is not part of GCC; it is part of the operating system). If
14038 the GOT size for the linked executable exceeds a machine-specific
14039 maximum size, you get an error message from the linker indicating that
14040 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14041 instead. (These maximums are 8k on the SPARC and 32k
14042 on the m68k and RS/6000. The 386 has no such limit.)
14044 Position-independent code requires special support, and therefore works
14045 only on certain machines. For the 386, GCC supports PIC for System V
14046 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14047 position-independent.
14049 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14054 If supported for the target machine, emit position-independent code,
14055 suitable for dynamic linking and avoiding any limit on the size of the
14056 global offset table. This option makes a difference on the m68k,
14057 PowerPC and SPARC@.
14059 Position-independent code requires special support, and therefore works
14060 only on certain machines.
14062 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14069 These options are similar to @option{-fpic} and @option{-fPIC}, but
14070 generated position independent code can be only linked into executables.
14071 Usually these options are used when @option{-pie} GCC option will be
14072 used during linking.
14074 @item -fno-jump-tables
14075 @opindex fno-jump-tables
14076 Do not use jump tables for switch statements even where it would be
14077 more efficient than other code generation strategies. This option is
14078 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14079 building code which forms part of a dynamic linker and cannot
14080 reference the address of a jump table. On some targets, jump tables
14081 do not require a GOT and this option is not needed.
14083 @item -ffixed-@var{reg}
14085 Treat the register named @var{reg} as a fixed register; generated code
14086 should never refer to it (except perhaps as a stack pointer, frame
14087 pointer or in some other fixed role).
14089 @var{reg} must be the name of a register. The register names accepted
14090 are machine-specific and are defined in the @code{REGISTER_NAMES}
14091 macro in the machine description macro file.
14093 This flag does not have a negative form, because it specifies a
14096 @item -fcall-used-@var{reg}
14097 @opindex fcall-used
14098 Treat the register named @var{reg} as an allocable register that is
14099 clobbered by function calls. It may be allocated for temporaries or
14100 variables that do not live across a call. Functions compiled this way
14101 will not save and restore the register @var{reg}.
14103 It is an error to used this flag with the frame pointer or stack pointer.
14104 Use of this flag for other registers that have fixed pervasive roles in
14105 the machine's execution model will produce disastrous results.
14107 This flag does not have a negative form, because it specifies a
14110 @item -fcall-saved-@var{reg}
14111 @opindex fcall-saved
14112 Treat the register named @var{reg} as an allocable register saved by
14113 functions. It may be allocated even for temporaries or variables that
14114 live across a call. Functions compiled this way will save and restore
14115 the register @var{reg} if they use it.
14117 It is an error to used this flag with the frame pointer or stack pointer.
14118 Use of this flag for other registers that have fixed pervasive roles in
14119 the machine's execution model will produce disastrous results.
14121 A different sort of disaster will result from the use of this flag for
14122 a register in which function values may be returned.
14124 This flag does not have a negative form, because it specifies a
14127 @item -fpack-struct[=@var{n}]
14128 @opindex fpack-struct
14129 Without a value specified, pack all structure members together without
14130 holes. When a value is specified (which must be a small power of two), pack
14131 structure members according to this value, representing the maximum
14132 alignment (that is, objects with default alignment requirements larger than
14133 this will be output potentially unaligned at the next fitting location.
14135 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14136 code that is not binary compatible with code generated without that switch.
14137 Additionally, it makes the code suboptimal.
14138 Use it to conform to a non-default application binary interface.
14140 @item -finstrument-functions
14141 @opindex finstrument-functions
14142 Generate instrumentation calls for entry and exit to functions. Just
14143 after function entry and just before function exit, the following
14144 profiling functions will be called with the address of the current
14145 function and its call site. (On some platforms,
14146 @code{__builtin_return_address} does not work beyond the current
14147 function, so the call site information may not be available to the
14148 profiling functions otherwise.)
14151 void __cyg_profile_func_enter (void *this_fn,
14153 void __cyg_profile_func_exit (void *this_fn,
14157 The first argument is the address of the start of the current function,
14158 which may be looked up exactly in the symbol table.
