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} -aux-info @var{filename} @gol
170 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
171 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
172 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
173 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
174 -fsigned-bitfields -fsigned-char @gol
175 -funsigned-bitfields -funsigned-char}
177 @item C++ Language Options
178 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
179 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
180 -fconserve-space -ffriend-injection @gol
181 -fno-elide-constructors @gol
182 -fno-enforce-eh-specs @gol
183 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
184 -fno-implicit-templates @gol
185 -fno-implicit-inline-templates @gol
186 -fno-implement-inlines -fms-extensions @gol
187 -fno-nonansi-builtins -fno-operator-names @gol
188 -fno-optional-diags -fpermissive @gol
189 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
190 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
191 -fno-default-inline -fvisibility-inlines-hidden @gol
192 -Wabi -Wctor-dtor-privacy @gol
193 -Wnon-virtual-dtor -Wreorder @gol
194 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
195 -Wno-non-template-friend -Wold-style-cast @gol
196 -Woverloaded-virtual -Wno-pmf-conversions @gol
199 @item Objective-C and Objective-C++ Language Options
200 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
201 Objective-C and Objective-C++ Dialects}.
202 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
203 -fgnu-runtime -fnext-runtime @gol
204 -fno-nil-receivers @gol
205 -fobjc-call-cxx-cdtors @gol
206 -fobjc-direct-dispatch @gol
207 -fobjc-exceptions @gol
209 -freplace-objc-classes @gol
212 -Wassign-intercept @gol
213 -Wno-protocol -Wselector @gol
214 -Wstrict-selector-match @gol
215 -Wundeclared-selector}
217 @item Language Independent Options
218 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
219 @gccoptlist{-fmessage-length=@var{n} @gol
220 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
221 -fdiagnostics-show-option}
223 @item Warning Options
224 @xref{Warning Options,,Options to Request or Suppress Warnings}.
225 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
226 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
227 -Wno-attributes -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
228 -Wchar-subscripts -Wclobbered -Wcomment @gol
229 -Wconversion -Wcoverage-mismatch -Wno-deprecated-declarations @gol
230 -Wdisabled-optimization -Wno-div-by-zero @gol
231 -Wempty-body -Wno-endif-labels @gol
232 -Werror -Werror=* @gol
233 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
234 -Wno-format-extra-args -Wformat-nonliteral @gol
235 -Wformat-security -Wformat-y2k @gol
236 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
237 -Wimport -Wno-import -Winit-self -Winline @gol
238 -Wno-int-to-pointer-cast @gol
239 -Wno-invalid-offsetof -Winvalid-pch @gol
240 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
241 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
242 -Wmissing-format-attribute -Wmissing-include-dirs @gol
243 -Wmissing-noreturn @gol
244 -Wno-multichar -Wnonnull -Wno-overflow @gol
245 -Woverlength-strings -Wpacked -Wpadded @gol
246 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
247 -Wredundant-decls @gol
248 -Wreturn-type -Wsequence-point -Wshadow @gol
249 -Wsign-compare -Wstack-protector @gol
250 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
251 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
252 -Wswitch -Wswitch-default -Wswitch-enum @gol
253 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
254 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
255 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
256 -Wunused-value -Wunused-variable -Wvariadic-macros @gol
257 -Wvolatile-register-var -Wwrite-strings}
259 @item C-only Warning Options
260 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
261 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
262 -Wold-style-declaration -Wold-style-definition @gol
263 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
264 -Wdeclaration-after-statement -Wpointer-sign}
266 @item Debugging Options
267 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
268 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
269 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
270 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
271 -fdump-ipa-all -fdump-ipa-cgraph @gol
273 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
278 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
283 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-nrv -fdump-tree-vect @gol
287 -fdump-tree-sink @gol
288 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-salias @gol
290 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
292 -ftree-vectorizer-verbose=@var{n} @gol
293 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
294 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
295 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
296 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
297 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
298 -ftest-coverage -ftime-report -fvar-tracking @gol
299 -g -g@var{level} -gcoff -gdwarf-2 @gol
300 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
301 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
302 -print-multi-directory -print-multi-lib @gol
303 -print-prog-name=@var{program} -print-search-dirs -Q @gol
306 @item Optimization Options
307 @xref{Optimize Options,,Options that Control Optimization}.
308 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
309 -falign-labels=@var{n} -falign-loops=@var{n} @gol
310 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
311 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
312 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
313 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
314 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
315 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
316 -fexpensive-optimizations -ffast-math -ffloat-store @gol
317 -fforce-addr -fforward-propagate -ffunction-sections @gol
318 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
319 -fcrossjumping -fif-conversion -fif-conversion2 @gol
320 -finline-functions -finline-functions-called-once @gol
321 -finline-limit=@var{n} -fkeep-inline-functions @gol
322 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
323 -fmodulo-sched -fno-branch-count-reg @gol
324 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
325 -fno-function-cse -fno-guess-branch-probability @gol
326 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
327 -funsafe-math-optimizations -funsafe-loop-optimizations @gol
328 -ffinite-math-only -fno-signed-zeros @gol
329 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
330 -fomit-frame-pointer -foptimize-register-move @gol
331 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
332 -fprofile-generate -fprofile-use @gol
333 -fregmove -frename-registers @gol
334 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
335 -frerun-cse-after-loop @gol
336 -frounding-math -frtl-abstract-sequences @gol
337 -fschedule-insns -fschedule-insns2 @gol
338 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
339 -fsched-spec-load-dangerous @gol
340 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
341 -fsched2-use-superblocks @gol
342 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
343 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
344 -fno-split-wide-types -fstack-protector -fstack-protector-all @gol
345 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
346 -funroll-all-loops -funroll-loops -fpeel-loops @gol
347 -fsplit-ivs-in-unroller -funswitch-loops @gol
348 -fvariable-expansion-in-unroller @gol
349 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
350 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
351 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
352 -ftree-ch -ftree-sra -ftree-ter -ftree-fre -ftree-vectorize @gol
353 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
354 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
355 --param @var{name}=@var{value}
356 -O -O0 -O1 -O2 -O3 -Os}
358 @item Preprocessor Options
359 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
360 @gccoptlist{-A@var{question}=@var{answer} @gol
361 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
362 -C -dD -dI -dM -dN @gol
363 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
364 -idirafter @var{dir} @gol
365 -include @var{file} -imacros @var{file} @gol
366 -iprefix @var{file} -iwithprefix @var{dir} @gol
367 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
368 -imultilib @var{dir} -isysroot @var{dir} @gol
369 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
370 -P -fworking-directory -remap @gol
371 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
372 -Xpreprocessor @var{option}}
374 @item Assembler Option
375 @xref{Assembler Options,,Passing Options to the Assembler}.
376 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
379 @xref{Link Options,,Options for Linking}.
380 @gccoptlist{@var{object-file-name} -l@var{library} @gol
381 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
382 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
383 -Wl,@var{option} -Xlinker @var{option} @gol
386 @item Directory Options
387 @xref{Directory Options,,Options for Directory Search}.
388 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
389 -specs=@var{file} -I- --sysroot=@var{dir}}
392 @c I wrote this xref this way to avoid overfull hbox. -- rms
393 @xref{Target Options}.
394 @gccoptlist{-V @var{version} -b @var{machine}}
396 @item Machine Dependent Options
397 @xref{Submodel Options,,Hardware Models and Configurations}.
398 @c This list is ordered alphanumerically by subsection name.
399 @c Try and put the significant identifier (CPU or system) first,
400 @c so users have a clue at guessing where the ones they want will be.
403 @gccoptlist{-EB -EL @gol
404 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
405 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
408 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
409 -mabi=@var{name} @gol
410 -mapcs-stack-check -mno-apcs-stack-check @gol
411 -mapcs-float -mno-apcs-float @gol
412 -mapcs-reentrant -mno-apcs-reentrant @gol
413 -msched-prolog -mno-sched-prolog @gol
414 -mlittle-endian -mbig-endian -mwords-little-endian @gol
415 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
416 -mthumb-interwork -mno-thumb-interwork @gol
417 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
418 -mstructure-size-boundary=@var{n} @gol
419 -mabort-on-noreturn @gol
420 -mlong-calls -mno-long-calls @gol
421 -msingle-pic-base -mno-single-pic-base @gol
422 -mpic-register=@var{reg} @gol
423 -mnop-fun-dllimport @gol
424 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
425 -mpoke-function-name @gol
427 -mtpcs-frame -mtpcs-leaf-frame @gol
428 -mcaller-super-interworking -mcallee-super-interworking @gol
432 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
433 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
435 @emph{Blackfin Options}
436 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
437 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
438 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
439 -mno-id-shared-library -mshared-library-id=@var{n} @gol
440 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
441 -msep-data -mno-sep-data -mlong-calls -mno-long-calls}
444 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
445 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
446 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
447 -mstack-align -mdata-align -mconst-align @gol
448 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
449 -melf -maout -melinux -mlinux -sim -sim2 @gol
450 -mmul-bug-workaround -mno-mul-bug-workaround}
453 @gccoptlist{-mmac -mpush-args}
455 @emph{Darwin Options}
456 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
457 -arch_only -bind_at_load -bundle -bundle_loader @gol
458 -client_name -compatibility_version -current_version @gol
460 -dependency-file -dylib_file -dylinker_install_name @gol
461 -dynamic -dynamiclib -exported_symbols_list @gol
462 -filelist -flat_namespace -force_cpusubtype_ALL @gol
463 -force_flat_namespace -headerpad_max_install_names @gol
464 -image_base -init -install_name -keep_private_externs @gol
465 -multi_module -multiply_defined -multiply_defined_unused @gol
466 -noall_load -no_dead_strip_inits_and_terms @gol
467 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
468 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
469 -private_bundle -read_only_relocs -sectalign @gol
470 -sectobjectsymbols -whyload -seg1addr @gol
471 -sectcreate -sectobjectsymbols -sectorder @gol
472 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
473 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
474 -segprot -segs_read_only_addr -segs_read_write_addr @gol
475 -single_module -static -sub_library -sub_umbrella @gol
476 -twolevel_namespace -umbrella -undefined @gol
477 -unexported_symbols_list -weak_reference_mismatches @gol
478 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
479 -mkernel -mone-byte-bool}
481 @emph{DEC Alpha Options}
482 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
483 -mieee -mieee-with-inexact -mieee-conformant @gol
484 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
485 -mtrap-precision=@var{mode} -mbuild-constants @gol
486 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
487 -mbwx -mmax -mfix -mcix @gol
488 -mfloat-vax -mfloat-ieee @gol
489 -mexplicit-relocs -msmall-data -mlarge-data @gol
490 -msmall-text -mlarge-text @gol
491 -mmemory-latency=@var{time}}
493 @emph{DEC Alpha/VMS Options}
494 @gccoptlist{-mvms-return-codes}
497 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
498 -mhard-float -msoft-float @gol
499 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
500 -mdouble -mno-double @gol
501 -mmedia -mno-media -mmuladd -mno-muladd @gol
502 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
503 -mlinked-fp -mlong-calls -malign-labels @gol
504 -mlibrary-pic -macc-4 -macc-8 @gol
505 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
506 -moptimize-membar -mno-optimize-membar @gol
507 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
508 -mvliw-branch -mno-vliw-branch @gol
509 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
510 -mno-nested-cond-exec -mtomcat-stats @gol
514 @emph{GNU/Linux Options}
515 @gccoptlist{-muclibc}
517 @emph{H8/300 Options}
518 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
521 @gccoptlist{-march=@var{architecture-type} @gol
522 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
523 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
524 -mfixed-range=@var{register-range} @gol
525 -mjump-in-delay -mlinker-opt -mlong-calls @gol
526 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
527 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
528 -mno-jump-in-delay -mno-long-load-store @gol
529 -mno-portable-runtime -mno-soft-float @gol
530 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
531 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
532 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
533 -munix=@var{unix-std} -nolibdld -static -threads}
535 @emph{i386 and x86-64 Options}
536 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
537 -mfpmath=@var{unit} @gol
538 -masm=@var{dialect} -mno-fancy-math-387 @gol
539 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
540 -mno-wide-multiply -mrtd -malign-double @gol
541 -mpreferred-stack-boundary=@var{num} @gol
542 -mmmx -msse -msse2 -msse3 -mssse3 -msse4a -m3dnow -mpopcnt -mabm @gol
543 -mthreads -mno-align-stringops -minline-all-stringops @gol
544 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
545 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
547 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
548 -mcmodel=@var{code-model} @gol
549 -m32 -m64 -mlarge-data-threshold=@var{num}}
552 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
553 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
554 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
555 -minline-float-divide-max-throughput @gol
556 -minline-int-divide-min-latency @gol
557 -minline-int-divide-max-throughput @gol
558 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
559 -mno-dwarf2-asm -mearly-stop-bits @gol
560 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
561 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
562 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
563 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
564 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
565 -mno-sched-prefer-non-data-spec-insns @gol
566 -mno-sched-prefer-non-control-spec-insns @gol
567 -mno-sched-count-spec-in-critical-path}
569 @emph{M32R/D Options}
570 @gccoptlist{-m32r2 -m32rx -m32r @gol
572 -malign-loops -mno-align-loops @gol
573 -missue-rate=@var{number} @gol
574 -mbranch-cost=@var{number} @gol
575 -mmodel=@var{code-size-model-type} @gol
576 -msdata=@var{sdata-type} @gol
577 -mno-flush-func -mflush-func=@var{name} @gol
578 -mno-flush-trap -mflush-trap=@var{number} @gol
582 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
584 @emph{M680x0 Options}
585 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
586 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
587 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
588 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
589 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
590 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
591 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
592 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
594 @emph{M68hc1x Options}
595 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
596 -mauto-incdec -minmax -mlong-calls -mshort @gol
597 -msoft-reg-count=@var{count}}
600 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
601 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
602 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
603 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
604 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
607 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
608 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
609 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
610 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
611 -mfp32 -mfp64 -mhard-float -msoft-float @gol
612 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
613 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
614 -G@var{num} -membedded-data -mno-embedded-data @gol
615 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
616 -msplit-addresses -mno-split-addresses @gol
617 -mexplicit-relocs -mno-explicit-relocs @gol
618 -mcheck-zero-division -mno-check-zero-division @gol
619 -mdivide-traps -mdivide-breaks @gol
620 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
621 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
622 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
623 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
624 -mfix-sb1 -mno-fix-sb1 @gol
625 -mflush-func=@var{func} -mno-flush-func @gol
626 -mbranch-likely -mno-branch-likely @gol
627 -mfp-exceptions -mno-fp-exceptions @gol
628 -mvr4130-align -mno-vr4130-align}
631 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
632 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
633 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
634 -mno-base-addresses -msingle-exit -mno-single-exit}
636 @emph{MN10300 Options}
637 @gccoptlist{-mmult-bug -mno-mult-bug @gol
638 -mam33 -mno-am33 @gol
639 -mam33-2 -mno-am33-2 @gol
640 -mreturn-pointer-on-d0 @gol
644 @gccoptlist{-mno-crt0 -mbacc -msim @gol
645 -march=@var{cpu-type} }
647 @emph{PDP-11 Options}
648 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
649 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
650 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
651 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
652 -mbranch-expensive -mbranch-cheap @gol
653 -msplit -mno-split -munix-asm -mdec-asm}
655 @emph{PowerPC Options}
656 See RS/6000 and PowerPC Options.
658 @emph{RS/6000 and PowerPC Options}
659 @gccoptlist{-mcpu=@var{cpu-type} @gol
660 -mtune=@var{cpu-type} @gol
661 -mpower -mno-power -mpower2 -mno-power2 @gol
662 -mpowerpc -mpowerpc64 -mno-powerpc @gol
663 -maltivec -mno-altivec @gol
664 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
665 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
666 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
667 -mmfpgpr -mno-mfpgpr @gol
668 -mnew-mnemonics -mold-mnemonics @gol
669 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
670 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
671 -malign-power -malign-natural @gol
672 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
673 -mstring -mno-string -mupdate -mno-update @gol
674 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
675 -mstrict-align -mno-strict-align -mrelocatable @gol
676 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
677 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
678 -mdynamic-no-pic -maltivec -mswdiv @gol
679 -mprioritize-restricted-insns=@var{priority} @gol
680 -msched-costly-dep=@var{dependence_type} @gol
681 -minsert-sched-nops=@var{scheme} @gol
682 -mcall-sysv -mcall-netbsd @gol
683 -maix-struct-return -msvr4-struct-return @gol
684 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
685 -misel -mno-isel @gol
686 -misel=yes -misel=no @gol
688 -mspe=yes -mspe=no @gol
689 -mvrsave -mno-vrsave @gol
690 -mmulhw -mno-mulhw @gol
691 -mdlmzb -mno-dlmzb @gol
692 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
693 -mprototype -mno-prototype @gol
694 -msim -mmvme -mads -myellowknife -memb -msdata @gol
695 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
697 @emph{S/390 and zSeries Options}
698 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
699 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
700 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
701 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
702 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
703 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
704 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
707 @gccoptlist{-mel -mel @gol
712 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
713 -m4-nofpu -m4-single-only -m4-single -m4 @gol
714 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
715 -m5-64media -m5-64media-nofpu @gol
716 -m5-32media -m5-32media-nofpu @gol
717 -m5-compact -m5-compact-nofpu @gol
718 -mb -ml -mdalign -mrelax @gol
719 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
720 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
721 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
722 -mdivsi3_libfunc=@var{name} @gol
723 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
727 @gccoptlist{-mcpu=@var{cpu-type} @gol
728 -mtune=@var{cpu-type} @gol
729 -mcmodel=@var{code-model} @gol
730 -m32 -m64 -mapp-regs -mno-app-regs @gol
731 -mfaster-structs -mno-faster-structs @gol
732 -mfpu -mno-fpu -mhard-float -msoft-float @gol
733 -mhard-quad-float -msoft-quad-float @gol
734 -mimpure-text -mno-impure-text -mlittle-endian @gol
735 -mstack-bias -mno-stack-bias @gol
736 -munaligned-doubles -mno-unaligned-doubles @gol
737 -mv8plus -mno-v8plus -mvis -mno-vis
738 -threads -pthreads -pthread}
741 @gccoptlist{-mwarn-reloc -merror-reloc @gol
742 -msafe-dma -munsafe-dma @gol
744 -msmall-mem -mlarge-mem -mstdmain @gol
745 -mfixed-range=@var{register-range}}
747 @emph{System V Options}
748 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
750 @emph{TMS320C3x/C4x Options}
751 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
752 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
753 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
754 -mparallel-insns -mparallel-mpy -mpreserve-float}
757 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
758 -mprolog-function -mno-prolog-function -mspace @gol
759 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
760 -mapp-regs -mno-app-regs @gol
761 -mdisable-callt -mno-disable-callt @gol
767 @gccoptlist{-mg -mgnu -munix}
769 @emph{x86-64 Options}
770 See i386 and x86-64 Options.
772 @emph{Xstormy16 Options}
775 @emph{Xtensa Options}
776 @gccoptlist{-mconst16 -mno-const16 @gol
777 -mfused-madd -mno-fused-madd @gol
778 -mtext-section-literals -mno-text-section-literals @gol
779 -mtarget-align -mno-target-align @gol
780 -mlongcalls -mno-longcalls}
782 @emph{zSeries Options}
783 See S/390 and zSeries Options.
785 @item Code Generation Options
786 @xref{Code Gen Options,,Options for Code Generation Conventions}.