14160 This instrumentation is also done for functions expanded inline in other
14161 functions. The profiling calls will indicate where, conceptually, the
14162 inline function is entered and exited. This means that addressable
14163 versions of such functions must be available. If all your uses of a
14164 function are expanded inline, this may mean an additional expansion of
14165 code size. If you use @samp{extern inline} in your C code, an
14166 addressable version of such functions must be provided. (This is
14167 normally the case anyways, but if you get lucky and the optimizer always
14168 expands the functions inline, you might have gotten away without
14169 providing static copies.)
14171 A function may be given the attribute @code{no_instrument_function}, in
14172 which case this instrumentation will not be done. This can be used, for
14173 example, for the profiling functions listed above, high-priority
14174 interrupt routines, and any functions from which the profiling functions
14175 cannot safely be called (perhaps signal handlers, if the profiling
14176 routines generate output or allocate memory).
14178 @item -fstack-check
14179 @opindex fstack-check
14180 Generate code to verify that you do not go beyond the boundary of the
14181 stack. You should specify this flag if you are running in an
14182 environment with multiple threads, but only rarely need to specify it in
14183 a single-threaded environment since stack overflow is automatically
14184 detected on nearly all systems if there is only one stack.
14186 Note that this switch does not actually cause checking to be done; the
14187 operating system must do that. The switch causes generation of code
14188 to ensure that the operating system sees the stack being extended.
14190 @item -fstack-limit-register=@var{reg}
14191 @itemx -fstack-limit-symbol=@var{sym}
14192 @itemx -fno-stack-limit
14193 @opindex fstack-limit-register
14194 @opindex fstack-limit-symbol
14195 @opindex fno-stack-limit
14196 Generate code to ensure that the stack does not grow beyond a certain value,
14197 either the value of a register or the address of a symbol. If the stack
14198 would grow beyond the value, a signal is raised. For most targets,
14199 the signal is raised before the stack overruns the boundary, so
14200 it is possible to catch the signal without taking special precautions.
14202 For instance, if the stack starts at absolute address @samp{0x80000000}
14203 and grows downwards, you can use the flags
14204 @option{-fstack-limit-symbol=__stack_limit} and
14205 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
14206 of 128KB@. Note that this may only work with the GNU linker.
14208 @cindex aliasing of parameters
14209 @cindex parameters, aliased
14210 @item -fargument-alias
14211 @itemx -fargument-noalias
14212 @itemx -fargument-noalias-global
14213 @itemx -fargument-noalias-anything
14214 @opindex fargument-alias
14215 @opindex fargument-noalias
14216 @opindex fargument-noalias-global
14217 @opindex fargument-noalias-anything
14218 Specify the possible relationships among parameters and between
14219 parameters and global data.
14221 @option{-fargument-alias} specifies that arguments (parameters) may
14222 alias each other and may alias global storage.@*
14223 @option{-fargument-noalias} specifies that arguments do not alias
14224 each other, but may alias global storage.@*
14225 @option{-fargument-noalias-global} specifies that arguments do not
14226 alias each other and do not alias global storage.
14227 @option{-fargument-noalias-anything} specifies that arguments do not
14228 alias any other storage.
14230 Each language will automatically use whatever option is required by
14231 the language standard. You should not need to use these options yourself.
14233 @item -fleading-underscore
14234 @opindex fleading-underscore
14235 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14236 change the way C symbols are represented in the object file. One use
14237 is to help link with legacy assembly code.
14239 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14240 generate code that is not binary compatible with code generated without that
14241 switch. Use it to conform to a non-default application binary interface.
14242 Not all targets provide complete support for this switch.
14244 @item -ftls-model=@var{model}
14245 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14246 The @var{model} argument should be one of @code{global-dynamic},
14247 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
14249 The default without @option{-fpic} is @code{initial-exec}; with
14250 @option{-fpic} the default is @code{global-dynamic}.
14252 @item -fvisibility=@var{default|internal|hidden|protected}
14253 @opindex fvisibility
14254 Set the default ELF image symbol visibility to the specified option---all
14255 symbols will be marked with this unless overridden within the code.
14256 Using this feature can very substantially improve linking and
14257 load times of shared object libraries, produce more optimized
14258 code, provide near-perfect API export and prevent symbol clashes.
14259 It is @strong{strongly} recommended that you use this in any shared objects
14262 Despite the nomenclature, @code{default} always means public ie;
14263 available to be linked against from outside the shared object.
14264 @code{protected} and @code{internal} are pretty useless in real-world
14265 usage so the only other commonly used option will be @code{hidden}.
14266 The default if @option{-fvisibility} isn't specified is
14267 @code{default}, i.e., make every
14268 symbol public---this causes the same behavior as previous versions of
14271 A good explanation of the benefits offered by ensuring ELF
14272 symbols have the correct visibility is given by ``How To Write
14273 Shared Libraries'' by Ulrich Drepper (which can be found at
14274 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
14275 solution made possible by this option to marking things hidden when
14276 the default is public is to make the default hidden and mark things
14277 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
14278 and @code{__attribute__ ((visibility("default")))} instead of
14279 @code{__declspec(dllexport)} you get almost identical semantics with
14280 identical syntax. This is a great boon to those working with
14281 cross-platform projects.
14283 For those adding visibility support to existing code, you may find
14284 @samp{#pragma GCC visibility} of use. This works by you enclosing
14285 the declarations you wish to set visibility for with (for example)
14286 @samp{#pragma GCC visibility push(hidden)} and
14287 @samp{#pragma GCC visibility pop}.
14288 Bear in mind that symbol visibility should be viewed @strong{as
14289 part of the API interface contract} and thus all new code should
14290 always specify visibility when it is not the default ie; declarations
14291 only for use within the local DSO should @strong{always} be marked explicitly
14292 as hidden as so to avoid PLT indirection overheads---making this
14293 abundantly clear also aids readability and self-documentation of the code.
14294 Note that due to ISO C++ specification requirements, operator new and
14295 operator delete must always be of default visibility.
14297 Be aware that headers from outside your project, in particular system
14298 headers and headers from any other library you use, may not be
14299 expecting to be compiled with visibility other than the default. You
14300 may need to explicitly say @samp{#pragma GCC visibility push(default)}
14301 before including any such headers.
14303 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
14304 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
14305 no modifications. However, this means that calls to @samp{extern}
14306 functions with no explicit visibility will use the PLT, so it is more
14307 effective to use @samp{__attribute ((visibility))} and/or
14308 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
14309 declarations should be treated as hidden.
14311 Note that @samp{-fvisibility} does affect C++ vague linkage
14312 entities. This means that, for instance, an exception class that will
14313 be thrown between DSOs must be explicitly marked with default
14314 visibility so that the @samp{type_info} nodes will be unified between
14317 An overview of these techniques, their benefits and how to use them
14318 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
14324 @node Environment Variables
14325 @section Environment Variables Affecting GCC
14326 @cindex environment variables
14328 @c man begin ENVIRONMENT
14329 This section describes several environment variables that affect how GCC
14330 operates. Some of them work by specifying directories or prefixes to use
14331 when searching for various kinds of files. Some are used to specify other
14332 aspects of the compilation environment.
14334 Note that you can also specify places to search using options such as
14335 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
14336 take precedence over places specified using environment variables, which
14337 in turn take precedence over those specified by the configuration of GCC@.
14338 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
14339 GNU Compiler Collection (GCC) Internals}.
14344 @c @itemx LC_COLLATE
14346 @c @itemx LC_MONETARY
14347 @c @itemx LC_NUMERIC
14352 @c @findex LC_COLLATE
14353 @findex LC_MESSAGES
14354 @c @findex LC_MONETARY
14355 @c @findex LC_NUMERIC
14359 These environment variables control the way that GCC uses
14360 localization information that allow GCC to work with different
14361 national conventions. GCC inspects the locale categories
14362 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
14363 so. These locale categories can be set to any value supported by your
14364 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
14365 Kingdom encoded in UTF-8.
14367 The @env{LC_CTYPE} environment variable specifies character
14368 classification. GCC uses it to determine the character boundaries in
14369 a string; this is needed for some multibyte encodings that contain quote
14370 and escape characters that would otherwise be interpreted as a string
14373 The @env{LC_MESSAGES} environment variable specifies the language to
14374 use in diagnostic messages.
14376 If the @env{LC_ALL} environment variable is set, it overrides the value
14377 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
14378 and @env{LC_MESSAGES} default to the value of the @env{LANG}
14379 environment variable. If none of these variables are set, GCC
14380 defaults to traditional C English behavior.