787 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
788 -ffixed-@var{reg} -fexceptions @gol
789 -fnon-call-exceptions -funwind-tables @gol
790 -fasynchronous-unwind-tables @gol
791 -finhibit-size-directive -finstrument-functions @gol
792 -fno-common -fno-ident @gol
793 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
794 -fno-jump-tables @gol
795 -frecord-gcc-switches @gol
796 -freg-struct-return -fshort-enums @gol
797 -fshort-double -fshort-wchar @gol
798 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
799 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
800 -fargument-alias -fargument-noalias @gol
801 -fargument-noalias-global -fargument-noalias-anything
802 -fleading-underscore -ftls-model=@var{model} @gol
803 -ftrapv -fwrapv -fbounds-check @gol
808 * Overall Options:: Controlling the kind of output:
809 an executable, object files, assembler files,
810 or preprocessed source.
811 * C Dialect Options:: Controlling the variant of C language compiled.
812 * C++ Dialect Options:: Variations on C++.
813 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
815 * Language Independent Options:: Controlling how diagnostics should be
817 * Warning Options:: How picky should the compiler be?
818 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
819 * Optimize Options:: How much optimization?
820 * Preprocessor Options:: Controlling header files and macro definitions.
821 Also, getting dependency information for Make.
822 * Assembler Options:: Passing options to the assembler.
823 * Link Options:: Specifying libraries and so on.
824 * Directory Options:: Where to find header files and libraries.
825 Where to find the compiler executable files.
826 * Spec Files:: How to pass switches to sub-processes.
827 * Target Options:: Running a cross-compiler, or an old version of GCC.
830 @node Overall Options
831 @section Options Controlling the Kind of Output
833 Compilation can involve up to four stages: preprocessing, compilation
834 proper, assembly and linking, always in that order. GCC is capable of
835 preprocessing and compiling several files either into several
836 assembler input files, or into one assembler input file; then each
837 assembler input file produces an object file, and linking combines all
838 the object files (those newly compiled, and those specified as input)
839 into an executable file.
841 @cindex file name suffix
842 For any given input file, the file name suffix determines what kind of
847 C source code which must be preprocessed.
850 C source code which should not be preprocessed.
853 C++ source code which should not be preprocessed.
856 Objective-C source code. Note that you must link with the @file{libobjc}
857 library to make an Objective-C program work.
860 Objective-C source code which should not be preprocessed.
864 Objective-C++ source code. Note that you must link with the @file{libobjc}
865 library to make an Objective-C++ program work. Note that @samp{.M} refers
866 to a literal capital M@.
869 Objective-C++ source code which should not be preprocessed.
872 C, C++, Objective-C or Objective-C++ header file to be turned into a
877 @itemx @var{file}.cxx
878 @itemx @var{file}.cpp
879 @itemx @var{file}.CPP
880 @itemx @var{file}.c++
882 C++ source code which must be preprocessed. Note that in @samp{.cxx},
883 the last two letters must both be literally @samp{x}. Likewise,
884 @samp{.C} refers to a literal capital C@.
888 Objective-C++ source code which must be preprocessed.
891 Objective-C++ source code which should not be preprocessed.
895 C++ header file to be turned into a precompiled header.
898 @itemx @var{file}.for
899 @itemx @var{file}.FOR
900 Fixed form Fortran source code which should not be preprocessed.
903 @itemx @var{file}.fpp
904 @itemx @var{file}.FPP
905 Fixed form Fortran source code which must be preprocessed (with the traditional
909 @itemx @var{file}.f95
910 Free form Fortran source code which should not be preprocessed.
913 @itemx @var{file}.F95
914 Free form Fortran source code which must be preprocessed (with the
915 traditional preprocessor).
917 @c FIXME: Descriptions of Java file types.
924 Ada source code file which contains a library unit declaration (a
925 declaration of a package, subprogram, or generic, or a generic
926 instantiation), or a library unit renaming declaration (a package,
927 generic, or subprogram renaming declaration). Such files are also
930 @itemx @var{file}.adb
931 Ada source code file containing a library unit body (a subprogram or
932 package body). Such files are also called @dfn{bodies}.
934 @c GCC also knows about some suffixes for languages not yet included:
945 Assembler code which must be preprocessed.
948 An object file to be fed straight into linking.
949 Any file name with no recognized suffix is treated this way.
953 You can specify the input language explicitly with the @option{-x} option:
956 @item -x @var{language}
957 Specify explicitly the @var{language} for the following input files
958 (rather than letting the compiler choose a default based on the file
959 name suffix). This option applies to all following input files until
960 the next @option{-x} option. Possible values for @var{language} are:
962 c c-header c-cpp-output
963 c++ c++-header c++-cpp-output
964 objective-c objective-c-header objective-c-cpp-output
965 objective-c++ objective-c++-header objective-c++-cpp-output
966 assembler assembler-with-cpp
974 Turn off any specification of a language, so that subsequent files are
975 handled according to their file name suffixes (as they are if @option{-x}
976 has not been used at all).
978 @item -pass-exit-codes
979 @opindex pass-exit-codes
980 Normally the @command{gcc} program will exit with the code of 1 if any
981 phase of the compiler returns a non-success return code. If you specify
982 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
983 numerically highest error produced by any phase that returned an error
984 indication. The C, C++, and Fortran frontends return 4, if an internal
985 compiler error is encountered.
988 If you only want some of the stages of compilation, you can use
989 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
990 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
991 @command{gcc} is to stop. Note that some combinations (for example,
992 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
997 Compile or assemble the source files, but do not link. The linking
998 stage simply is not done. The ultimate output is in the form of an
999 object file for each source file.
1001 By default, the object file name for a source file is made by replacing
1002 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1004 Unrecognized input files, not requiring compilation or assembly, are
1009 Stop after the stage of compilation proper; do not assemble. The output
1010 is in the form of an assembler code file for each non-assembler input
1013 By default, the assembler file name for a source file is made by
1014 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1016 Input files that don't require compilation are ignored.
1020 Stop after the preprocessing stage; do not run the compiler proper. The
1021 output is in the form of preprocessed source code, which is sent to the
1024 Input files which don't require preprocessing are ignored.
1026 @cindex output file option
1029 Place output in file @var{file}. This applies regardless to whatever
1030 sort of output is being produced, whether it be an executable file,
1031 an object file, an assembler file or preprocessed C code.
1033 If @option{-o} is not specified, the default is to put an executable
1034 file in @file{a.out}, the object file for
1035 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1036 assembler file in @file{@var{source}.s}, a precompiled header file in
1037 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1042 Print (on standard error output) the commands executed to run the stages
1043 of compilation. Also print the version number of the compiler driver
1044 program and of the preprocessor and the compiler proper.
1048 Like @option{-v} except the commands are not executed and all command
1049 arguments are quoted. This is useful for shell scripts to capture the
1050 driver-generated command lines.
1054 Use pipes rather than temporary files for communication between the
1055 various stages of compilation. This fails to work on some systems where
1056 the assembler is unable to read from a pipe; but the GNU assembler has
1061 If you are compiling multiple source files, this option tells the driver
1062 to pass all the source files to the compiler at once (for those
1063 languages for which the compiler can handle this). This will allow
1064 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1065 language for which this is supported is C@. If you pass source files for
1066 multiple languages to the driver, using this option, the driver will invoke
1067 the compiler(s) that support IMA once each, passing each compiler all the
1068 source files appropriate for it. For those languages that do not support
1069 IMA this option will be ignored, and the compiler will be invoked once for
1070 each source file in that language. If you use this option in conjunction
1071 with @option{-save-temps}, the compiler will generate multiple
1073 (one for each source file), but only one (combined) @file{.o} or
1078 Print (on the standard output) a description of the command line options
1079 understood by @command{gcc}. If the @option{-v} option is also specified
1080 then @option{--help} will also be passed on to the various processes
1081 invoked by @command{gcc}, so that they can display the command line options
1082 they accept. If the @option{-Wextra} option has also been specified
1083 (prior to the @option{--help} option), then command line options which
1084 have no documentation associated with them will also be displayed.
1087 @opindex target-help
1088 Print (on the standard output) a description of target-specific command
1089 line options for each tool.
1091 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1092 Print (on the standard output) a description of the command line
1093 options understood by the compiler that fit into a specific class.
1094 The class can be one of @var{optimizers}, @var{warnings}, @var{target}
1098 @item @var{optimizers}
1099 This will display all of the optimization options supported by the
1102 @item @var{warnings}
1103 This will display all of the options controlling warning messages
1104 produced by the compiler.
1107 This will display target-specific options. Unlike the
1108 @option{--target-help} option however, target-specific options of the
1109 linker and assembler will not be displayed. This is because those
1110 tools do not currently support the extended @option{--help=} syntax.
1113 This will display the values recognized by the @option{--param}
1117 It is possible to further refine the output of the @option{--help=}
1118 option by adding a comma separated list of qualifiers after the
1119 class. These can be any from the following list:
1123 Display only those options which are undocumented.
1126 Display options which take an argument that appears after an equal
1127 sign in the same continuous piece of text, such as:
1128 @samp{--help=target}.
1131 Display options which take an argument that appears as a separate word
1132 following the original option, such as: @samp{-o output-file}.
1135 Thus for example to display all the undocumented target-specific
1136 switches supported by the compiler the following can be used:
1139 --help=target,undocumented
1142 The sense of a qualifier can be inverted by prefixing it with the
1143 @var{^} character, so for example to display all binary warning
1144 options (i.e. ones that are either on or off and that do not take an
1145 argument), which have a description the following can be used:
1148 --help=warnings,^joined,^undocumented
1151 A class can also be used as a qualifier, although this usually
1152 restricts the output by so much that there is nothing to display. One
1153 case where it does work however is when one of the classes is
1154 @var{target}. So for example to display all the target-specific
1155 optimization options the following can be used:
1158 --help=target,optimizers
1161 The @option{--help=} option can be repeated on the command line. Each
1162 successive use will display its requested class of options, skipping
1163 those that have already been displayed.
1165 If the @option{-Q} option appears on the command line before the
1166 @option{--help=} option, then the descriptive text displayed by
1167 @option{--help=} is changed. Instead of describing the displayed
1168 options, an indication is given as to whether the option is enabled,
1169 disabled or set to a specific value (assuming that the compiler
1170 knows this at the point where the @option{--help=} option is used).
1172 Here is a truncated example from the ARM port of @command{gcc}:
1175 % gcc -Q -mabi=2 --help=target -c
1176 The following options are target specific:
1178 -mabort-on-noreturn [disabled]
1182 The output is sensitive to the effects of previous command line
1183 options, so for example it is possible to find out which optimizations
1184 are enabled at @option{-O2} by using:
1187 -O2 --help=optimizers
1190 Alternatively you can discover which binary optimizations are enabled
1191 by @option{-O3} by using:
1194 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1195 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1196 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1201 Display the version number and copyrights of the invoked GCC@.
1203 @include @value{srcdir}/../libiberty/at-file.texi
1207 @section Compiling C++ Programs
1209 @cindex suffixes for C++ source
1210 @cindex C++ source file suffixes
1211 C++ source files conventionally use one of the suffixes @samp{.C},
1212 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1213 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1214 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1215 files with these names and compiles them as C++ programs even if you
1216 call the compiler the same way as for compiling C programs (usually
1217 with the name @command{gcc}).
1221 However, the use of @command{gcc} does not add the C++ library.
1222 @command{g++} is a program that calls GCC and treats @samp{.c},
1223 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1224 files unless @option{-x} is used, and automatically specifies linking
1225 against the C++ library. This program is also useful when
1226 precompiling a C header file with a @samp{.h} extension for use in C++
1227 compilations. On many systems, @command{g++} is also installed with
1228 the name @command{c++}.
1230 @cindex invoking @command{g++}
1231 When you compile C++ programs, you may specify many of the same
1232 command-line options that you use for compiling programs in any
1233 language; or command-line options meaningful for C and related
1234 languages; or options that are meaningful only for C++ programs.
1235 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1236 explanations of options for languages related to C@.
1237 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1238 explanations of options that are meaningful only for C++ programs.
1240 @node C Dialect Options
1241 @section Options Controlling C Dialect
1242 @cindex dialect options
1243 @cindex language dialect options
1244 @cindex options, dialect
1246 The following options control the dialect of C (or languages derived
1247 from C, such as C++, Objective-C and Objective-C++) that the compiler
1251 @cindex ANSI support
1255 In C mode, support all ISO C90 programs. In C++ mode,
1256 remove GNU extensions that conflict with ISO C++.
1258 This turns off certain features of GCC that are incompatible with ISO
1259 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1260 such as the @code{asm} and @code{typeof} keywords, and
1261 predefined macros such as @code{unix} and @code{vax} that identify the
1262 type of system you are using. It also enables the undesirable and
1263 rarely used ISO trigraph feature. For the C compiler,
1264 it disables recognition of C++ style @samp{//} comments as well as
1265 the @code{inline} keyword.
1267 The alternate keywords @code{__asm__}, @code{__extension__},
1268 @code{__inline__} and @code{__typeof__} continue to work despite
1269 @option{-ansi}. You would not want to use them in an ISO C program, of
1270 course, but it is useful to put them in header files that might be included
1271 in compilations done with @option{-ansi}. Alternate predefined macros
1272 such as @code{__unix__} and @code{__vax__} are also available, with or
1273 without @option{-ansi}.
1275 The @option{-ansi} option does not cause non-ISO programs to be
1276 rejected gratuitously. For that, @option{-pedantic} is required in
1277 addition to @option{-ansi}. @xref{Warning Options}.
1279 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1280 option is used. Some header files may notice this macro and refrain
1281 from declaring certain functions or defining certain macros that the
1282 ISO standard doesn't call for; this is to avoid interfering with any
1283 programs that might use these names for other things.
1285 Functions which would normally be built in but do not have semantics
1286 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1287 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1288 built-in functions provided by GCC}, for details of the functions
1293 Determine the language standard. This option is currently only
1294 supported when compiling C or C++. A value for this option must be
1295 provided; possible values are
1300 ISO C90 (same as @option{-ansi}).
1302 @item iso9899:199409
1303 ISO C90 as modified in amendment 1.
1309 ISO C99. Note that this standard is not yet fully supported; see
1310 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1311 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1314 Default, ISO C90 plus GNU extensions (including some C99 features).
1318 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1319 this will become the default. The name @samp{gnu9x} is deprecated.
1322 The 1998 ISO C++ standard plus amendments.
1325 The same as @option{-std=c++98} plus GNU extensions. This is the
1326 default for C++ code.
1329 The working draft of the upcoming ISO C++0x standard. This option
1330 enables experimental features that are likely to be included in
1331 C++0x. The working draft is constantly changing, and any feature that is
1332 enabled by this flag may be removed from future versions of GCC if it is
1333 not part of the C++0x standard.
1336 The same as @option{-std=c++0x} plus GNU extensions. As with
1337 @option{-std=c++0x}, this option enables experimental features that may
1338 be removed in future versions of GCC.
1341 Even when this option is not specified, you can still use some of the
1342 features of newer standards in so far as they do not conflict with
1343 previous C standards. For example, you may use @code{__restrict__} even
1344 when @option{-std=c99} is not specified.
1346 The @option{-std} options specifying some version of ISO C have the same
1347 effects as @option{-ansi}, except that features that were not in ISO C90
1348 but are in the specified version (for example, @samp{//} comments and
1349 the @code{inline} keyword in ISO C99) are not disabled.
1351 @xref{Standards,,Language Standards Supported by GCC}, for details of
1352 these standard versions.
1354 @item -aux-info @var{filename}
1356 Output to the given filename prototyped declarations for all functions
1357 declared and/or defined in a translation unit, including those in header
1358 files. This option is silently ignored in any language other than C@.
1360 Besides declarations, the file indicates, in comments, the origin of
1361 each declaration (source file and line), whether the declaration was
1362 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1363 @samp{O} for old, respectively, in the first character after the line
1364 number and the colon), and whether it came from a declaration or a
1365 definition (@samp{C} or @samp{F}, respectively, in the following
1366 character). In the case of function definitions, a K&R-style list of
1367 arguments followed by their declarations is also provided, inside
1368 comments, after the declaration.
1372 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1373 keyword, so that code can use these words as identifiers. You can use
1374 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1375 instead. @option{-ansi} implies @option{-fno-asm}.
1377 In C++, this switch only affects the @code{typeof} keyword, since
1378 @code{asm} and @code{inline} are standard keywords. You may want to
1379 use the @option{-fno-gnu-keywords} flag instead, which has the same
1380 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1381 switch only affects the @code{asm} and @code{typeof} keywords, since
1382 @code{inline} is a standard keyword in ISO C99.
1385 @itemx -fno-builtin-@var{function}
1386 @opindex fno-builtin
1387 @cindex built-in functions
1388 Don't recognize built-in functions that do not begin with
1389 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1390 functions provided by GCC}, for details of the functions affected,
1391 including those which are not built-in functions when @option{-ansi} or
1392 @option{-std} options for strict ISO C conformance are used because they
1393 do not have an ISO standard meaning.
1395 GCC normally generates special code to handle certain built-in functions
1396 more efficiently; for instance, calls to @code{alloca} may become single
1397 instructions that adjust the stack directly, and calls to @code{memcpy}
1398 may become inline copy loops. The resulting code is often both smaller
1399 and faster, but since the function calls no longer appear as such, you
1400 cannot set a breakpoint on those calls, nor can you change the behavior
1401 of the functions by linking with a different library. In addition,
1402 when a function is recognized as a built-in function, GCC may use
1403 information about that function to warn about problems with calls to
1404 that function, or to generate more efficient code, even if the
1405 resulting code still contains calls to that function. For example,
1406 warnings are given with @option{-Wformat} for bad calls to
1407 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1408 known not to modify global memory.
1410 With the @option{-fno-builtin-@var{function}} option
1411 only the built-in function @var{function} is
1412 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1413 function is named this is not built-in in this version of GCC, this
1414 option is ignored. There is no corresponding
1415 @option{-fbuiltin-@var{function}} option; if you wish to enable
1416 built-in functions selectively when using @option{-fno-builtin} or
1417 @option{-ffreestanding}, you may define macros such as:
1420 #define abs(n) __builtin_abs ((n))
1421 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1426 @cindex hosted environment
1428 Assert that compilation takes place in a hosted environment. This implies
1429 @option{-fbuiltin}. A hosted environment is one in which the
1430 entire standard library is available, and in which @code{main} has a return
1431 type of @code{int}. Examples are nearly everything except a kernel.
1432 This is equivalent to @option{-fno-freestanding}.
1434 @item -ffreestanding
1435 @opindex ffreestanding
1436 @cindex hosted environment
1438 Assert that compilation takes place in a freestanding environment. This
1439 implies @option{-fno-builtin}. A freestanding environment
1440 is one in which the standard library may not exist, and program startup may
1441 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1442 This is equivalent to @option{-fno-hosted}.
1444 @xref{Standards,,Language Standards Supported by GCC}, for details of
1445 freestanding and hosted environments.