14384 If @env{TMPDIR} is set, it specifies the directory to use for temporary
14385 files. GCC uses temporary files to hold the output of one stage of
14386 compilation which is to be used as input to the next stage: for example,
14387 the output of the preprocessor, which is the input to the compiler
14390 @item GCC_EXEC_PREFIX
14391 @findex GCC_EXEC_PREFIX
14392 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
14393 names of the subprograms executed by the compiler. No slash is added
14394 when this prefix is combined with the name of a subprogram, but you can
14395 specify a prefix that ends with a slash if you wish.
14397 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
14398 an appropriate prefix to use based on the pathname it was invoked with.
14400 If GCC cannot find the subprogram using the specified prefix, it
14401 tries looking in the usual places for the subprogram.
14403 The default value of @env{GCC_EXEC_PREFIX} is
14404 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
14405 the installed compiler. In many cases @var{prefix} is the value
14406 of @code{prefix} when you ran the @file{configure} script.
14408 Other prefixes specified with @option{-B} take precedence over this prefix.
14410 This prefix is also used for finding files such as @file{crt0.o} that are
14413 In addition, the prefix is used in an unusual way in finding the
14414 directories to search for header files. For each of the standard
14415 directories whose name normally begins with @samp{/usr/local/lib/gcc}
14416 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
14417 replacing that beginning with the specified prefix to produce an
14418 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
14419 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
14420 These alternate directories are searched first; the standard directories
14421 come next. If a standard directory begins with the configured
14422 @var{prefix} then the value of @var{prefix} is replaced by
14423 @env{GCC_EXEC_PREFIX} when looking for header files.
14425 @item COMPILER_PATH
14426 @findex COMPILER_PATH
14427 The value of @env{COMPILER_PATH} is a colon-separated list of
14428 directories, much like @env{PATH}. GCC tries the directories thus
14429 specified when searching for subprograms, if it can't find the
14430 subprograms using @env{GCC_EXEC_PREFIX}.
14433 @findex LIBRARY_PATH
14434 The value of @env{LIBRARY_PATH} is a colon-separated list of
14435 directories, much like @env{PATH}. When configured as a native compiler,
14436 GCC tries the directories thus specified when searching for special
14437 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
14438 using GCC also uses these directories when searching for ordinary
14439 libraries for the @option{-l} option (but directories specified with
14440 @option{-L} come first).
14444 @cindex locale definition
14445 This variable is used to pass locale information to the compiler. One way in
14446 which this information is used is to determine the character set to be used
14447 when character literals, string literals and comments are parsed in C and C++.
14448 When the compiler is configured to allow multibyte characters,
14449 the following values for @env{LANG} are recognized:
14453 Recognize JIS characters.
14455 Recognize SJIS characters.
14457 Recognize EUCJP characters.
14460 If @env{LANG} is not defined, or if it has some other value, then the
14461 compiler will use mblen and mbtowc as defined by the default locale to
14462 recognize and translate multibyte characters.
14466 Some additional environments variables affect the behavior of the
14469 @include cppenv.texi
14473 @node Precompiled Headers
14474 @section Using Precompiled Headers
14475 @cindex precompiled headers
14476 @cindex speed of compilation
14478 Often large projects have many header files that are included in every
14479 source file. The time the compiler takes to process these header files
14480 over and over again can account for nearly all of the time required to
14481 build the project. To make builds faster, GCC allows users to
14482 `precompile' a header file; then, if builds can use the precompiled
14483 header file they will be much faster.
14485 To create a precompiled header file, simply compile it as you would any
14486 other file, if necessary using the @option{-x} option to make the driver
14487 treat it as a C or C++ header file. You will probably want to use a
14488 tool like @command{make} to keep the precompiled header up-to-date when
14489 the headers it contains change.
14491 A precompiled header file will be searched for when @code{#include} is
14492 seen in the compilation. As it searches for the included file
14493 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
14494 compiler looks for a precompiled header in each directory just before it
14495 looks for the include file in that directory. The name searched for is
14496 the name specified in the @code{#include} with @samp{.gch} appended. If
14497 the precompiled header file can't be used, it is ignored.
14499 For instance, if you have @code{#include "all.h"}, and you have
14500 @file{all.h.gch} in the same directory as @file{all.h}, then the
14501 precompiled header file will be used if possible, and the original
14502 header will be used otherwise.
14504 Alternatively, you might decide to put the precompiled header file in a
14505 directory and use @option{-I} to ensure that directory is searched
14506 before (or instead of) the directory containing the original header.