1449 @cindex openmp parallel
1450 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1451 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1452 compiler generates parallel code according to the OpenMP Application
1453 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1455 @item -fms-extensions
1456 @opindex fms-extensions
1457 Accept some non-standard constructs used in Microsoft header files.
1459 Some cases of unnamed fields in structures and unions are only
1460 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1461 fields within structs/unions}, for details.
1465 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1466 options for strict ISO C conformance) implies @option{-trigraphs}.
1468 @item -no-integrated-cpp
1469 @opindex no-integrated-cpp
1470 Performs a compilation in two passes: preprocessing and compiling. This
1471 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1472 @option{-B} option. The user supplied compilation step can then add in
1473 an additional preprocessing step after normal preprocessing but before
1474 compiling. The default is to use the integrated cpp (internal cpp)
1476 The semantics of this option will change if "cc1", "cc1plus", and
1477 "cc1obj" are merged.
1479 @cindex traditional C language
1480 @cindex C language, traditional
1482 @itemx -traditional-cpp
1483 @opindex traditional-cpp
1484 @opindex traditional
1485 Formerly, these options caused GCC to attempt to emulate a pre-standard
1486 C compiler. They are now only supported with the @option{-E} switch.
1487 The preprocessor continues to support a pre-standard mode. See the GNU
1488 CPP manual for details.
1490 @item -fcond-mismatch
1491 @opindex fcond-mismatch
1492 Allow conditional expressions with mismatched types in the second and
1493 third arguments. The value of such an expression is void. This option
1494 is not supported for C++.
1496 @item -flax-vector-conversions
1497 @opindex flax-vector-conversions
1498 Allow implicit conversions between vectors with differing numbers of
1499 elements and/or incompatible element types. This option should not be
1502 @item -funsigned-char
1503 @opindex funsigned-char
1504 Let the type @code{char} be unsigned, like @code{unsigned char}.
1506 Each kind of machine has a default for what @code{char} should
1507 be. It is either like @code{unsigned char} by default or like
1508 @code{signed char} by default.
1510 Ideally, a portable program should always use @code{signed char} or
1511 @code{unsigned char} when it depends on the signedness of an object.
1512 But many programs have been written to use plain @code{char} and
1513 expect it to be signed, or expect it to be unsigned, depending on the
1514 machines they were written for. This option, and its inverse, let you
1515 make such a program work with the opposite default.
1517 The type @code{char} is always a distinct type from each of
1518 @code{signed char} or @code{unsigned char}, even though its behavior
1519 is always just like one of those two.
1522 @opindex fsigned-char
1523 Let the type @code{char} be signed, like @code{signed char}.
1525 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1526 the negative form of @option{-funsigned-char}. Likewise, the option
1527 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1529 @item -fsigned-bitfields
1530 @itemx -funsigned-bitfields
1531 @itemx -fno-signed-bitfields
1532 @itemx -fno-unsigned-bitfields
1533 @opindex fsigned-bitfields
1534 @opindex funsigned-bitfields
1535 @opindex fno-signed-bitfields
1536 @opindex fno-unsigned-bitfields
1537 These options control whether a bit-field is signed or unsigned, when the
1538 declaration does not use either @code{signed} or @code{unsigned}. By
1539 default, such a bit-field is signed, because this is consistent: the
1540 basic integer types such as @code{int} are signed types.
1543 @node C++ Dialect Options
1544 @section Options Controlling C++ Dialect
1546 @cindex compiler options, C++
1547 @cindex C++ options, command line
1548 @cindex options, C++
1549 This section describes the command-line options that are only meaningful
1550 for C++ programs; but you can also use most of the GNU compiler options
1551 regardless of what language your program is in. For example, you
1552 might compile a file @code{firstClass.C} like this:
1555 g++ -g -frepo -O -c firstClass.C
1559 In this example, only @option{-frepo} is an option meant
1560 only for C++ programs; you can use the other options with any
1561 language supported by GCC@.
1563 Here is a list of options that are @emph{only} for compiling C++ programs:
1567 @item -fabi-version=@var{n}
1568 @opindex fabi-version
1569 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1570 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1571 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1572 the version that conforms most closely to the C++ ABI specification.
1573 Therefore, the ABI obtained using version 0 will change as ABI bugs
1576 The default is version 2.
1578 @item -fno-access-control
1579 @opindex fno-access-control
1580 Turn off all access checking. This switch is mainly useful for working
1581 around bugs in the access control code.
1585 Check that the pointer returned by @code{operator new} is non-null
1586 before attempting to modify the storage allocated. This check is
1587 normally unnecessary because the C++ standard specifies that
1588 @code{operator new} will only return @code{0} if it is declared
1589 @samp{throw()}, in which case the compiler will always check the
1590 return value even without this option. In all other cases, when
1591 @code{operator new} has a non-empty exception specification, memory
1592 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1593 @samp{new (nothrow)}.
1595 @item -fconserve-space
1596 @opindex fconserve-space
1597 Put uninitialized or runtime-initialized global variables into the
1598 common segment, as C does. This saves space in the executable at the
1599 cost of not diagnosing duplicate definitions. If you compile with this
1600 flag and your program mysteriously crashes after @code{main()} has
1601 completed, you may have an object that is being destroyed twice because
1602 two definitions were merged.
1604 This option is no longer useful on most targets, now that support has
1605 been added for putting variables into BSS without making them common.
1607 @item -ffriend-injection
1608 @opindex ffriend-injection
1609 Inject friend functions into the enclosing namespace, so that they are
1610 visible outside the scope of the class in which they are declared.
1611 Friend functions were documented to work this way in the old Annotated
1612 C++ Reference Manual, and versions of G++ before 4.1 always worked
1613 that way. However, in ISO C++ a friend function which is not declared
1614 in an enclosing scope can only be found using argument dependent
1615 lookup. This option causes friends to be injected as they were in
1618 This option is for compatibility, and may be removed in a future
1621 @item -fno-elide-constructors
1622 @opindex fno-elide-constructors
1623 The C++ standard allows an implementation to omit creating a temporary
1624 which is only used to initialize another object of the same type.
1625 Specifying this option disables that optimization, and forces G++ to
1626 call the copy constructor in all cases.
1628 @item -fno-enforce-eh-specs
1629 @opindex fno-enforce-eh-specs
1630 Don't generate code to check for violation of exception specifications
1631 at runtime. This option violates the C++ standard, but may be useful
1632 for reducing code size in production builds, much like defining
1633 @samp{NDEBUG}. This does not give user code permission to throw
1634 exceptions in violation of the exception specifications; the compiler
1635 will still optimize based on the specifications, so throwing an
1636 unexpected exception will result in undefined behavior.
1639 @itemx -fno-for-scope
1641 @opindex fno-for-scope
1642 If @option{-ffor-scope} is specified, the scope of variables declared in
1643 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1644 as specified by the C++ standard.
1645 If @option{-fno-for-scope} is specified, the scope of variables declared in
1646 a @i{for-init-statement} extends to the end of the enclosing scope,
1647 as was the case in old versions of G++, and other (traditional)
1648 implementations of C++.
1650 The default if neither flag is given to follow the standard,
1651 but to allow and give a warning for old-style code that would
1652 otherwise be invalid, or have different behavior.
1654 @item -fno-gnu-keywords
1655 @opindex fno-gnu-keywords
1656 Do not recognize @code{typeof} as a keyword, so that code can use this
1657 word as an identifier. You can use the keyword @code{__typeof__} instead.
1658 @option{-ansi} implies @option{-fno-gnu-keywords}.
1660 @item -fno-implicit-templates
1661 @opindex fno-implicit-templates
1662 Never emit code for non-inline templates which are instantiated
1663 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1664 @xref{Template Instantiation}, for more information.
1666 @item -fno-implicit-inline-templates
1667 @opindex fno-implicit-inline-templates
1668 Don't emit code for implicit instantiations of inline templates, either.
1669 The default is to handle inlines differently so that compiles with and
1670 without optimization will need the same set of explicit instantiations.
1672 @item -fno-implement-inlines
1673 @opindex fno-implement-inlines
1674 To save space, do not emit out-of-line copies of inline functions
1675 controlled by @samp{#pragma implementation}. This will cause linker
1676 errors if these functions are not inlined everywhere they are called.
1678 @item -fms-extensions
1679 @opindex fms-extensions
1680 Disable pedantic warnings about constructs used in MFC, such as implicit
1681 int and getting a pointer to member function via non-standard syntax.
1683 @item -fno-nonansi-builtins
1684 @opindex fno-nonansi-builtins
1685 Disable built-in declarations of functions that are not mandated by
1686 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1687 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1689 @item -fno-operator-names
1690 @opindex fno-operator-names
1691 Do not treat the operator name keywords @code{and}, @code{bitand},
1692 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1693 synonyms as keywords.
1695 @item -fno-optional-diags
1696 @opindex fno-optional-diags
1697 Disable diagnostics that the standard says a compiler does not need to
1698 issue. Currently, the only such diagnostic issued by G++ is the one for
1699 a name having multiple meanings within a class.
1702 @opindex fpermissive
1703 Downgrade some diagnostics about nonconformant code from errors to
1704 warnings. Thus, using @option{-fpermissive} will allow some
1705 nonconforming code to compile.
1709 Enable automatic template instantiation at link time. This option also
1710 implies @option{-fno-implicit-templates}. @xref{Template
1711 Instantiation}, for more information.
1715 Disable generation of information about every class with virtual
1716 functions for use by the C++ runtime type identification features
1717 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1718 of the language, you can save some space by using this flag. Note that
1719 exception handling uses the same information, but it will generate it as
1720 needed. The @samp{dynamic_cast} operator can still be used for casts that
1721 do not require runtime type information, i.e. casts to @code{void *} or to
1722 unambiguous base classes.
1726 Emit statistics about front-end processing at the end of the compilation.
1727 This information is generally only useful to the G++ development team.
1729 @item -ftemplate-depth-@var{n}
1730 @opindex ftemplate-depth
1731 Set the maximum instantiation depth for template classes to @var{n}.
1732 A limit on the template instantiation depth is needed to detect
1733 endless recursions during template class instantiation. ANSI/ISO C++
1734 conforming programs must not rely on a maximum depth greater than 17.
1736 @item -fno-threadsafe-statics
1737 @opindex fno-threadsafe-statics
1738 Do not emit the extra code to use the routines specified in the C++
1739 ABI for thread-safe initialization of local statics. You can use this
1740 option to reduce code size slightly in code that doesn't need to be
1743 @item -fuse-cxa-atexit
1744 @opindex fuse-cxa-atexit
1745 Register destructors for objects with static storage duration with the
1746 @code{__cxa_atexit} function rather than the @code{atexit} function.
1747 This option is required for fully standards-compliant handling of static
1748 destructors, but will only work if your C library supports
1749 @code{__cxa_atexit}.
1751 @item -fno-use-cxa-get-exception-ptr
1752 @opindex fno-use-cxa-get-exception-ptr
1753 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1754 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1755 if the runtime routine is not available.
1757 @item -fvisibility-inlines-hidden
1758 @opindex fvisibility-inlines-hidden
1759 This switch declares that the user does not attempt to compare
1760 pointers to inline methods where the addresses of the two functions
1761 were taken in different shared objects.
1763 The effect of this is that GCC may, effectively, mark inline methods with
1764 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1765 appear in the export table of a DSO and do not require a PLT indirection
1766 when used within the DSO@. Enabling this option can have a dramatic effect
1767 on load and link times of a DSO as it massively reduces the size of the
1768 dynamic export table when the library makes heavy use of templates.
1770 The behavior of this switch is not quite the same as marking the
1771 methods as hidden directly, because it does not affect static variables
1772 local to the function or cause the compiler to deduce that
1773 the function is defined in only one shared object.
1775 You may mark a method as having a visibility explicitly to negate the
1776 effect of the switch for that method. For example, if you do want to
1777 compare pointers to a particular inline method, you might mark it as
1778 having default visibility. Marking the enclosing class with explicit
1779 visibility will have no effect.
1781 Explicitly instantiated inline methods are unaffected by this option
1782 as their linkage might otherwise cross a shared library boundary.
1783 @xref{Template Instantiation}.
1787 Do not use weak symbol support, even if it is provided by the linker.
1788 By default, G++ will use weak symbols if they are available. This
1789 option exists only for testing, and should not be used by end-users;
1790 it will result in inferior code and has no benefits. This option may
1791 be removed in a future release of G++.
1795 Do not search for header files in the standard directories specific to
1796 C++, but do still search the other standard directories. (This option
1797 is used when building the C++ library.)
1800 In addition, these optimization, warning, and code generation options
1801 have meanings only for C++ programs:
1804 @item -fno-default-inline
1805 @opindex fno-default-inline
1806 Do not assume @samp{inline} for functions defined inside a class scope.
1807 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1808 functions will have linkage like inline functions; they just won't be
1811 @item -Wabi @r{(C++ only)}
1813 Warn when G++ generates code that is probably not compatible with the
1814 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1815 all such cases, there are probably some cases that are not warned about,
1816 even though G++ is generating incompatible code. There may also be
1817 cases where warnings are emitted even though the code that is generated
1820 You should rewrite your code to avoid these warnings if you are
1821 concerned about the fact that code generated by G++ may not be binary
1822 compatible with code generated by other compilers.
1824 The known incompatibilities at this point include:
1829 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1830 pack data into the same byte as a base class. For example:
1833 struct A @{ virtual void f(); int f1 : 1; @};
1834 struct B : public A @{ int f2 : 1; @};
1838 In this case, G++ will place @code{B::f2} into the same byte
1839 as@code{A::f1}; other compilers will not. You can avoid this problem
1840 by explicitly padding @code{A} so that its size is a multiple of the
1841 byte size on your platform; that will cause G++ and other compilers to
1842 layout @code{B} identically.
1845 Incorrect handling of tail-padding for virtual bases. G++ does not use
1846 tail padding when laying out virtual bases. For example:
1849 struct A @{ virtual void f(); char c1; @};
1850 struct B @{ B(); char c2; @};
1851 struct C : public A, public virtual B @{@};
1855 In this case, G++ will not place @code{B} into the tail-padding for
1856 @code{A}; other compilers will. You can avoid this problem by
1857 explicitly padding @code{A} so that its size is a multiple of its
1858 alignment (ignoring virtual base classes); that will cause G++ and other
1859 compilers to layout @code{C} identically.
1862 Incorrect handling of bit-fields with declared widths greater than that
1863 of their underlying types, when the bit-fields appear in a union. For
1867 union U @{ int i : 4096; @};
1871 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1872 union too small by the number of bits in an @code{int}.
1875 Empty classes can be placed at incorrect offsets. For example:
1885 struct C : public B, public A @{@};
1889 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1890 it should be placed at offset zero. G++ mistakenly believes that the
1891 @code{A} data member of @code{B} is already at offset zero.
1894 Names of template functions whose types involve @code{typename} or
1895 template template parameters can be mangled incorrectly.
1898 template <typename Q>
1899 void f(typename Q::X) @{@}
1901 template <template <typename> class Q>
1902 void f(typename Q<int>::X) @{@}
1906 Instantiations of these templates may be mangled incorrectly.
1910 @item -Wctor-dtor-privacy @r{(C++ only)}
1911 @opindex Wctor-dtor-privacy
1912 Warn when a class seems unusable because all the constructors or
1913 destructors in that class are private, and it has neither friends nor
1914 public static member functions.
1916 @item -Wnon-virtual-dtor @r{(C++ only)}
1917 @opindex Wnon-virtual-dtor
1918 Warn when a class appears to be polymorphic, thereby requiring a virtual
1919 destructor, yet it declares a non-virtual one. This warning is also
1920 enabled if -Weffc++ is specified.
1922 @item -Wreorder @r{(C++ only)}
1924 @cindex reordering, warning
1925 @cindex warning for reordering of member initializers
1926 Warn when the order of member initializers given in the code does not
1927 match the order in which they must be executed. For instance:
1933 A(): j (0), i (1) @{ @}
1937 The compiler will rearrange the member initializers for @samp{i}
1938 and @samp{j} to match the declaration order of the members, emitting
1939 a warning to that effect. This warning is enabled by @option{-Wall}.
1942 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1945 @item -Weffc++ @r{(C++ only)}
1947 Warn about violations of the following style guidelines from Scott Meyers'
1948 @cite{Effective C++} book:
1952 Item 11: Define a copy constructor and an assignment operator for classes
1953 with dynamically allocated memory.
1956 Item 12: Prefer initialization to assignment in constructors.
1959 Item 14: Make destructors virtual in base classes.
1962 Item 15: Have @code{operator=} return a reference to @code{*this}.
1965 Item 23: Don't try to return a reference when you must return an object.
1969 Also warn about violations of the following style guidelines from
1970 Scott Meyers' @cite{More Effective C++} book:
1974 Item 6: Distinguish between prefix and postfix forms of increment and
1975 decrement operators.
1978 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1982 When selecting this option, be aware that the standard library
1983 headers do not obey all of these guidelines; use @samp{grep -v}
1984 to filter out those warnings.
1986 @item -Wno-deprecated @r{(C++ only)}
1987 @opindex Wno-deprecated
1988 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1990 @item -Wstrict-null-sentinel @r{(C++ only)}
1991 @opindex Wstrict-null-sentinel
1992 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1993 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1994 to @code{__null}. Although it is a null pointer constant not a null pointer,
1995 it is guaranteed to of the same size as a pointer. But this use is
1996 not portable across different compilers.
1998 @item -Wno-non-template-friend @r{(C++ only)}
1999 @opindex Wno-non-template-friend
2000 Disable warnings when non-templatized friend functions are declared
2001 within a template. Since the advent of explicit template specification
2002 support in G++, if the name of the friend is an unqualified-id (i.e.,
2003 @samp{friend foo(int)}), the C++ language specification demands that the
2004 friend declare or define an ordinary, nontemplate function. (Section
2005 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2006 could be interpreted as a particular specialization of a templatized
2007 function. Because this non-conforming behavior is no longer the default
2008 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2009 check existing code for potential trouble spots and is on by default.
2010 This new compiler behavior can be turned off with
2011 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2012 but disables the helpful warning.
2014 @item -Wold-style-cast @r{(C++ only)}
2015 @opindex Wold-style-cast
2016 Warn if an old-style (C-style) cast to a non-void type is used within
2017 a C++ program. The new-style casts (@samp{dynamic_cast},
2018 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2019 less vulnerable to unintended effects and much easier to search for.
2021 @item -Woverloaded-virtual @r{(C++ only)}
2022 @opindex Woverloaded-virtual
2023 @cindex overloaded virtual fn, warning
2024 @cindex warning for overloaded virtual fn
2025 Warn when a function declaration hides virtual functions from a
2026 base class. For example, in:
2033 struct B: public A @{
2038 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2046 will fail to compile.
2048 @item -Wno-pmf-conversions @r{(C++ only)}
2049 @opindex Wno-pmf-conversions
2050 Disable the diagnostic for converting a bound pointer to member function
2053 @item -Wsign-promo @r{(C++ only)}
2054 @opindex Wsign-promo
2055 Warn when overload resolution chooses a promotion from unsigned or
2056 enumerated type to a signed type, over a conversion to an unsigned type of
2057 the same size. Previous versions of G++ would try to preserve
2058 unsignedness, but the standard mandates the current behavior.