14507 Then, if you want to check that the precompiled header file is always
14508 used, you can put a file of the same name as the original header in this
14509 directory containing an @code{#error} command.
14511 This also works with @option{-include}. So yet another way to use
14512 precompiled headers, good for projects not designed with precompiled
14513 header files in mind, is to simply take most of the header files used by
14514 a project, include them from another header file, precompile that header
14515 file, and @option{-include} the precompiled header. If the header files
14516 have guards against multiple inclusion, they will be skipped because
14517 they've already been included (in the precompiled header).
14519 If you need to precompile the same header file for different
14520 languages, targets, or compiler options, you can instead make a
14521 @emph{directory} named like @file{all.h.gch}, and put each precompiled
14522 header in the directory, perhaps using @option{-o}. It doesn't matter
14523 what you call the files in the directory, every precompiled header in
14524 the directory will be considered. The first precompiled header
14525 encountered in the directory that is valid for this compilation will
14526 be used; they're searched in no particular order.
14528 There are many other possibilities, limited only by your imagination,
14529 good sense, and the constraints of your build system.
14531 A precompiled header file can be used only when these conditions apply:
14535 Only one precompiled header can be used in a particular compilation.
14538 A precompiled header can't be used once the first C token is seen. You
14539 can have preprocessor directives before a precompiled header; you can
14540 even include a precompiled header from inside another header, so long as
14541 there are no C tokens before the @code{#include}.
14544 The precompiled header file must be produced for the same language as
14545 the current compilation. You can't use a C precompiled header for a C++
14549 The precompiled header file must have been produced by the same compiler
14550 binary as the current compilation is using.
14553 Any macros defined before the precompiled header is included must
14554 either be defined in the same way as when the precompiled header was
14555 generated, or must not affect the precompiled header, which usually
14556 means that they don't appear in the precompiled header at all.
14558 The @option{-D} option is one way to define a macro before a
14559 precompiled header is included; using a @code{#define} can also do it.
14560 There are also some options that define macros implicitly, like
14561 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
14564 @item If debugging information is output when using the precompiled
14565 header, using @option{-g} or similar, the same kind of debugging information
14566 must have been output when building the precompiled header. However,
14567 a precompiled header built using @option{-g} can be used in a compilation
14568 when no debugging information is being output.
14570 @item The same @option{-m} options must generally be used when building
14571 and using the precompiled header. @xref{Submodel Options},
14572 for any cases where this rule is relaxed.
14574 @item Each of the following options must be the same when building and using
14575 the precompiled header:
14577 @gccoptlist{-fexceptions -funit-at-a-time}
14580 Some other command-line options starting with @option{-f},
14581 @option{-p}, or @option{-O} must be defined in the same way as when
14582 the precompiled header was generated. At present, it's not clear
14583 which options are safe to change and which are not; the safest choice
14584 is to use exactly the same options when generating and using the
14585 precompiled header. The following are known to be safe:
14587 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
14588 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
14589 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
14594 For all of these except the last, the compiler will automatically
14595 ignore the precompiled header if the conditions aren't met. If you
14596 find an option combination that doesn't work and doesn't cause the
14597 precompiled header to be ignored, please consider filing a bug report,
14600 If you do use differing options when generating and using the
14601 precompiled header, the actual behavior will be a mixture of the
14602 behavior for the options. For instance, if you use @option{-g} to
14603 generate the precompiled header but not when using it, you may or may
14604 not get debugging information for routines in the precompiled header.
14606 @node Running Protoize
14607 @section Running Protoize
14609 The program @code{protoize} is an optional part of GCC@. You can use
14610 it to add prototypes to a program, thus converting the program to ISO
14611 C in one respect. The companion program @code{unprotoize} does the
14612 reverse: it removes argument types from any prototypes that are found.
14614 When you run these programs, you must specify a set of source files as
14615 command line arguments. The conversion programs start out by compiling
14616 these files to see what functions they define. The information gathered
14617 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
14619 After scanning comes actual conversion. The specified files are all
14620 eligible to be converted; any files they include (whether sources or
14621 just headers) are eligible as well.
14623 But not all the eligible files are converted. By default,
14624 @code{protoize} and @code{unprotoize} convert only source and header
14625 files in the current directory. You can specify additional directories
14626 whose files should be converted with the @option{-d @var{directory}}
14627 option. You can also specify particular files to exclude with the
14628 @option{-x @var{file}} option. A file is converted if it is eligible, its
14629 directory name matches one of the specified directory names, and its
14630 name within the directory has not been excluded.