2063 A& operator = (int);
2073 In this example, G++ will synthesize a default @samp{A& operator =
2074 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2077 @node Objective-C and Objective-C++ Dialect Options
2078 @section Options Controlling Objective-C and Objective-C++ Dialects
2080 @cindex compiler options, Objective-C and Objective-C++
2081 @cindex Objective-C and Objective-C++ options, command line
2082 @cindex options, Objective-C and Objective-C++
2083 (NOTE: This manual does not describe the Objective-C and Objective-C++
2084 languages themselves. See @xref{Standards,,Language Standards
2085 Supported by GCC}, for references.)
2087 This section describes the command-line options that are only meaningful
2088 for Objective-C and Objective-C++ programs, but you can also use most of
2089 the language-independent GNU compiler options.
2090 For example, you might compile a file @code{some_class.m} like this:
2093 gcc -g -fgnu-runtime -O -c some_class.m
2097 In this example, @option{-fgnu-runtime} is an option meant only for
2098 Objective-C and Objective-C++ programs; you can use the other options with
2099 any language supported by GCC@.
2101 Note that since Objective-C is an extension of the C language, Objective-C
2102 compilations may also use options specific to the C front-end (e.g.,
2103 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2104 C++-specific options (e.g., @option{-Wabi}).
2106 Here is a list of options that are @emph{only} for compiling Objective-C
2107 and Objective-C++ programs:
2110 @item -fconstant-string-class=@var{class-name}
2111 @opindex fconstant-string-class
2112 Use @var{class-name} as the name of the class to instantiate for each
2113 literal string specified with the syntax @code{@@"@dots{}"}. The default
2114 class name is @code{NXConstantString} if the GNU runtime is being used, and
2115 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2116 @option{-fconstant-cfstrings} option, if also present, will override the
2117 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2118 to be laid out as constant CoreFoundation strings.
2121 @opindex fgnu-runtime
2122 Generate object code compatible with the standard GNU Objective-C
2123 runtime. This is the default for most types of systems.
2125 @item -fnext-runtime
2126 @opindex fnext-runtime
2127 Generate output compatible with the NeXT runtime. This is the default
2128 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2129 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2132 @item -fno-nil-receivers
2133 @opindex fno-nil-receivers
2134 Assume that all Objective-C message dispatches (e.g.,
2135 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2136 is not @code{nil}. This allows for more efficient entry points in the runtime
2137 to be used. Currently, this option is only available in conjunction with
2138 the NeXT runtime on Mac OS X 10.3 and later.
2140 @item -fobjc-call-cxx-cdtors
2141 @opindex fobjc-call-cxx-cdtors
2142 For each Objective-C class, check if any of its instance variables is a
2143 C++ object with a non-trivial default constructor. If so, synthesize a
2144 special @code{- (id) .cxx_construct} instance method that will run
2145 non-trivial default constructors on any such instance variables, in order,
2146 and then return @code{self}. Similarly, check if any instance variable
2147 is a C++ object with a non-trivial destructor, and if so, synthesize a
2148 special @code{- (void) .cxx_destruct} method that will run
2149 all such default destructors, in reverse order.
2151 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2152 thusly generated will only operate on instance variables declared in the
2153 current Objective-C class, and not those inherited from superclasses. It
2154 is the responsibility of the Objective-C runtime to invoke all such methods
2155 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2156 will be invoked by the runtime immediately after a new object
2157 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2158 be invoked immediately before the runtime deallocates an object instance.
2160 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2161 support for invoking the @code{- (id) .cxx_construct} and
2162 @code{- (void) .cxx_destruct} methods.
2164 @item -fobjc-direct-dispatch
2165 @opindex fobjc-direct-dispatch
2166 Allow fast jumps to the message dispatcher. On Darwin this is
2167 accomplished via the comm page.
2169 @item -fobjc-exceptions
2170 @opindex fobjc-exceptions
2171 Enable syntactic support for structured exception handling in Objective-C,
2172 similar to what is offered by C++ and Java. This option is
2173 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2182 @@catch (AnObjCClass *exc) @{
2189 @@catch (AnotherClass *exc) @{
2192 @@catch (id allOthers) @{
2202 The @code{@@throw} statement may appear anywhere in an Objective-C or
2203 Objective-C++ program; when used inside of a @code{@@catch} block, the
2204 @code{@@throw} may appear without an argument (as shown above), in which case
2205 the object caught by the @code{@@catch} will be rethrown.
2207 Note that only (pointers to) Objective-C objects may be thrown and
2208 caught using this scheme. When an object is thrown, it will be caught
2209 by the nearest @code{@@catch} clause capable of handling objects of that type,
2210 analogously to how @code{catch} blocks work in C++ and Java. A
2211 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2212 any and all Objective-C exceptions not caught by previous @code{@@catch}
2215 The @code{@@finally} clause, if present, will be executed upon exit from the
2216 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2217 regardless of whether any exceptions are thrown, caught or rethrown
2218 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2219 of the @code{finally} clause in Java.
2221 There are several caveats to using the new exception mechanism:
2225 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2226 idioms provided by the @code{NSException} class, the new
2227 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2228 systems, due to additional functionality needed in the (NeXT) Objective-C
2232 As mentioned above, the new exceptions do not support handling
2233 types other than Objective-C objects. Furthermore, when used from
2234 Objective-C++, the Objective-C exception model does not interoperate with C++
2235 exceptions at this time. This means you cannot @code{@@throw} an exception
2236 from Objective-C and @code{catch} it in C++, or vice versa
2237 (i.e., @code{throw @dots{} @@catch}).
2240 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2241 blocks for thread-safe execution:
2244 @@synchronized (ObjCClass *guard) @{
2249 Upon entering the @code{@@synchronized} block, a thread of execution shall
2250 first check whether a lock has been placed on the corresponding @code{guard}
2251 object by another thread. If it has, the current thread shall wait until
2252 the other thread relinquishes its lock. Once @code{guard} becomes available,
2253 the current thread will place its own lock on it, execute the code contained in
2254 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2255 making @code{guard} available to other threads).
2257 Unlike Java, Objective-C does not allow for entire methods to be marked
2258 @code{@@synchronized}. Note that throwing exceptions out of
2259 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2260 to be unlocked properly.
2264 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2266 @item -freplace-objc-classes
2267 @opindex freplace-objc-classes
2268 Emit a special marker instructing @command{ld(1)} not to statically link in
2269 the resulting object file, and allow @command{dyld(1)} to load it in at
2270 run time instead. This is used in conjunction with the Fix-and-Continue
2271 debugging mode, where the object file in question may be recompiled and
2272 dynamically reloaded in the course of program execution, without the need
2273 to restart the program itself. Currently, Fix-and-Continue functionality
2274 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2279 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2280 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2281 compile time) with static class references that get initialized at load time,
2282 which improves run-time performance. Specifying the @option{-fzero-link} flag
2283 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2284 to be retained. This is useful in Zero-Link debugging mode, since it allows
2285 for individual class implementations to be modified during program execution.
2289 Dump interface declarations for all classes seen in the source file to a
2290 file named @file{@var{sourcename}.decl}.
2292 @item -Wassign-intercept
2293 @opindex Wassign-intercept
2294 Warn whenever an Objective-C assignment is being intercepted by the
2298 @opindex Wno-protocol
2299 If a class is declared to implement a protocol, a warning is issued for
2300 every method in the protocol that is not implemented by the class. The
2301 default behavior is to issue a warning for every method not explicitly
2302 implemented in the class, even if a method implementation is inherited
2303 from the superclass. If you use the @option{-Wno-protocol} option, then
2304 methods inherited from the superclass are considered to be implemented,
2305 and no warning is issued for them.
2309 Warn if multiple methods of different types for the same selector are
2310 found during compilation. The check is performed on the list of methods
2311 in the final stage of compilation. Additionally, a check is performed
2312 for each selector appearing in a @code{@@selector(@dots{})}
2313 expression, and a corresponding method for that selector has been found
2314 during compilation. Because these checks scan the method table only at
2315 the end of compilation, these warnings are not produced if the final
2316 stage of compilation is not reached, for example because an error is
2317 found during compilation, or because the @option{-fsyntax-only} option is
2320 @item -Wstrict-selector-match
2321 @opindex Wstrict-selector-match
2322 Warn if multiple methods with differing argument and/or return types are
2323 found for a given selector when attempting to send a message using this
2324 selector to a receiver of type @code{id} or @code{Class}. When this flag
2325 is off (which is the default behavior), the compiler will omit such warnings
2326 if any differences found are confined to types which share the same size
2329 @item -Wundeclared-selector
2330 @opindex Wundeclared-selector
2331 Warn if a @code{@@selector(@dots{})} expression referring to an
2332 undeclared selector is found. A selector is considered undeclared if no
2333 method with that name has been declared before the
2334 @code{@@selector(@dots{})} expression, either explicitly in an
2335 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2336 an @code{@@implementation} section. This option always performs its
2337 checks as soon as a @code{@@selector(@dots{})} expression is found,
2338 while @option{-Wselector} only performs its checks in the final stage of
2339 compilation. This also enforces the coding style convention
2340 that methods and selectors must be declared before being used.
2342 @item -print-objc-runtime-info
2343 @opindex print-objc-runtime-info
2344 Generate C header describing the largest structure that is passed by
2349 @node Language Independent Options
2350 @section Options to Control Diagnostic Messages Formatting
2351 @cindex options to control diagnostics formatting
2352 @cindex diagnostic messages
2353 @cindex message formatting
2355 Traditionally, diagnostic messages have been formatted irrespective of
2356 the output device's aspect (e.g.@: its width, @dots{}). The options described
2357 below can be used to control the diagnostic messages formatting
2358 algorithm, e.g.@: how many characters per line, how often source location
2359 information should be reported. Right now, only the C++ front end can
2360 honor these options. However it is expected, in the near future, that
2361 the remaining front ends would be able to digest them correctly.
2364 @item -fmessage-length=@var{n}
2365 @opindex fmessage-length
2366 Try to format error messages so that they fit on lines of about @var{n}
2367 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2368 the front ends supported by GCC@. If @var{n} is zero, then no
2369 line-wrapping will be done; each error message will appear on a single
2372 @opindex fdiagnostics-show-location
2373 @item -fdiagnostics-show-location=once
2374 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2375 reporter to emit @emph{once} source location information; that is, in
2376 case the message is too long to fit on a single physical line and has to
2377 be wrapped, the source location won't be emitted (as prefix) again,
2378 over and over, in subsequent continuation lines. This is the default
2381 @item -fdiagnostics-show-location=every-line
2382 Only meaningful in line-wrapping mode. Instructs the diagnostic
2383 messages reporter to emit the same source location information (as
2384 prefix) for physical lines that result from the process of breaking
2385 a message which is too long to fit on a single line.
2387 @item -fdiagnostics-show-option
2388 @opindex fdiagnostics-show-option
2389 This option instructs the diagnostic machinery to add text to each
2390 diagnostic emitted, which indicates which command line option directly
2391 controls that diagnostic, when such an option is known to the
2392 diagnostic machinery.
2394 @item -Wcoverage-mismatch
2395 @opindex Wcoverage-mismatch
2396 Warn if feedback profiles do not match when using the
2397 @option{-fprofile-use} option.
2398 If a source file was changed between @option{-fprofile-gen} and
2399 @option{-fprofile-use}, the files with the profile feedback can fail
2400 to match the source file and GCC can not use the profile feedback
2401 information. By default, GCC emits an error message in this case.
2402 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2403 error. GCC does not use appropriate feedback profiles, so using this
2404 option can result in poorly optimized code. This option is useful
2405 only in the case of very minor changes such as bug fixes to an
2410 @node Warning Options
2411 @section Options to Request or Suppress Warnings
2412 @cindex options to control warnings
2413 @cindex warning messages
2414 @cindex messages, warning
2415 @cindex suppressing warnings
2417 Warnings are diagnostic messages that report constructions which
2418 are not inherently erroneous but which are risky or suggest there
2419 may have been an error.
2421 You can request many specific warnings with options beginning @samp{-W},
2422 for example @option{-Wimplicit} to request warnings on implicit
2423 declarations. Each of these specific warning options also has a
2424 negative form beginning @samp{-Wno-} to turn off warnings;
2425 for example, @option{-Wno-implicit}. This manual lists only one of the
2426 two forms, whichever is not the default.
2428 The following options control the amount and kinds of warnings produced
2429 by GCC; for further, language-specific options also refer to
2430 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2434 @cindex syntax checking
2436 @opindex fsyntax-only
2437 Check the code for syntax errors, but don't do anything beyond that.
2441 Issue all the warnings demanded by strict ISO C and ISO C++;
2442 reject all programs that use forbidden extensions, and some other
2443 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2444 version of the ISO C standard specified by any @option{-std} option used.
2446 Valid ISO C and ISO C++ programs should compile properly with or without
2447 this option (though a rare few will require @option{-ansi} or a
2448 @option{-std} option specifying the required version of ISO C)@. However,
2449 without this option, certain GNU extensions and traditional C and C++
2450 features are supported as well. With this option, they are rejected.
2452 @option{-pedantic} does not cause warning messages for use of the
2453 alternate keywords whose names begin and end with @samp{__}. Pedantic
2454 warnings are also disabled in the expression that follows
2455 @code{__extension__}. However, only system header files should use
2456 these escape routes; application programs should avoid them.
2457 @xref{Alternate Keywords}.
2459 Some users try to use @option{-pedantic} to check programs for strict ISO
2460 C conformance. They soon find that it does not do quite what they want:
2461 it finds some non-ISO practices, but not all---only those for which
2462 ISO C @emph{requires} a diagnostic, and some others for which
2463 diagnostics have been added.
2465 A feature to report any failure to conform to ISO C might be useful in
2466 some instances, but would require considerable additional work and would
2467 be quite different from @option{-pedantic}. We don't have plans to
2468 support such a feature in the near future.
2470 Where the standard specified with @option{-std} represents a GNU
2471 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2472 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2473 extended dialect is based. Warnings from @option{-pedantic} are given
2474 where they are required by the base standard. (It would not make sense
2475 for such warnings to be given only for features not in the specified GNU
2476 C dialect, since by definition the GNU dialects of C include all
2477 features the compiler supports with the given option, and there would be
2478 nothing to warn about.)
2480 @item -pedantic-errors
2481 @opindex pedantic-errors
2482 Like @option{-pedantic}, except that errors are produced rather than
2487 Inhibit all warning messages.
2491 Inhibit warning messages about the use of @samp{#import}.
2493 @item -Wchar-subscripts
2494 @opindex Wchar-subscripts
2495 Warn if an array subscript has type @code{char}. This is a common cause
2496 of error, as programmers often forget that this type is signed on some
2498 This warning is enabled by @option{-Wall}.
2502 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2503 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2504 This warning is enabled by @option{-Wall}.
2506 @item -Wfatal-errors
2507 @opindex Wfatal-errors
2508 This option causes the compiler to abort compilation on the first error
2509 occurred rather than trying to keep going and printing further error
2514 @opindex ffreestanding
2515 @opindex fno-builtin
2516 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2517 the arguments supplied have types appropriate to the format string
2518 specified, and that the conversions specified in the format string make
2519 sense. This includes standard functions, and others specified by format
2520 attributes (@pxref{Function Attributes}), in the @code{printf},
2521 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2522 not in the C standard) families (or other target-specific families).
2523 Which functions are checked without format attributes having been
2524 specified depends on the standard version selected, and such checks of
2525 functions without the attribute specified are disabled by
2526 @option{-ffreestanding} or @option{-fno-builtin}.
2528 The formats are checked against the format features supported by GNU
2529 libc version 2.2. These include all ISO C90 and C99 features, as well
2530 as features from the Single Unix Specification and some BSD and GNU
2531 extensions. Other library implementations may not support all these
2532 features; GCC does not support warning about features that go beyond a
2533 particular library's limitations. However, if @option{-pedantic} is used
2534 with @option{-Wformat}, warnings will be given about format features not
2535 in the selected standard version (but not for @code{strfmon} formats,
2536 since those are not in any version of the C standard). @xref{C Dialect
2537 Options,,Options Controlling C Dialect}.
2539 Since @option{-Wformat} also checks for null format arguments for
2540 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2542 @option{-Wformat} is included in @option{-Wall}. For more control over some
2543 aspects of format checking, the options @option{-Wformat-y2k},
2544 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2545 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2546 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2549 @opindex Wformat-y2k
2550 If @option{-Wformat} is specified, also warn about @code{strftime}
2551 formats which may yield only a two-digit year.
2553 @item -Wno-format-extra-args
2554 @opindex Wno-format-extra-args
2555 If @option{-Wformat} is specified, do not warn about excess arguments to a
2556 @code{printf} or @code{scanf} format function. The C standard specifies
2557 that such arguments are ignored.
2559 Where the unused arguments lie between used arguments that are
2560 specified with @samp{$} operand number specifications, normally
2561 warnings are still given, since the implementation could not know what
2562 type to pass to @code{va_arg} to skip the unused arguments. However,
2563 in the case of @code{scanf} formats, this option will suppress the
2564 warning if the unused arguments are all pointers, since the Single
2565 Unix Specification says that such unused arguments are allowed.
2567 @item -Wno-format-zero-length
2568 @opindex Wno-format-zero-length
2569 If @option{-Wformat} is specified, do not warn about zero-length formats.
2570 The C standard specifies that zero-length formats are allowed.
2572 @item -Wformat-nonliteral
2573 @opindex Wformat-nonliteral
2574 If @option{-Wformat} is specified, also warn if the format string is not a
2575 string literal and so cannot be checked, unless the format function
2576 takes its format arguments as a @code{va_list}.
2578 @item -Wformat-security
2579 @opindex Wformat-security
2580 If @option{-Wformat} is specified, also warn about uses of format
2581 functions that represent possible security problems. At present, this
2582 warns about calls to @code{printf} and @code{scanf} functions where the
2583 format string is not a string literal and there are no format arguments,
2584 as in @code{printf (foo);}. This may be a security hole if the format
2585 string came from untrusted input and contains @samp{%n}. (This is
2586 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2587 in future warnings may be added to @option{-Wformat-security} that are not
2588 included in @option{-Wformat-nonliteral}.)
2592 Enable @option{-Wformat} plus format checks not included in
2593 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2594 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2598 Warn about passing a null pointer for arguments marked as
2599 requiring a non-null value by the @code{nonnull} function attribute.
2601 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2602 can be disabled with the @option{-Wno-nonnull} option.
2604 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2606 Warn about uninitialized variables which are initialized with themselves.
2607 Note this option can only be used with the @option{-Wuninitialized} option,
2608 which in turn only works with @option{-O1} and above.
2610 For example, GCC will warn about @code{i} being uninitialized in the
2611 following snippet only when @option{-Winit-self} has been specified:
2622 @item -Wimplicit-int
2623 @opindex Wimplicit-int
2624 Warn when a declaration does not specify a type.
2625 This warning is enabled by @option{-Wall}.
2627 @item -Wimplicit-function-declaration
2628 @opindex Wimplicit-function-declaration
2629 @opindex Wno-implicit-function-declaration
2630 Give a warning whenever a function is used before being declared. In
2631 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2632 enabled by default and it is made into an error by
2633 @option{-pedantic-errors}. This warning is also enabled by
2638 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2639 This warning is enabled by @option{-Wall}.
2643 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2644 function with external linkage, returning int, taking either zero
2645 arguments, two, or three arguments of appropriate types.
2646 This warning is enabled by @option{-Wall}.
2648 @item -Wmissing-braces
2649 @opindex Wmissing-braces
2650 Warn if an aggregate or union initializer is not fully bracketed. In
2651 the following example, the initializer for @samp{a} is not fully
2652 bracketed, but that for @samp{b} is fully bracketed.
2655 int a[2][2] = @{ 0, 1, 2, 3 @};
2656 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2659 This warning is enabled by @option{-Wall}.
2661 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2662 @opindex Wmissing-include-dirs
2663 Warn if a user-supplied include directory does not exist.
2666 @opindex Wparentheses
2667 Warn if parentheses are omitted in certain contexts, such
2668 as when there is an assignment in a context where a truth value
2669 is expected, or when operators are nested whose precedence people
2670 often get confused about.
2672 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2673 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2674 interpretation from that of ordinary mathematical notation.
2676 Also warn about constructions where there may be confusion to which
2677 @code{if} statement an @code{else} branch belongs. Here is an example of
2692 In C/C++, every @code{else} branch belongs to the innermost possible
2693 @code{if} statement, which in this example is @code{if (b)}. This is
2694 often not what the programmer expected, as illustrated in the above
2695 example by indentation the programmer chose. When there is the
2696 potential for this confusion, GCC will issue a warning when this flag
2697 is specified. To eliminate the warning, add explicit braces around
2698 the innermost @code{if} statement so there is no way the @code{else}
2699 could belong to the enclosing @code{if}. The resulting code would
2716 This warning is enabled by @option{-Wall}.
2718 @item -Wsequence-point
2719 @opindex Wsequence-point
2720 Warn about code that may have undefined semantics because of violations
2721 of sequence point rules in the C and C++ standards.
2723 The C and C++ standards defines the order in which expressions in a C/C++
2724 program are evaluated in terms of @dfn{sequence points}, which represent
2725 a partial ordering between the execution of parts of the program: those
2726 executed before the sequence point, and those executed after it. These
2727 occur after the evaluation of a full expression (one which is not part
2728 of a larger expression), after the evaluation of the first operand of a
2729 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2730 function is called (but after the evaluation of its arguments and the
2731 expression denoting the called function), and in certain other places.
2732 Other than as expressed by the sequence point rules, the order of
2733 evaluation of subexpressions of an expression is not specified. All
2734 these rules describe only a partial order rather than a total order,
2735 since, for example, if two functions are called within one expression
2736 with no sequence point between them, the order in which the functions
2737 are called is not specified. However, the standards committee have
2738 ruled that function calls do not overlap.
2740 It is not specified when between sequence points modifications to the
2741 values of objects take effect. Programs whose behavior depends on this
2742 have undefined behavior; the C and C++ standards specify that ``Between
2743 the previous and next sequence point an object shall have its stored
2744 value modified at most once by the evaluation of an expression.
2745 Furthermore, the prior value shall be read only to determine the value
2746 to be stored.''. If a program breaks these rules, the results on any
2747 particular implementation are entirely unpredictable.
2749 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2750 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2751 diagnosed by this option, and it may give an occasional false positive
2752 result, but in general it has been found fairly effective at detecting
2753 this sort of problem in programs.
2755 The standard is worded confusingly, therefore there is some debate
2756 over the precise meaning of the sequence point rules in subtle cases.
2757 Links to discussions of the problem, including proposed formal
2758 definitions, may be found on the GCC readings page, at
2759 @w{@uref{http://gcc.gnu.org/readings.html}}.
2761 This warning is enabled by @option{-Wall} for C and C++.
2764 @opindex Wreturn-type
2765 Warn whenever a function is defined with a return-type that defaults to
2766 @code{int}. Also warn about any @code{return} statement with no
2767 return-value in a function whose return-type is not @code{void}.
2769 Also warn if the return type of a function has a type qualifier
2770 such as @code{const}. For ISO C such a type qualifier has no effect,
2771 since the value returned by a function is not an lvalue.
2772 For C++, the warning is only emitted for scalar types or @code{void}.
2773 ISO C prohibits qualified @code{void} return types on function
2774 definitions, so such return types always receive a warning
2775 even without this option.
2777 For C++, a function without return type always produces a diagnostic
2778 message, even when @option{-Wno-return-type} is specified. The only
2779 exceptions are @samp{main} and functions defined in system headers.
2781 This warning is enabled by @option{-Wall}.
2785 Warn whenever a @code{switch} statement has an index of enumerated type
2786 and lacks a @code{case} for one or more of the named codes of that
2787 enumeration. (The presence of a @code{default} label prevents this
2788 warning.) @code{case} labels outside the enumeration range also
2789 provoke warnings when this option is used.
2790 This warning is enabled by @option{-Wall}.
2792 @item -Wswitch-default
2793 @opindex Wswitch-switch
2794 Warn whenever a @code{switch} statement does not have a @code{default}
2798 @opindex Wswitch-enum
2799 Warn whenever a @code{switch} statement has an index of enumerated type
2800 and lacks a @code{case} for one or more of the named codes of that
2801 enumeration. @code{case} labels outside the enumeration range also
2802 provoke warnings when this option is used.
2806 Warn if any trigraphs are encountered that might change the meaning of
2807 the program (trigraphs within comments are not warned about).
2808 This warning is enabled by @option{-Wall}.
2810 @item -Wunused-function
2811 @opindex Wunused-function
2812 Warn whenever a static function is declared but not defined or a
2813 non-inline static function is unused.
2814 This warning is enabled by @option{-Wall}.
2816 @item -Wunused-label
2817 @opindex Wunused-label
2818 Warn whenever a label is declared but not used.
2819 This warning is enabled by @option{-Wall}.
2821 To suppress this warning use the @samp{unused} attribute
2822 (@pxref{Variable Attributes}).
2824 @item -Wunused-parameter
2825 @opindex Wunused-parameter
2826 Warn whenever a function parameter is unused aside from its declaration.
2828 To suppress this warning use the @samp{unused} attribute
2829 (@pxref{Variable Attributes}).
2831 @item -Wunused-variable
2832 @opindex Wunused-variable
2833 Warn whenever a local variable or non-constant static variable is unused
2834 aside from its declaration.
2835 This warning is enabled by @option{-Wall}.
2837 To suppress this warning use the @samp{unused} attribute
2838 (@pxref{Variable Attributes}).
2840 @item -Wunused-value
2841 @opindex Wunused-value
2842 Warn whenever a statement computes a result that is explicitly not
2843 used. To suppress this warning cast the unused expression to
2844 @samp{void}. This includes an expression-statement or the left-hand
2845 side of a comma expression that contains no side effects. For example,
2846 an expression such as @samp{x[i,j]} will cause a warning, while
2847 @samp{x[(void)i,j]} will not.
2849 This warning is enabled by @option{-Wall}.
2853 All the above @option{-Wunused} options combined.
2855 In order to get a warning about an unused function parameter, you must
2856 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2857 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2859 @item -Wuninitialized
2860 @opindex Wuninitialized
2861 Warn if an automatic variable is used without first being initialized or
2862 if a variable may be clobbered by a @code{setjmp} call.
2864 These warnings are possible only in optimizing compilation,
2865 because they require data flow information that is computed only
2866 when optimizing. If you do not specify @option{-O}, you will not get
2867 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2868 requiring @option{-O}.
2870 If you want to warn about code which uses the uninitialized value of the
2871 variable in its own initializer, use the @option{-Winit-self} option.
2873 These warnings occur for individual uninitialized or clobbered
2874 elements of structure, union or array variables as well as for
2875 variables which are uninitialized or clobbered as a whole. They do
2876 not occur for variables or elements declared @code{volatile}. Because
2877 these warnings depend on optimization, the exact variables or elements
2878 for which there are warnings will depend on the precise optimization
2879 options and version of GCC used.
2881 Note that there may be no warning about a variable that is used only
2882 to compute a value that itself is never used, because such
2883 computations may be deleted by data flow analysis before the warnings
2886 These warnings are made optional because GCC is not smart
2887 enough to see all the reasons why the code might be correct
2888 despite appearing to have an error. Here is one example of how
2909 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2910 always initialized, but GCC doesn't know this. Here is
2911 another common case:
2916 if (change_y) save_y = y, y = new_y;
2918 if (change_y) y = save_y;
2923 This has no bug because @code{save_y} is used only if it is set.
2925 @cindex @code{longjmp} warnings
2926 This option also warns when a non-volatile automatic variable might be
2927 changed by a call to @code{longjmp}. These warnings as well are possible
2928 only in optimizing compilation.
2930 The compiler sees only the calls to @code{setjmp}. It cannot know
2931 where @code{longjmp} will be called; in fact, a signal handler could
2932 call it at any point in the code. As a result, you may get a warning
2933 even when there is in fact no problem because @code{longjmp} cannot
2934 in fact be called at the place which would cause a problem.
2936 Some spurious warnings can be avoided if you declare all the functions
2937 you use that never return as @code{noreturn}. @xref{Function
2940 This warning is enabled by @option{-Wall}.
2942 @item -Wunknown-pragmas
2943 @opindex Wunknown-pragmas
2944 @cindex warning for unknown pragmas
2945 @cindex unknown pragmas, warning
2946 @cindex pragmas, warning of unknown
2947 Warn when a #pragma directive is encountered which is not understood by
2948 GCC@. If this command line option is used, warnings will even be issued
2949 for unknown pragmas in system header files. This is not the case if
2950 the warnings were only enabled by the @option{-Wall} command line option.
2953 @opindex Wno-pragmas
2955 Do not warn about misuses of pragmas, such as incorrect parameters,
2956 invalid syntax, or conflicts between pragmas. See also
2957 @samp{-Wunknown-pragmas}.
2959 @item -Wstrict-aliasing
2960 @opindex Wstrict-aliasing
2961 This option is only active when @option{-fstrict-aliasing} is active.
2962 It warns about code which might break the strict aliasing rules that the
2963 compiler is using for optimization. The warning does not catch all
2964 cases, but does attempt to catch the more common pitfalls. It is
2965 included in @option{-Wall}.
2967 @item -Wstrict-aliasing=2
2968 @opindex Wstrict-aliasing=2
2969 This option is only active when @option{-fstrict-aliasing} is active.
2970 It warns about code which might break the strict aliasing rules that the
2971 compiler is using for optimization. This warning catches more cases than
2972 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2973 cases that are safe.
2975 @item -Wstrict-overflow
2976 @item -Wstrict-overflow=@var{n}
2977 @opindex -Wstrict-overflow
2978 This option is only active when @option{-fstrict-overflow} is active.
2979 It warns about cases where the compiler optimizes based on the
2980 assumption that signed overflow does not occur. Note that it does not
2981 warn about all cases where the code might overflow: it only warns
2982 about cases where the compiler implements some optimization. Thus
2983 this warning depends on the optimization level.
2985 An optimization which assumes that signed overflow does not occur is
2986 perfectly safe if the values of the variables involved are such that
2987 overflow never does, in fact, occur. Therefore this warning can
2988 easily give a false positive: a warning about code which is not
2989 actually a problem. To help focus on important issues, several
2990 warning levels are defined.
2993 @item -Wstrict-overflow=1
2994 Warn about cases which are both questionable and easy to avoid. For
2995 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
2996 compiler will simplify this to @code{1}. @option{-Wstrict-overflow}
2997 (with no level) is the same as @option{-Wstrict-overflow=1}. This
2998 level of @option{-Wstrict-overflow} is enabled by @option{-Wall};
2999 higher levels are not, and must be explicitly requested.
3001 @item -Wstrict-overflow=2
3002 Also warn about other cases where a comparison is simplified to a
3003 constant. For example: @code{abs (x) >= 0}. This can only be
3004 simplified when @option{-fstrict-overflow} is in effect, because
3005 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3008 @item -Wstrict-overflow=3
3009 Also warn about other cases where a comparison is simplified. For
3010 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3012 @item -Wstrict-overflow=4
3013 Also warn about other simplifications not covered by the above cases.
3014 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3016 @item -Wstrict-overflow=5
3017 Also warn about cases where the compiler reduces the magnitude of a
3018 constant involved in a comparison. For example: @code{x + 2 > y} will
3019 be simplified to @code{x + 1 >= y}. This is reported only at the
3020 highest warning level because this simplification applies to many
3021 comparisons, so this warning level will give a very large number of
3025 @item -Warray-bounds
3026 @opindex Wno-array-bounds
3027 @opindex Warray-bounds
3028 This option is only active when @option{-ftree-vrp} is active
3029 (default for -O2 and above). It warns about subscripts to arrays
3030 that are always out of bounds. This warning is enabled by @option{-Wall}.
3034 All of the above @samp{-W} options combined. This enables all the
3035 warnings about constructions that some users consider questionable, and
3036 that are easy to avoid (or modify to prevent the warning), even in
3037 conjunction with macros. This also enables some language-specific
3038 warnings described in @ref{C++ Dialect Options} and
3039 @ref{Objective-C and Objective-C++ Dialect Options}.
3042 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
3043 Some of them warn about constructions that users generally do not
3044 consider questionable, but which occasionally you might wish to check
3045 for; others warn about constructions that are necessary or hard to avoid
3046 in some cases, and there is no simple way to modify the code to suppress
3053 (This option used to be called @option{-W}. The older name is still
3054 supported, but the newer name is more descriptive.) Print extra warning
3055 messages for these events:
3059 A function can return either with or without a value. (Falling
3060 off the end of the function body is considered returning without
3061 a value.) For example, this function would evoke such a
3075 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
3078 Storage-class specifiers like @code{static} are not the first things
3079 in a declaration. According to the C Standard, this usage is
3080 obsolescent. This warning can be independently controlled by
3081 @option{-Wold-style-declaration}.
3084 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
3088 A comparison between signed and unsigned values could produce an
3089 incorrect result when the signed value is converted to unsigned.
3090 (But don't warn if @option{-Wno-sign-compare} is also specified.)
3093 An aggregate has an initializer which does not initialize all members.
3094 This warning can be independently controlled by
3095 @option{-Wmissing-field-initializers}.
3098 An initialized field without side effects is overridden when using
3099 designated initializers (@pxref{Designated Inits, , Designated
3100 Initializers}). This warning can be independently controlled by
3101 @option{-Woverride-init}.
3104 A function parameter is declared without a type specifier in K&R-style
3105 functions. This warning can be independently controlled by
3106 @option{-Wmissing-parameter-type}.
3109 An empty body occurs in an @samp{if} or @samp{else} statement. This
3110 warning can be independently controlled by @option{-Wempty-body}.
3113 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3114 @samp{>}, or @samp{>=}.
3117 A variable might be changed by @samp{longjmp} or @samp{vfork}.
3118 This warning can be independently controlled by @option{-Wclobbered}.
3120 @item @r{(C++ only)}
3121 An enumerator and a non-enumerator both appear in a conditional expression.
3123 @item @r{(C++ only)}
3124 A non-static reference or non-static @samp{const} member appears in a
3125 class without constructors.
3127 @item @r{(C++ only)}
3128 Ambiguous virtual bases.
3130 @item @r{(C++ only)}
3131 Subscripting an array which has been declared @samp{register}.
3133 @item @r{(C++ only)}
3134 Taking the address of a variable which has been declared @samp{register}.
3136 @item @r{(C++ only)}
3137 A base class is not initialized in a derived class' copy constructor.
3140 @item -Wno-div-by-zero
3141 @opindex Wno-div-by-zero
3142 @opindex Wdiv-by-zero
3143 Do not warn about compile-time integer division by zero. Floating point
3144 division by zero is not warned about, as it can be a legitimate way of
3145 obtaining infinities and NaNs.
3147 @item -Wsystem-headers
3148 @opindex Wsystem-headers
3149 @cindex warnings from system headers
3150 @cindex system headers, warnings from
3151 Print warning messages for constructs found in system header files.
3152 Warnings from system headers are normally suppressed, on the assumption
3153 that they usually do not indicate real problems and would only make the
3154 compiler output harder to read. Using this command line option tells
3155 GCC to emit warnings from system headers as if they occurred in user
3156 code. However, note that using @option{-Wall} in conjunction with this
3157 option will @emph{not} warn about unknown pragmas in system
3158 headers---for that, @option{-Wunknown-pragmas} must also be used.
3161 @opindex Wfloat-equal
3162 Warn if floating point values are used in equality comparisons.
3164 The idea behind this is that sometimes it is convenient (for the
3165 programmer) to consider floating-point values as approximations to
3166 infinitely precise real numbers. If you are doing this, then you need
3167 to compute (by analyzing the code, or in some other way) the maximum or
3168 likely maximum error that the computation introduces, and allow for it
3169 when performing comparisons (and when producing output, but that's a
3170 different problem). In particular, instead of testing for equality, you
3171 would check to see whether the two values have ranges that overlap; and
3172 this is done with the relational operators, so equality comparisons are
3175 @item -Wtraditional @r{(C only)}
3176 @opindex Wtraditional
3177 Warn about certain constructs that behave differently in traditional and
3178 ISO C@. Also warn about ISO C constructs that have no traditional C
3179 equivalent, and/or problematic constructs which should be avoided.
3183 Macro parameters that appear within string literals in the macro body.
3184 In traditional C macro replacement takes place within string literals,
3185 but does not in ISO C@.
3188 In traditional C, some preprocessor directives did not exist.
3189 Traditional preprocessors would only consider a line to be a directive
3190 if the @samp{#} appeared in column 1 on the line. Therefore
3191 @option{-Wtraditional} warns about directives that traditional C
3192 understands but would ignore because the @samp{#} does not appear as the
3193 first character on the line. It also suggests you hide directives like
3194 @samp{#pragma} not understood by traditional C by indenting them. Some
3195 traditional implementations would not recognize @samp{#elif}, so it
3196 suggests avoiding it altogether.
3199 A function-like macro that appears without arguments.
3202 The unary plus operator.
3205 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3206 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3207 constants.) Note, these suffixes appear in macros defined in the system
3208 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3209 Use of these macros in user code might normally lead to spurious
3210 warnings, however GCC's integrated preprocessor has enough context to
3211 avoid warning in these cases.
3214 A function declared external in one block and then used after the end of
3218 A @code{switch} statement has an operand of type @code{long}.
3221 A non-@code{static} function declaration follows a @code{static} one.
3222 This construct is not accepted by some traditional C compilers.
3225 The ISO type of an integer constant has a different width or
3226 signedness from its traditional type. This warning is only issued if
3227 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3228 typically represent bit patterns, are not warned about.
3231 Usage of ISO string concatenation is detected.
3234 Initialization of automatic aggregates.
3237 Identifier conflicts with labels. Traditional C lacks a separate
3238 namespace for labels.