14632 Basic conversion with @code{protoize} consists of rewriting most
14633 function definitions and function declarations to specify the types of
14634 the arguments. The only ones not rewritten are those for varargs
14637 @code{protoize} optionally inserts prototype declarations at the
14638 beginning of the source file, to make them available for any calls that
14639 precede the function's definition. Or it can insert prototype
14640 declarations with block scope in the blocks where undeclared functions
14643 Basic conversion with @code{unprotoize} consists of rewriting most
14644 function declarations to remove any argument types, and rewriting
14645 function definitions to the old-style pre-ISO form.
14647 Both conversion programs print a warning for any function declaration or
14648 definition that they can't convert. You can suppress these warnings
14651 The output from @code{protoize} or @code{unprotoize} replaces the
14652 original source file. The original file is renamed to a name ending
14653 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
14654 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
14655 for DOS) file already exists, then the source file is simply discarded.
14657 @code{protoize} and @code{unprotoize} both depend on GCC itself to
14658 scan the program and collect information about the functions it uses.
14659 So neither of these programs will work until GCC is installed.
14661 Here is a table of the options you can use with @code{protoize} and
14662 @code{unprotoize}. Each option works with both programs unless
14666 @item -B @var{directory}
14667 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
14668 usual directory (normally @file{/usr/local/lib}). This file contains
14669 prototype information about standard system functions. This option
14670 applies only to @code{protoize}.
14672 @item -c @var{compilation-options}
14673 Use @var{compilation-options} as the options when running @command{gcc} to
14674 produce the @samp{.X} files. The special option @option{-aux-info} is
14675 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
14677 Note that the compilation options must be given as a single argument to
14678 @code{protoize} or @code{unprotoize}. If you want to specify several
14679 @command{gcc} options, you must quote the entire set of compilation options
14680 to make them a single word in the shell.
14682 There are certain @command{gcc} arguments that you cannot use, because they
14683 would produce the wrong kind of output. These include @option{-g},
14684 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14685 the @var{compilation-options}, they are ignored.
14688 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14689 systems) instead of @samp{.c}. This is convenient if you are converting
14690 a C program to C++. This option applies only to @code{protoize}.
14693 Add explicit global declarations. This means inserting explicit
14694 declarations at the beginning of each source file for each function
14695 that is called in the file and was not declared. These declarations
14696 precede the first function definition that contains a call to an
14697 undeclared function. This option applies only to @code{protoize}.
14699 @item -i @var{string}
14700 Indent old-style parameter declarations with the string @var{string}.
14701 This option applies only to @code{protoize}.
14703 @code{unprotoize} converts prototyped function definitions to old-style
14704 function definitions, where the arguments are declared between the
14705 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14706 uses five spaces as the indentation. If you want to indent with just
14707 one space instead, use @option{-i " "}.
14710 Keep the @samp{.X} files. Normally, they are deleted after conversion
14714 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14715 a prototype declaration for each function in each block which calls the
14716 function without any declaration. This option applies only to
14720 Make no real changes. This mode just prints information about the conversions
14721 that would have been done without @option{-n}.
14724 Make no @samp{.save} files. The original files are simply deleted.
14725 Use this option with caution.
14727 @item -p @var{program}
14728 Use the program @var{program} as the compiler. Normally, the name
14729 @file{gcc} is used.
14732 Work quietly. Most warnings are suppressed.
14735 Print the version number, just like @option{-v} for @command{gcc}.
14738 If you need special compiler options to compile one of your program's
14739 source files, then you should generate that file's @samp{.X} file
14740 specially, by running @command{gcc} on that source file with the
14741 appropriate options and the option @option{-aux-info}. Then run
14742 @code{protoize} on the entire set of files. @code{protoize} will use
14743 the existing @samp{.X} file because it is newer than the source file.
14747 gcc -Dfoo=bar file1.c -aux-info file1.X
14752 You need to include the special files along with the rest in the
14753 @code{protoize} command, even though their @samp{.X} files already
14754 exist, because otherwise they won't get converted.
14756 @xref{Protoize Caveats}, for more information on how to use
14757 @code{protoize} successfully.