3241 Initialization of unions. If the initializer is zero, the warning is
3242 omitted. This is done under the assumption that the zero initializer in
3243 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3244 initializer warnings and relies on default initialization to zero in the
3248 Conversions by prototypes between fixed/floating point values and vice
3249 versa. The absence of these prototypes when compiling with traditional
3250 C would cause serious problems. This is a subset of the possible
3251 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3254 Use of ISO C style function definitions. This warning intentionally is
3255 @emph{not} issued for prototype declarations or variadic functions
3256 because these ISO C features will appear in your code when using
3257 libiberty's traditional C compatibility macros, @code{PARAMS} and
3258 @code{VPARAMS}. This warning is also bypassed for nested functions
3259 because that feature is already a GCC extension and thus not relevant to
3260 traditional C compatibility.
3263 @item -Wtraditional-conversion @r{(C only)}
3264 @opindex Wtraditional-conversion
3265 Warn if a prototype causes a type conversion that is different from what
3266 would happen to the same argument in the absence of a prototype. This
3267 includes conversions of fixed point to floating and vice versa, and
3268 conversions changing the width or signedness of a fixed point argument
3269 except when the same as the default promotion.
3271 @item -Wdeclaration-after-statement @r{(C only)}
3272 @opindex Wdeclaration-after-statement
3273 Warn when a declaration is found after a statement in a block. This
3274 construct, known from C++, was introduced with ISO C99 and is by default
3275 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3276 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3280 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3282 @item -Wno-endif-labels
3283 @opindex Wno-endif-labels
3284 @opindex Wendif-labels
3285 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3289 Warn whenever a local variable shadows another local variable, parameter or
3290 global variable or whenever a built-in function is shadowed.
3292 @item -Wlarger-than-@var{len}
3293 @opindex Wlarger-than
3294 Warn whenever an object of larger than @var{len} bytes is defined.
3296 @item -Wunsafe-loop-optimizations
3297 @opindex Wunsafe-loop-optimizations
3298 Warn if the loop cannot be optimized because the compiler could not
3299 assume anything on the bounds of the loop indices. With
3300 @option{-funsafe-loop-optimizations} warn if the compiler made
3303 @item -Wpointer-arith
3304 @opindex Wpointer-arith
3305 Warn about anything that depends on the ``size of'' a function type or
3306 of @code{void}. GNU C assigns these types a size of 1, for
3307 convenience in calculations with @code{void *} pointers and pointers
3308 to functions. In C++, warn also when an arithmetic operation involves
3309 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3311 @item -Wbad-function-cast @r{(C only)}
3312 @opindex Wbad-function-cast
3313 Warn whenever a function call is cast to a non-matching type.
3314 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3317 Warn about ISO C constructs that are outside of the common subset of
3318 ISO C and ISO C++, e.g.@: request for implicit conversion from
3319 @code{void *} to a pointer to non-@code{void} type.
3321 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3322 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3323 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3324 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3328 Warn whenever a pointer is cast so as to remove a type qualifier from
3329 the target type. For example, warn if a @code{const char *} is cast
3330 to an ordinary @code{char *}.
3333 @opindex Wcast-align
3334 Warn whenever a pointer is cast such that the required alignment of the
3335 target is increased. For example, warn if a @code{char *} is cast to
3336 an @code{int *} on machines where integers can only be accessed at
3337 two- or four-byte boundaries.
3339 @item -Wwrite-strings
3340 @opindex Wwrite-strings
3341 When compiling C, give string constants the type @code{const
3342 char[@var{length}]} so that
3343 copying the address of one into a non-@code{const} @code{char *}
3344 pointer will get a warning; when compiling C++, warn about the
3345 deprecated conversion from string literals to @code{char *}. This
3346 warning, by default, is enabled for C++ programs.
3347 These warnings will help you find at
3348 compile time code that can try to write into a string constant, but
3349 only if you have been very careful about using @code{const} in
3350 declarations and prototypes. Otherwise, it will just be a nuisance;
3351 this is why we did not make @option{-Wall} request these warnings.
3355 Warn for variables that might be changed by @samp{longjmp} or
3356 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3359 @opindex Wconversion
3360 Warn for implicit conversions that may alter a value. This includes
3361 conversions between real and integer, like @code{abs (x)} when
3362 @code{x} is @code{double}; conversions between signed and unsigned,
3363 like @code{unsigned ui = -1}; and conversions to smaller types, like
3364 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3365 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3366 changed by the conversion like in @code{abs (2.0)}.
3369 @opindex Wempty-body
3370 An empty body occurs in an @samp{if} or @samp{else} statement.
3371 This warning is also enabled by @option{-Wextra}.
3373 @item -Wsign-compare
3374 @opindex Wsign-compare
3375 @cindex warning for comparison of signed and unsigned values
3376 @cindex comparison of signed and unsigned values, warning
3377 @cindex signed and unsigned values, comparison warning
3378 Warn when a comparison between signed and unsigned values could produce
3379 an incorrect result when the signed value is converted to unsigned.
3380 This warning is also enabled by @option{-Wextra}; to get the other warnings
3381 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3385 @opindex Wno-address
3386 Warn about suspicious uses of memory addresses. These include using
3387 the address of a function in a conditional expression, such as
3388 @code{void func(void); if (func)}, and comparisons against the memory
3389 address of a string literal, such as @code{if (x == "abc")}. Such
3390 uses typically indicate a programmer error: the address of a function
3391 always evaluates to true, so their use in a conditional usually
3392 indicate that the programmer forgot the parentheses in a function
3393 call; and comparisons against string literals result in unspecified
3394 behavior and are not portable in C, so they usually indicate that the
3395 programmer intended to use @code{strcmp}. This warning is enabled by
3398 @item -Waggregate-return
3399 @opindex Waggregate-return
3400 Warn if any functions that return structures or unions are defined or
3401 called. (In languages where you can return an array, this also elicits
3404 @item -Wno-attributes
3405 @opindex Wno-attributes
3406 @opindex Wattributes
3407 Do not warn if an unexpected @code{__attribute__} is used, such as
3408 unrecognized attributes, function attributes applied to variables,
3409 etc. This will not stop errors for incorrect use of supported
3412 @item -Wstrict-prototypes @r{(C only)}
3413 @opindex Wstrict-prototypes
3414 Warn if a function is declared or defined without specifying the
3415 argument types. (An old-style function definition is permitted without
3416 a warning if preceded by a declaration which specifies the argument
3419 @item -Wold-style-declaration @r{(C only)}
3420 @opindex Wold-style-declaration
3421 Warn for obsolescent usages, according to the C Standard, in a
3422 declaration. For example, warn if storage-class specifiers like
3423 @code{static} are not the first things in a declaration. This warning
3424 is also enabled by @option{-Wextra}.
3426 @item -Wold-style-definition @r{(C only)}
3427 @opindex Wold-style-definition
3428 Warn if an old-style function definition is used. A warning is given
3429 even if there is a previous prototype.
3431 @item -Wmissing-parameter-type @r{(C only)}
3432 @opindex Wmissing-parameter-type
3433 A function parameter is declared without a type specifier in K&R-style
3440 This warning is also enabled by @option{-Wextra}.
3442 @item -Wmissing-prototypes @r{(C only)}
3443 @opindex Wmissing-prototypes
3444 Warn if a global function is defined without a previous prototype
3445 declaration. This warning is issued even if the definition itself
3446 provides a prototype. The aim is to detect global functions that fail
3447 to be declared in header files.
3449 @item -Wmissing-declarations @r{(C and C++ only)}
3450 @opindex Wmissing-declarations
3451 Warn if a global function is defined without a previous declaration.
3452 Do so even if the definition itself provides a prototype.
3453 Use this option to detect global functions that are not declared in
3454 header files. In C++, no warnings are issued for function templates,
3455 or for inline functions, or for functions in anonymous namespaces.
3457 @item -Wmissing-field-initializers
3458 @opindex Wmissing-field-initializers
3461 Warn if a structure's initializer has some fields missing. For
3462 example, the following code would cause such a warning, because
3463 @code{x.h} is implicitly zero:
3466 struct s @{ int f, g, h; @};
3467 struct s x = @{ 3, 4 @};
3470 This option does not warn about designated initializers, so the following
3471 modification would not trigger a warning:
3474 struct s @{ int f, g, h; @};
3475 struct s x = @{ .f = 3, .g = 4 @};
3478 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3479 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3481 @item -Wmissing-noreturn
3482 @opindex Wmissing-noreturn
3483 Warn about functions which might be candidates for attribute @code{noreturn}.
3484 Note these are only possible candidates, not absolute ones. Care should
3485 be taken to manually verify functions actually do not ever return before
3486 adding the @code{noreturn} attribute, otherwise subtle code generation
3487 bugs could be introduced. You will not get a warning for @code{main} in
3488 hosted C environments.
3490 @item -Wmissing-format-attribute
3491 @opindex Wmissing-format-attribute
3493 Warn about function pointers which might be candidates for @code{format}
3494 attributes. Note these are only possible candidates, not absolute ones.
3495 GCC will guess that function pointers with @code{format} attributes that
3496 are used in assignment, initialization, parameter passing or return
3497 statements should have a corresponding @code{format} attribute in the
3498 resulting type. I.e.@: the left-hand side of the assignment or
3499 initialization, the type of the parameter variable, or the return type
3500 of the containing function respectively should also have a @code{format}
3501 attribute to avoid the warning.
3503 GCC will also warn about function definitions which might be
3504 candidates for @code{format} attributes. Again, these are only
3505 possible candidates. GCC will guess that @code{format} attributes
3506 might be appropriate for any function that calls a function like
3507 @code{vprintf} or @code{vscanf}, but this might not always be the
3508 case, and some functions for which @code{format} attributes are
3509 appropriate may not be detected.
3511 @item -Wno-multichar
3512 @opindex Wno-multichar
3514 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3515 Usually they indicate a typo in the user's code, as they have
3516 implementation-defined values, and should not be used in portable code.
3518 @item -Wnormalized=<none|id|nfc|nfkc>
3519 @opindex Wnormalized
3522 @cindex character set, input normalization
3523 In ISO C and ISO C++, two identifiers are different if they are
3524 different sequences of characters. However, sometimes when characters
3525 outside the basic ASCII character set are used, you can have two
3526 different character sequences that look the same. To avoid confusion,
3527 the ISO 10646 standard sets out some @dfn{normalization rules} which
3528 when applied ensure that two sequences that look the same are turned into
3529 the same sequence. GCC can warn you if you are using identifiers which
3530 have not been normalized; this option controls that warning.
3532 There are four levels of warning that GCC supports. The default is
3533 @option{-Wnormalized=nfc}, which warns about any identifier which is
3534 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3535 recommended form for most uses.
3537 Unfortunately, there are some characters which ISO C and ISO C++ allow
3538 in identifiers that when turned into NFC aren't allowable as
3539 identifiers. That is, there's no way to use these symbols in portable
3540 ISO C or C++ and have all your identifiers in NFC.
3541 @option{-Wnormalized=id} suppresses the warning for these characters.
3542 It is hoped that future versions of the standards involved will correct
3543 this, which is why this option is not the default.
3545 You can switch the warning off for all characters by writing
3546 @option{-Wnormalized=none}. You would only want to do this if you
3547 were using some other normalization scheme (like ``D''), because
3548 otherwise you can easily create bugs that are literally impossible to see.
3550 Some characters in ISO 10646 have distinct meanings but look identical
3551 in some fonts or display methodologies, especially once formatting has
3552 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3553 LETTER N'', will display just like a regular @code{n} which has been
3554 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3555 normalization scheme to convert all these into a standard form as
3556 well, and GCC will warn if your code is not in NFKC if you use
3557 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3558 about every identifier that contains the letter O because it might be
3559 confused with the digit 0, and so is not the default, but may be
3560 useful as a local coding convention if the programming environment is
3561 unable to be fixed to display these characters distinctly.
3563 @item -Wno-deprecated-declarations
3564 @opindex Wno-deprecated-declarations
3565 Do not warn about uses of functions (@pxref{Function Attributes}),
3566 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3567 Attributes}) marked as deprecated by using the @code{deprecated}
3571 @opindex Wno-overflow
3572 Do not warn about compile-time overflow in constant expressions.
3574 @item -Woverride-init
3575 @opindex Woverride-init
3578 Warn if an initialized field without side effects is overridden when
3579 using designated initializers (@pxref{Designated Inits, , Designated
3582 This warning is included in @option{-Wextra}. To get other
3583 @option{-Wextra} warnings without this one, use @samp{-Wextra
3584 -Wno-override-init}.
3588 Warn if a structure is given the packed attribute, but the packed
3589 attribute has no effect on the layout or size of the structure.
3590 Such structures may be mis-aligned for little benefit. For
3591 instance, in this code, the variable @code{f.x} in @code{struct bar}
3592 will be misaligned even though @code{struct bar} does not itself
3593 have the packed attribute:
3600 @} __attribute__((packed));
3610 Warn if padding is included in a structure, either to align an element
3611 of the structure or to align the whole structure. Sometimes when this
3612 happens it is possible to rearrange the fields of the structure to
3613 reduce the padding and so make the structure smaller.
3615 @item -Wredundant-decls
3616 @opindex Wredundant-decls
3617 Warn if anything is declared more than once in the same scope, even in
3618 cases where multiple declaration is valid and changes nothing.
3620 @item -Wnested-externs @r{(C only)}
3621 @opindex Wnested-externs
3622 Warn if an @code{extern} declaration is encountered within a function.
3624 @item -Wunreachable-code
3625 @opindex Wunreachable-code
3626 Warn if the compiler detects that code will never be executed.
3628 This option is intended to warn when the compiler detects that at
3629 least a whole line of source code will never be executed, because
3630 some condition is never satisfied or because it is after a
3631 procedure that never returns.
3633 It is possible for this option to produce a warning even though there
3634 are circumstances under which part of the affected line can be executed,
3635 so care should be taken when removing apparently-unreachable code.
3637 For instance, when a function is inlined, a warning may mean that the
3638 line is unreachable in only one inlined copy of the function.
3640 This option is not made part of @option{-Wall} because in a debugging
3641 version of a program there is often substantial code which checks
3642 correct functioning of the program and is, hopefully, unreachable
3643 because the program does work. Another common use of unreachable
3644 code is to provide behavior which is selectable at compile-time.
3648 Warn if a function can not be inlined and it was declared as inline.
3649 Even with this option, the compiler will not warn about failures to
3650 inline functions declared in system headers.
3652 The compiler uses a variety of heuristics to determine whether or not
3653 to inline a function. For example, the compiler takes into account
3654 the size of the function being inlined and the amount of inlining
3655 that has already been done in the current function. Therefore,
3656 seemingly insignificant changes in the source program can cause the
3657 warnings produced by @option{-Winline} to appear or disappear.
3659 @item -Wno-invalid-offsetof @r{(C++ only)}
3660 @opindex Wno-invalid-offsetof
3661 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3662 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3663 to a non-POD type is undefined. In existing C++ implementations,
3664 however, @samp{offsetof} typically gives meaningful results even when
3665 applied to certain kinds of non-POD types. (Such as a simple
3666 @samp{struct} that fails to be a POD type only by virtue of having a
3667 constructor.) This flag is for users who are aware that they are
3668 writing nonportable code and who have deliberately chosen to ignore the
3671 The restrictions on @samp{offsetof} may be relaxed in a future version
3672 of the C++ standard.
3674 @item -Wno-int-to-pointer-cast @r{(C only)}
3675 @opindex Wno-int-to-pointer-cast
3676 Suppress warnings from casts to pointer type of an integer of a
3679 @item -Wno-pointer-to-int-cast @r{(C only)}
3680 @opindex Wno-pointer-to-int-cast
3681 Suppress warnings from casts from a pointer to an integer type of a
3685 @opindex Winvalid-pch
3686 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3687 the search path but can't be used.
3691 @opindex Wno-long-long
3692 Warn if @samp{long long} type is used. This is default. To inhibit
3693 the warning messages, use @option{-Wno-long-long}. Flags
3694 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3695 only when @option{-pedantic} flag is used.
3697 @item -Wvariadic-macros
3698 @opindex Wvariadic-macros
3699 @opindex Wno-variadic-macros
3700 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3701 alternate syntax when in pedantic ISO C99 mode. This is default.
3702 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3704 @item -Wvolatile-register-var
3705 @opindex Wvolatile-register-var
3706 @opindex Wno-volatile-register-var
3707 Warn if a register variable is declared volatile. The volatile
3708 modifier does not inhibit all optimizations that may eliminate reads
3709 and/or writes to register variables.
3711 @item -Wdisabled-optimization
3712 @opindex Wdisabled-optimization
3713 Warn if a requested optimization pass is disabled. This warning does
3714 not generally indicate that there is anything wrong with your code; it
3715 merely indicates that GCC's optimizers were unable to handle the code
3716 effectively. Often, the problem is that your code is too big or too
3717 complex; GCC will refuse to optimize programs when the optimization
3718 itself is likely to take inordinate amounts of time.
3720 @item -Wpointer-sign
3721 @opindex Wpointer-sign
3722 @opindex Wno-pointer-sign
3723 Warn for pointer argument passing or assignment with different signedness.
3724 This option is only supported for C and Objective-C@. It is implied by
3725 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3726 @option{-Wno-pointer-sign}.
3730 Make all warnings into errors.
3734 Make the specified warning into an errors. The specifier for a
3735 warning is appended, for example @option{-Werror=switch} turns the
3736 warnings controlled by @option{-Wswitch} into errors. This switch
3737 takes a negative form, to be used to negate @option{-Werror} for
3738 specific warnings, for example @option{-Wno-error=switch} makes
3739 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3740 is in effect. You can use the @option{-fdiagnostics-show-option}
3741 option to have each controllable warning amended with the option which
3742 controls it, to determine what to use with this option.
3744 Note that specifying @option{-Werror=}@var{foo} automatically implies
3745 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3748 @item -Wstack-protector
3749 @opindex Wstack-protector
3750 This option is only active when @option{-fstack-protector} is active. It
3751 warns about functions that will not be protected against stack smashing.
3753 @item -Woverlength-strings
3754 @opindex Woverlength-strings
3755 Warn about string constants which are longer than the ``minimum
3756 maximum'' length specified in the C standard. Modern compilers
3757 generally allow string constants which are much longer than the
3758 standard's minimum limit, but very portable programs should avoid
3759 using longer strings.
3761 The limit applies @emph{after} string constant concatenation, and does
3762 not count the trailing NUL@. In C89, the limit was 509 characters; in
3763 C99, it was raised to 4095. C++98 does not specify a normative
3764 minimum maximum, so we do not diagnose overlength strings in C++@.
3766 This option is implied by @option{-pedantic}, and can be disabled with
3767 @option{-Wno-overlength-strings}.
3770 @node Debugging Options
3771 @section Options for Debugging Your Program or GCC
3772 @cindex options, debugging
3773 @cindex debugging information options
3775 GCC has various special options that are used for debugging
3776 either your program or GCC:
3781 Produce debugging information in the operating system's native format
3782 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3785 On most systems that use stabs format, @option{-g} enables use of extra
3786 debugging information that only GDB can use; this extra information
3787 makes debugging work better in GDB but will probably make other debuggers
3789 refuse to read the program. If you want to control for certain whether
3790 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3791 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3793 GCC allows you to use @option{-g} with
3794 @option{-O}. The shortcuts taken by optimized code may occasionally
3795 produce surprising results: some variables you declared may not exist
3796 at all; flow of control may briefly move where you did not expect it;
3797 some statements may not be executed because they compute constant
3798 results or their values were already at hand; some statements may
3799 execute in different places because they were moved out of loops.
3801 Nevertheless it proves possible to debug optimized output. This makes
3802 it reasonable to use the optimizer for programs that might have bugs.
3804 The following options are useful when GCC is generated with the
3805 capability for more than one debugging format.
3809 Produce debugging information for use by GDB@. This means to use the
3810 most expressive format available (DWARF 2, stabs, or the native format
3811 if neither of those are supported), including GDB extensions if at all
3816 Produce debugging information in stabs format (if that is supported),
3817 without GDB extensions. This is the format used by DBX on most BSD
3818 systems. On MIPS, Alpha and System V Release 4 systems this option
3819 produces stabs debugging output which is not understood by DBX or SDB@.
3820 On System V Release 4 systems this option requires the GNU assembler.
3822 @item -feliminate-unused-debug-symbols
3823 @opindex feliminate-unused-debug-symbols
3824 Produce debugging information in stabs format (if that is supported),
3825 for only symbols that are actually used.
3827 @item -femit-class-debug-always
3828 Instead of emitting debugging information for a C++ class in only one
3829 object file, emit it in all object files using the class. This option
3830 should be used only with debuggers that are unable to handle the way GCC
3831 normally emits debugging information for classes because using this
3832 option will increase the size of debugging information by as much as a
3837 Produce debugging information in stabs format (if that is supported),
3838 using GNU extensions understood only by the GNU debugger (GDB)@. The
3839 use of these extensions is likely to make other debuggers crash or
3840 refuse to read the program.
3844 Produce debugging information in COFF format (if that is supported).
3845 This is the format used by SDB on most System V systems prior to
3850 Produce debugging information in XCOFF format (if that is supported).
3851 This is the format used by the DBX debugger on IBM RS/6000 systems.
3855 Produce debugging information in XCOFF format (if that is supported),
3856 using GNU extensions understood only by the GNU debugger (GDB)@. The
3857 use of these extensions is likely to make other debuggers crash or
3858 refuse to read the program, and may cause assemblers other than the GNU
3859 assembler (GAS) to fail with an error.
3863 Produce debugging information in DWARF version 2 format (if that is
3864 supported). This is the format used by DBX on IRIX 6. With this
3865 option, GCC uses features of DWARF version 3 when they are useful;
3866 version 3 is upward compatible with version 2, but may still cause
3867 problems for older debuggers.
3871 Produce debugging information in VMS debug format (if that is
3872 supported). This is the format used by DEBUG on VMS systems.
3875 @itemx -ggdb@var{level}
3876 @itemx -gstabs@var{level}
3877 @itemx -gcoff@var{level}
3878 @itemx -gxcoff@var{level}
3879 @itemx -gvms@var{level}
3880 Request debugging information and also use @var{level} to specify how
3881 much information. The default level is 2.
3883 Level 0 produces no debug information at all. Thus, @option{-g0} negates
3886 Level 1 produces minimal information, enough for making backtraces in
3887 parts of the program that you don't plan to debug. This includes
3888 descriptions of functions and external variables, but no information
3889 about local variables and no line numbers.
3891 Level 3 includes extra information, such as all the macro definitions
3892 present in the program. Some debuggers support macro expansion when
3893 you use @option{-g3}.
3895 @option{-gdwarf-2} does not accept a concatenated debug level, because
3896 GCC used to support an option @option{-gdwarf} that meant to generate
3897 debug information in version 1 of the DWARF format (which is very
3898 different from version 2), and it would have been too confusing. That
3899 debug format is long obsolete, but the option cannot be changed now.
3900 Instead use an additional @option{-g@var{level}} option to change the
3901 debug level for DWARF2.
3903 @item -feliminate-dwarf2-dups
3904 @opindex feliminate-dwarf2-dups
3905 Compress DWARF2 debugging information by eliminating duplicated
3906 information about each symbol. This option only makes sense when
3907 generating DWARF2 debugging information with @option{-gdwarf-2}.
3909 @cindex @command{prof}
3912 Generate extra code to write profile information suitable for the
3913 analysis program @command{prof}. You must use this option when compiling
3914 the source files you want data about, and you must also use it when
3917 @cindex @command{gprof}
3920 Generate extra code to write profile information suitable for the
3921 analysis program @command{gprof}. You must use this option when compiling
3922 the source files you want data about, and you must also use it when
3927 Makes the compiler print out each function name as it is compiled, and
3928 print some statistics about each pass when it finishes.
3931 @opindex ftime-report
3932 Makes the compiler print some statistics about the time consumed by each
3933 pass when it finishes.
3936 @opindex fmem-report
3937 Makes the compiler print some statistics about permanent memory
3938 allocation when it finishes.
3940 @item -fpre-ipa-mem-report
3941 @opindex fpre-ipa-mem-report
3942 @item -fpost-ipa-mem-report
3943 @opindex fpost-ipa-mem-report
3944 Makes the compiler print some statistics about permanent memory
3945 allocation before or after interprocedural optimization.
3947 @item -fprofile-arcs
3948 @opindex fprofile-arcs
3949 Add code so that program flow @dfn{arcs} are instrumented. During
3950 execution the program records how many times each branch and call is
3951 executed and how many times it is taken or returns. When the compiled
3952 program exits it saves this data to a file called
3953 @file{@var{auxname}.gcda} for each source file. The data may be used for
3954 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3955 test coverage analysis (@option{-ftest-coverage}). Each object file's
3956 @var{auxname} is generated from the name of the output file, if
3957 explicitly specified and it is not the final executable, otherwise it is
3958 the basename of the source file. In both cases any suffix is removed
3959 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3960 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3961 @xref{Cross-profiling}.
3963 @cindex @command{gcov}
3967 This option is used to compile and link code instrumented for coverage
3968 analysis. The option is a synonym for @option{-fprofile-arcs}
3969 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3970 linking). See the documentation for those options for more details.
3975 Compile the source files with @option{-fprofile-arcs} plus optimization
3976 and code generation options. For test coverage analysis, use the
3977 additional @option{-ftest-coverage} option. You do not need to profile
3978 every source file in a program.
3981 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3982 (the latter implies the former).
3985 Run the program on a representative workload to generate the arc profile
3986 information. This may be repeated any number of times. You can run
3987 concurrent instances of your program, and provided that the file system
3988 supports locking, the data files will be correctly updated. Also
3989 @code{fork} calls are detected and correctly handled (double counting
3993 For profile-directed optimizations, compile the source files again with
3994 the same optimization and code generation options plus
3995 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3996 Control Optimization}).
3999 For test coverage analysis, use @command{gcov} to produce human readable
4000 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4001 @command{gcov} documentation for further information.
4005 With @option{-fprofile-arcs}, for each function of your program GCC
4006 creates a program flow graph, then finds a spanning tree for the graph.
4007 Only arcs that are not on the spanning tree have to be instrumented: the
4008 compiler adds code to count the number of times that these arcs are
4009 executed. When an arc is the only exit or only entrance to a block, the
4010 instrumentation code can be added to the block; otherwise, a new basic
4011 block must be created to hold the instrumentation code.
4014 @item -ftest-coverage
4015 @opindex ftest-coverage
4016 Produce a notes file that the @command{gcov} code-coverage utility
4017 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4018 show program coverage. Each source file's note file is called
4019 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4020 above for a description of @var{auxname} and instructions on how to
4021 generate test coverage data. Coverage data will match the source files
4022 more closely, if you do not optimize.
4024 @item -d@var{letters}
4025 @item -fdump-rtl-@var{pass}
4027 Says to make debugging dumps during compilation at times specified by
4028 @var{letters}. This is used for debugging the RTL-based passes of the
4029 compiler. The file names for most of the dumps are made by appending a
4030 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4031 from the name of the output file, if explicitly specified and it is not
4032 an executable, otherwise it is the basename of the source file.
4034 Most debug dumps can be enabled either passing a letter to the @option{-d}
4035 option, or with a long @option{-fdump-rtl} switch; here are the possible
4036 letters for use in @var{letters} and @var{pass}, and their meanings:
4041 Annotate the assembler output with miscellaneous debugging information.
4044 @itemx -fdump-rtl-bbro
4046 @opindex fdump-rtl-bbro
4047 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4050 @itemx -fdump-rtl-combine
4052 @opindex fdump-rtl-combine
4053 Dump after the RTL instruction combination pass, to the file
4054 @file{@var{file}.129r.combine}.
4057 @itemx -fdump-rtl-ce1
4058 @itemx -fdump-rtl-ce2
4060 @opindex fdump-rtl-ce1
4061 @opindex fdump-rtl-ce2
4062 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
4063 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
4064 and @option{-fdump-rtl-ce2} enable dumping after the second if
4065 conversion, to the file @file{@var{file}.130r.ce2}.
4068 @itemx -fdump-rtl-btl
4069 @itemx -fdump-rtl-dbr
4071 @opindex fdump-rtl-btl
4072 @opindex fdump-rtl-dbr
4073 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
4074 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
4075 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
4076 scheduling, to @file{@var{file}.36.dbr}.
4080 Dump all macro definitions, at the end of preprocessing, in addition to
4084 @itemx -fdump-rtl-ce3
4086 @opindex fdump-rtl-ce3
4087 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4090 @itemx -fdump-rtl-cfg
4091 @itemx -fdump-rtl-life
4093 @opindex fdump-rtl-cfg
4094 @opindex fdump-rtl-life
4095 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
4096 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
4097 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4098 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4101 @itemx -fdump-rtl-greg
4103 @opindex fdump-rtl-greg
4104 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4107 @itemx -fdump-rtl-gcse
4108 @itemx -fdump-rtl-bypass
4110 @opindex fdump-rtl-gcse
4111 @opindex fdump-rtl-bypass
4112 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4113 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
4114 enable dumping after jump bypassing and control flow optimizations, to
4115 @file{@var{file}.115r.bypass}.
4118 @itemx -fdump-rtl-eh
4120 @opindex fdump-rtl-eh
4121 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4124 @itemx -fdump-rtl-sibling
4126 @opindex fdump-rtl-sibling
4127 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4130 @itemx -fdump-rtl-jump
4132 @opindex fdump-rtl-jump
4133 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4136 @itemx -fdump-rtl-stack
4138 @opindex fdump-rtl-stack
4139 Dump after conversion from GCC's "flat register file" registers to the
4140 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4143 @itemx -fdump-rtl-lreg
4145 @opindex fdump-rtl-lreg
4146 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4149 @itemx -fdump-rtl-loop2
4151 @opindex fdump-rtl-loop2
4152 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4153 loop optimization pass, to @file{@var{file}.119r.loop2},
4154 @file{@var{file}.120r.loop2_init},
4155 @file{@var{file}.121r.loop2_invariant}, and
4156 @file{@var{file}.125r.loop2_done}.
4159 @itemx -fdump-rtl-sms
4161 @opindex fdump-rtl-sms
4162 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4165 @itemx -fdump-rtl-mach
4167 @opindex fdump-rtl-mach
4168 Dump after performing the machine dependent reorganization pass, to
4169 @file{@var{file}.155r.mach}.
4172 @itemx -fdump-rtl-rnreg
4174 @opindex fdump-rtl-rnreg
4175 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4178 @itemx -fdump-rtl-regmove
4180 @opindex fdump-rtl-regmove
4181 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4184 @itemx -fdump-rtl-postreload
4186 @opindex fdump-rtl-postreload
4187 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4190 @itemx -fdump-rtl-expand
4192 @opindex fdump-rtl-expand
4193 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4196 @itemx -fdump-rtl-sched2
4198 @opindex fdump-rtl-sched2
4199 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4202 @itemx -fdump-rtl-cse
4204 @opindex fdump-rtl-cse
4205 Dump after CSE (including the jump optimization that sometimes follows
4206 CSE), to @file{@var{file}.113r.cse}.
4209 @itemx -fdump-rtl-sched1
4211 @opindex fdump-rtl-sched1
4212 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4215 @itemx -fdump-rtl-cse2
4217 @opindex fdump-rtl-cse2
4218 Dump after the second CSE pass (including the jump optimization that
4219 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4222 @itemx -fdump-rtl-tracer
4224 @opindex fdump-rtl-tracer
4225 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4228 @itemx -fdump-rtl-vpt
4229 @itemx -fdump-rtl-vartrack
4231 @opindex fdump-rtl-vpt
4232 @opindex fdump-rtl-vartrack
4233 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4234 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4235 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4236 to @file{@var{file}.154r.vartrack}.
4239 @itemx -fdump-rtl-flow2
4241 @opindex fdump-rtl-flow2
4242 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4245 @itemx -fdump-rtl-peephole2
4247 @opindex fdump-rtl-peephole2
4248 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4251 @itemx -fdump-rtl-web
4253 @opindex fdump-rtl-web
4254 Dump after live range splitting, to @file{@var{file}.126r.web}.
4257 @itemx -fdump-rtl-all
4259 @opindex fdump-rtl-all
4260 Produce all the dumps listed above.
4264 Produce a core dump whenever an error occurs.
4268 Print statistics on memory usage, at the end of the run, to
4273 Annotate the assembler output with a comment indicating which
4274 pattern and alternative was used. The length of each instruction is
4279 Dump the RTL in the assembler output as a comment before each instruction.
4280 Also turns on @option{-dp} annotation.
4284 For each of the other indicated dump files (either with @option{-d} or
4285 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4286 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4290 Just generate RTL for a function instead of compiling it. Usually used
4291 with @samp{r} (@option{-fdump-rtl-expand}).
4295 Dump debugging information during parsing, to standard error.
4299 @opindex fdump-noaddr
4300 When doing debugging dumps (see @option{-d} option above), suppress
4301 address output. This makes it more feasible to use diff on debugging
4302 dumps for compiler invocations with different compiler binaries and/or
4303 different text / bss / data / heap / stack / dso start locations.
4305 @item -fdump-unnumbered
4306 @opindex fdump-unnumbered
4307 When doing debugging dumps (see @option{-d} option above), suppress instruction
4308 numbers, line number note and address output. This makes it more feasible to
4309 use diff on debugging dumps for compiler invocations with different
4310 options, in particular with and without @option{-g}.
4312 @item -fdump-translation-unit @r{(C++ only)}
4313 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4314 @opindex fdump-translation-unit
4315 Dump a representation of the tree structure for the entire translation
4316 unit to a file. The file name is made by appending @file{.tu} to the
4317 source file name. If the @samp{-@var{options}} form is used, @var{options}
4318 controls the details of the dump as described for the
4319 @option{-fdump-tree} options.
4321 @item -fdump-class-hierarchy @r{(C++ only)}
4322 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4323 @opindex fdump-class-hierarchy
4324 Dump a representation of each class's hierarchy and virtual function
4325 table layout to a file. The file name is made by appending @file{.class}
4326 to the source file name. If the @samp{-@var{options}} form is used,
4327 @var{options} controls the details of the dump as described for the
4328 @option{-fdump-tree} options.
4330 @item -fdump-ipa-@var{switch}
4332 Control the dumping at various stages of inter-procedural analysis
4333 language tree to a file. The file name is generated by appending a switch
4334 specific suffix to the source file name. The following dumps are possible:
4338 Enables all inter-procedural analysis dumps; currently the only produced
4339 dump is the @samp{cgraph} dump.
4342 Dumps information about call-graph optimization, unused function removal,
4343 and inlining decisions.
4346 @item -fdump-tree-@var{switch}
4347 @itemx -fdump-tree-@var{switch}-@var{options}
4349 Control the dumping at various stages of processing the intermediate
4350 language tree to a file. The file name is generated by appending a switch
4351 specific suffix to the source file name. If the @samp{-@var{options}}
4352 form is used, @var{options} is a list of @samp{-} separated options that
4353 control the details of the dump. Not all options are applicable to all
4354 dumps, those which are not meaningful will be ignored. The following
4355 options are available
4359 Print the address of each node. Usually this is not meaningful as it
4360 changes according to the environment and source file. Its primary use
4361 is for tying up a dump file with a debug environment.
4363 Inhibit dumping of members of a scope or body of a function merely
4364 because that scope has been reached. Only dump such items when they
4365 are directly reachable by some other path. When dumping pretty-printed
4366 trees, this option inhibits dumping the bodies of control structures.
4368 Print a raw representation of the tree. By default, trees are
4369 pretty-printed into a C-like representation.
4371 Enable more detailed dumps (not honored by every dump option).
4373 Enable dumping various statistics about the pass (not honored by every dump
4376 Enable showing basic block boundaries (disabled in raw dumps).
4378 Enable showing virtual operands for every statement.
4380 Enable showing line numbers for statements.
4382 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4384 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4387 The following tree dumps are possible:
4391 Dump before any tree based optimization, to @file{@var{file}.original}.
4394 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4397 Dump after function inlining, to @file{@var{file}.inlined}.
4400 @opindex fdump-tree-gimple
4401 Dump each function before and after the gimplification pass to a file. The
4402 file name is made by appending @file{.gimple} to the source file name.
4405 @opindex fdump-tree-cfg
4406 Dump the control flow graph of each function to a file. The file name is
4407 made by appending @file{.cfg} to the source file name.
4410 @opindex fdump-tree-vcg
4411 Dump the control flow graph of each function to a file in VCG format. The
4412 file name is made by appending @file{.vcg} to the source file name. Note
4413 that if the file contains more than one function, the generated file cannot
4414 be used directly by VCG@. You will need to cut and paste each function's
4415 graph into its own separate file first.
4418 @opindex fdump-tree-ch
4419 Dump each function after copying loop headers. The file name is made by
4420 appending @file{.ch} to the source file name.
4423 @opindex fdump-tree-ssa
4424 Dump SSA related information to a file. The file name is made by appending
4425 @file{.ssa} to the source file name.
4428 @opindex fdump-tree-salias
4429 Dump structure aliasing variable information to a file. This file name
4430 is made by appending @file{.salias} to the source file name.
4433 @opindex fdump-tree-alias
4434 Dump aliasing information for each function. The file name is made by
4435 appending @file{.alias} to the source file name.
4438 @opindex fdump-tree-ccp
4439 Dump each function after CCP@. The file name is made by appending
4440 @file{.ccp} to the source file name.
4443 @opindex fdump-tree-storeccp
4444 Dump each function after STORE-CCP. The file name is made by appending
4445 @file{.storeccp} to the source file name.
4448 @opindex fdump-tree-pre
4449 Dump trees after partial redundancy elimination. The file name is made
4450 by appending @file{.pre} to the source file name.
4453 @opindex fdump-tree-fre
4454 Dump trees after full redundancy elimination. The file name is made
4455 by appending @file{.fre} to the source file name.
4458 @opindex fdump-tree-copyprop
4459 Dump trees after copy propagation. The file name is made
4460 by appending @file{.copyprop} to the source file name.
4462 @item store_copyprop
4463 @opindex fdump-tree-store_copyprop
4464 Dump trees after store copy-propagation. The file name is made
4465 by appending @file{.store_copyprop} to the source file name.
4468 @opindex fdump-tree-dce
4469 Dump each function after dead code elimination. The file name is made by
4470 appending @file{.dce} to the source file name.
4473 @opindex fdump-tree-mudflap
4474 Dump each function after adding mudflap instrumentation. The file name is
4475 made by appending @file{.mudflap} to the source file name.
4478 @opindex fdump-tree-sra
4479 Dump each function after performing scalar replacement of aggregates. The
4480 file name is made by appending @file{.sra} to the source file name.
4483 @opindex fdump-tree-sink
4484 Dump each function after performing code sinking. The file name is made
4485 by appending @file{.sink} to the source file name.
4488 @opindex fdump-tree-dom
4489 Dump each function after applying dominator tree optimizations. The file
4490 name is made by appending @file{.dom} to the source file name.
4493 @opindex fdump-tree-dse
4494 Dump each function after applying dead store elimination. The file
4495 name is made by appending @file{.dse} to the source file name.
4498 @opindex fdump-tree-phiopt
4499 Dump each function after optimizing PHI nodes into straightline code. The file
4500 name is made by appending @file{.phiopt} to the source file name.
4503 @opindex fdump-tree-forwprop
4504 Dump each function after forward propagating single use variables. The file
4505 name is made by appending @file{.forwprop} to the source file name.
4508 @opindex fdump-tree-copyrename
4509 Dump each function after applying the copy rename optimization. The file
4510 name is made by appending @file{.copyrename} to the source file name.
4513 @opindex fdump-tree-nrv
4514 Dump each function after applying the named return value optimization on
4515 generic trees. The file name is made by appending @file{.nrv} to the source
4519 @opindex fdump-tree-vect
4520 Dump each function after applying vectorization of loops. The file name is
4521 made by appending @file{.vect} to the source file name.
4524 @opindex fdump-tree-vrp
4525 Dump each function after Value Range Propagation (VRP). The file name
4526 is made by appending @file{.vrp} to the source file name.
4529 @opindex fdump-tree-all
4530 Enable all the available tree dumps with the flags provided in this option.
4533 @item -ftree-vectorizer-verbose=@var{n}
4534 @opindex ftree-vectorizer-verbose
4535 This option controls the amount of debugging output the vectorizer prints.
4536 This information is written to standard error, unless
4537 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4538 in which case it is output to the usual dump listing file, @file{.vect}.
4539 For @var{n}=0 no diagnostic information is reported.
4540 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4541 and the total number of loops that got vectorized.
4542 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4543 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4544 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4545 level that @option{-fdump-tree-vect-stats} uses.
4546 Higher verbosity levels mean either more information dumped for each
4547 reported loop, or same amount of information reported for more loops:
4548 If @var{n}=3, alignment related information is added to the reports.
4549 If @var{n}=4, data-references related information (e.g. memory dependences,
4550 memory access-patterns) is added to the reports.
4551 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4552 that did not pass the first analysis phase (i.e. may not be countable, or
4553 may have complicated control-flow).
4554 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4555 For @var{n}=7, all the information the vectorizer generates during its
4556 analysis and transformation is reported. This is the same verbosity level
4557 that @option{-fdump-tree-vect-details} uses.
4559 @item -frandom-seed=@var{string}
4560 @opindex frandom-string
4561 This option provides a seed that GCC uses when it would otherwise use
4562 random numbers. It is used to generate certain symbol names
4563 that have to be different in every compiled file. It is also used to
4564 place unique stamps in coverage data files and the object files that
4565 produce them. You can use the @option{-frandom-seed} option to produce
4566 reproducibly identical object files.
4568 The @var{string} should be different for every file you compile.
4570 @item -fsched-verbose=@var{n}
4571 @opindex fsched-verbose
4572 On targets that use instruction scheduling, this option controls the
4573 amount of debugging output the scheduler prints. This information is
4574 written to standard error, unless @option{-dS} or @option{-dR} is
4575 specified, in which case it is output to the usual dump
4576 listing file, @file{.sched} or @file{.sched2} respectively. However
4577 for @var{n} greater than nine, the output is always printed to standard
4580 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4581 same information as @option{-dRS}. For @var{n} greater than one, it
4582 also output basic block probabilities, detailed ready list information
4583 and unit/insn info. For @var{n} greater than two, it includes RTL
4584 at abort point, control-flow and regions info. And for @var{n} over
4585 four, @option{-fsched-verbose} also includes dependence info.
4589 Store the usual ``temporary'' intermediate files permanently; place them
4590 in the current directory and name them based on the source file. Thus,
4591 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4592 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4593 preprocessed @file{foo.i} output file even though the compiler now
4594 normally uses an integrated preprocessor.
4596 When used in combination with the @option{-x} command line option,
4597 @option{-save-temps} is sensible enough to avoid over writing an
4598 input source file with the same extension as an intermediate file.
4599 The corresponding intermediate file may be obtained by renaming the
4600 source file before using @option{-save-temps}.
4604 Report the CPU time taken by each subprocess in the compilation
4605 sequence. For C source files, this is the compiler proper and assembler
4606 (plus the linker if linking is done). The output looks like this:
4613 The first number on each line is the ``user time'', that is time spent
4614 executing the program itself. The second number is ``system time'',
4615 time spent executing operating system routines on behalf of the program.
4616 Both numbers are in seconds.
4618 @item -fvar-tracking
4619 @opindex fvar-tracking
4620 Run variable tracking pass. It computes where variables are stored at each
4621 position in code. Better debugging information is then generated
4622 (if the debugging information format supports this information).
4624 It is enabled by default when compiling with optimization (@option{-Os},
4625 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4626 the debug info format supports it.
4628 @item -print-file-name=@var{library}
4629 @opindex print-file-name
4630 Print the full absolute name of the library file @var{library} that
4631 would be used when linking---and don't do anything else. With this
4632 option, GCC does not compile or link anything; it just prints the
4635 @item -print-multi-directory
4636 @opindex print-multi-directory
4637 Print the directory name corresponding to the multilib selected by any
4638 other switches present in the command line. This directory is supposed
4639 to exist in @env{GCC_EXEC_PREFIX}.
4641 @item -print-multi-lib
4642 @opindex print-multi-lib
4643 Print the mapping from multilib directory names to compiler switches
4644 that enable them. The directory name is separated from the switches by
4645 @samp{;}, and each switch starts with an @samp{@@} instead of the
4646 @samp{-}, without spaces between multiple switches. This is supposed to
4647 ease shell-processing.
4649 @item -print-prog-name=@var{program}
4650 @opindex print-prog-name
4651 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4653 @item -print-libgcc-file-name
4654 @opindex print-libgcc-file-name
4655 Same as @option{-print-file-name=libgcc.a}.
4657 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4658 but you do want to link with @file{libgcc.a}. You can do
4661 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4664 @item -print-search-dirs
4665 @opindex print-search-dirs
4666 Print the name of the configured installation directory and a list of
4667 program and library directories @command{gcc} will search---and don't do anything else.
4669 This is useful when @command{gcc} prints the error message
4670 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4671 To resolve this you either need to put @file{cpp0} and the other compiler
4672 components where @command{gcc} expects to find them, or you can set the environment
4673 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4674 Don't forget the trailing @samp{/}.
4675 @xref{Environment Variables}.
4678 @opindex dumpmachine
4679 Print the compiler's target machine (for example,
4680 @samp{i686-pc-linux-gnu})---and don't do anything else.
4683 @opindex dumpversion
4684 Print the compiler version (for example, @samp{3.0})---and don't do
4689 Print the compiler's built-in specs---and don't do anything else. (This
4690 is used when GCC itself is being built.) @xref{Spec Files}.
4692 @item -feliminate-unused-debug-types
4693 @opindex feliminate-unused-debug-types
4694 Normally, when producing DWARF2 output, GCC will emit debugging
4695 information for all types declared in a compilation
4696 unit, regardless of whether or not they are actually used
4697 in that compilation unit. Sometimes this is useful, such as
4698 if, in the debugger, you want to cast a value to a type that is
4699 not actually used in your program (but is declared). More often,
4700 however, this results in a significant amount of wasted space.
4701 With this option, GCC will avoid producing debug symbol output
4702 for types that are nowhere used in the source file being compiled.
4705 @node Optimize Options
4706 @section Options That Control Optimization
4707 @cindex optimize options
4708 @cindex options, optimization
4710 These options control various sorts of optimizations.
4712 Without any optimization option, the compiler's goal is to reduce the
4713 cost of compilation and to make debugging produce the expected
4714 results. Statements are independent: if you stop the program with a
4715 breakpoint between statements, you can then assign a new value to any
4716 variable or change the program counter to any other statement in the
4717 function and get exactly the results you would expect from the source
4720 Turning on optimization flags makes the compiler attempt to improve
4721 the performance and/or code size at the expense of compilation time
4722 and possibly the ability to debug the program.
4724 The compiler performs optimization based on the knowledge it has of
4725 the program. Optimization levels @option{-O} and above, in
4726 particular, enable @emph{unit-at-a-time} mode, which allows the
4727 compiler to consider information gained from later functions in
4728 the file when compiling a function. Compiling multiple files at
4729 once to a single output file in @emph{unit-at-a-time} mode allows
4730 the compiler to use information gained from all of the files when
4731 compiling each of them.
4733 Not all optimizations are controlled directly by a flag. Only
4734 optimizations that have a flag are listed.
4741 Optimize. Optimizing compilation takes somewhat more time, and a lot
4742 more memory for a large function.
4744 With @option{-O}, the compiler tries to reduce code size and execution
4745 time, without performing any optimizations that take a great deal of
4748 @option{-O} turns on the following optimization flags:
4749 @gccoptlist{-fdefer-pop @gol
4750 -fdelayed-branch @gol
4751 -fguess-branch-probability @gol
4752 -fcprop-registers @gol
4753 -fif-conversion @gol
4754 -fif-conversion2 @gol
4755 -fsplit-wide-types @gol
4758 -ftree-dominator-opts @gol
4762 -ftree-copyrename @gol
4765 -funit-at-a-time @gol
4768 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4769 where doing so does not interfere with debugging.
4773 Optimize even more. GCC performs nearly all supported optimizations
4774 that do not involve a space-speed tradeoff. The compiler does not
4775 perform loop unrolling or function inlining when you specify @option{-O2}.
4776 As compared to @option{-O}, this option increases both compilation time
4777 and the performance of the generated code.
4779 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4780 also turns on the following optimization flags:
4781 @gccoptlist{-fthread-jumps @gol
4783 -foptimize-sibling-calls @gol
4784 -fcse-follow-jumps -fcse-skip-blocks @gol
4785 -fgcse -fgcse-lm @gol
4786 -fexpensive-optimizations @gol
4787 -frerun-cse-after-loop @gol
4790 -fschedule-insns -fschedule-insns2 @gol
4791 -fsched-interblock -fsched-spec @gol
4793 -fstrict-aliasing -fstrict-overflow @gol
4794 -fdelete-null-pointer-checks @gol
4795 -freorder-blocks -freorder-functions @gol
4796 -falign-functions -falign-jumps @gol
4797 -falign-loops -falign-labels @gol
4801 Please note the warning under @option{-fgcse} about
4802 invoking @option{-O2} on programs that use computed gotos.
4806 Optimize yet more. @option{-O3} turns on all optimizations specified by
4807 @option{-O2} and also turns on the @option{-finline-functions},
4808 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4812 Reduce compilation time and make debugging produce the expected
4813 results. This is the default.
4817 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4818 do not typically increase code size. It also performs further
4819 optimizations designed to reduce code size.
4821 @option{-Os} disables the following optimization flags:
4822 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4823 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4824 -fprefetch-loop-arrays -ftree-vect-loop-version}
4826 If you use multiple @option{-O} options, with or without level numbers,
4827 the last such option is the one that is effective.
4830 Options of the form @option{-f@var{flag}} specify machine-independent
4831 flags. Most flags have both positive and negative forms; the negative
4832 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4833 below, only one of the forms is listed---the one you typically will
4834 use. You can figure out the other form by either removing @samp{no-}
4837 The following options control specific optimizations. They are either
4838 activated by @option{-O} options or are related to ones that are. You
4839 can use the following flags in the rare cases when ``fine-tuning'' of
4840 optimizations to be performed is desired.
4843 @item -fno-default-inline
4844 @opindex fno-default-inline
4845 Do not make member functions inline by default merely because they are
4846 defined inside the class scope (C++ only). Otherwise, when you specify
4847 @w{@option{-O}}, member functions defined inside class scope are compiled
4848 inline by default; i.e., you don't need to add @samp{inline} in front of
4849 the member function name.
4851 @item -fno-defer-pop
4852 @opindex fno-defer-pop
4853 Always pop the arguments to each function call as soon as that function
4854 returns. For machines which must pop arguments after a function call,
4855 the compiler normally lets arguments accumulate on the stack for several
4856 function calls and pops them all at once.
4858 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4861 @opindex fforce-addr
4862 Force memory address constants to be copied into registers before
4863 doing arithmetic on them.
4865 @item -fforward-propagate
4866 @opindex fforward-propagate
4867 Perform a forward propagation pass on RTL. The pass tries to combine two
4868 instructions and checks if the result can be simplified. If loop unrolling
4869 is active, two passes are performed and the second is scheduled after
4872 This option is enabled by default at optimization levels @option{-O2},
4873 @option{-O3}, @option{-Os}.
4875 @item -fomit-frame-pointer
4876 @opindex fomit-frame-pointer
4877 Don't keep the frame pointer in a register for functions that
4878 don't need one. This avoids the instructions to save, set up and
4879 restore frame pointers; it also makes an extra register available
4880 in many functions. @strong{It also makes debugging impossible on
4883 On some machines, such as the VAX, this flag has no effect, because
4884 the standard calling sequence automatically handles the frame pointer
4885 and nothing is saved by pretending it doesn't exist. The
4886 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4887 whether a target machine supports this flag. @xref{Registers,,Register
4888 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4890 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4892 @item -foptimize-sibling-calls
4893 @opindex foptimize-sibling-calls
4894 Optimize sibling and tail recursive calls.
4896 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4900 Don't pay attention to the @code{inline} keyword. Normally this option
4901 is used to keep the compiler from expanding any functions inline.
4902 Note that if you are not optimizing, no functions can be expanded inline.
4904 @item -finline-functions
4905 @opindex finline-functions
4906 Integrate all simple functions into their callers. The compiler
4907 heuristically decides which functions are simple enough to be worth
4908 integrating in this way.
4910 If all calls to a given function are integrated, and the function is
4911 declared @code{static}, then the function is normally not output as
4912 assembler code in its own right.
4914 Enabled at level @option{-O3}.
4916 @item -finline-functions-called-once
4917 @opindex finline-functions-called-once
4918 Consider all @code{static} functions called once for inlining into their
4919 caller even if they are not marked @code{inline}. If a call to a given
4920 function is integrated, then the function is not output as assembler code
4923 Enabled if @option{-funit-at-a-time} is enabled.
4925 @item -fearly-inlining
4926 @opindex fearly-inlining
4927 Inline functions marked by @code{always_inline} and functions whose body seems
4928 smaller than the function call overhead early before doing
4929 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4930 makes profiling significantly cheaper and usually inlining faster on programs
4931 having large chains of nested wrapper functions.
4935 @item -finline-limit=@var{n}
4936 @opindex finline-limit
4937 By default, GCC limits the size of functions that can be inlined. This flag
4938 allows the control of this limit for functions that are explicitly marked as
4939 inline (i.e., marked with the inline keyword or defined within the class
4940 definition in c++). @var{n} is the size of functions that can be inlined in
4941 number of pseudo instructions (not counting parameter handling). The default
4942 value of @var{n} is 600.
4943 Increasing this value can result in more inlined code at
4944 the cost of compilation time and memory consumption. Decreasing usually makes
4945 the compilation faster and less code will be inlined (which presumably
4946 means slower programs). This option is particularly useful for programs that
4947 use inlining heavily such as those based on recursive templates with C++.
4949 Inlining is actually controlled by a number of parameters, which may be
4950 specified individually by using @option{--param @var{name}=@var{value}}.
4951 The @option{-finline-limit=@var{n}} option sets some of these parameters
4955 @item max-inline-insns-single
4956 is set to @var{n}/2.
4957 @item max-inline-insns-auto
4958 is set to @var{n}/2.
4959 @item min-inline-insns
4960 is set to 130 or @var{n}/4, whichever is smaller.
4961 @item max-inline-insns-rtl
4965 See below for a documentation of the individual
4966 parameters controlling inlining.
4968 @emph{Note:} pseudo instruction represents, in this particular context, an
4969 abstract measurement of function's size. In no way does it represent a count
4970 of assembly instructions and as such its exact meaning might change from one
4971 release to an another.
4973 @item -fkeep-inline-functions
4974 @opindex fkeep-inline-functions
4975 In C, emit @code{static} functions that are declared @code{inline}
4976 into the object file, even if the function has been inlined into all
4977 of its callers. This switch does not affect functions using the
4978 @code{extern inline} extension in GNU C89@. In C++, emit any and all
4979 inline functions into the object file.
4981 @item -fkeep-static-consts
4982 @opindex fkeep-static-consts
4983 Emit variables declared @code{static const} when optimization isn't turned
4984 on, even if the variables aren't referenced.
4986 GCC enables this option by default. If you want to force the compiler to
4987 check if the variable was referenced, regardless of whether or not
4988 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4990 @item -fmerge-constants
4991 Attempt to merge identical constants (string constants and floating point
4992 constants) across compilation units.
4994 This option is the default for optimized compilation if the assembler and
4995 linker support it. Use @option{-fno-merge-constants} to inhibit this
4998 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5000 @item -fmerge-all-constants
5001 Attempt to merge identical constants and identical variables.
5003 This option implies @option{-fmerge-constants}. In addition to
5004 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5005 arrays or initialized constant variables with integral or floating point
5006 types. Languages like C or C++ require each non-automatic variable to
5007 have distinct location, so using this option will result in non-conforming
5010 @item -fmodulo-sched
5011 @opindex fmodulo-sched
5012 Perform swing modulo scheduling immediately before the first scheduling
5013 pass. This pass looks at innermost loops and reorders their
5014 instructions by overlapping different iterations.
5016 @item -fno-branch-count-reg
5017 @opindex fno-branch-count-reg
5018 Do not use ``decrement and branch'' instructions on a count register,
5019 but instead generate a sequence of instructions that decrement a
5020 register, compare it against zero, then branch based upon the result.
5021 This option is only meaningful on architectures that support such
5022 instructions, which include x86, PowerPC, IA-64 and S/390.
5024 The default is @option{-fbranch-count-reg}.
5026 @item -fno-function-cse
5027 @opindex fno-function-cse
5028 Do not put function addresses in registers; make each instruction that
5029 calls a constant function contain the function's address explicitly.
5031 This option results in less efficient code, but some strange hacks
5032 that alter the assembler output may be confused by the optimizations
5033 performed when this option is not used.
5035 The default is @option{-ffunction-cse}
5037 @item -fno-zero-initialized-in-bss
5038 @opindex fno-zero-initialized-in-bss
5039 If the target supports a BSS section, GCC by default puts variables that
5040 are initialized to zero into BSS@. This can save space in the resulting
5043 This option turns off this behavior because some programs explicitly
5044 rely on variables going to the data section. E.g., so that the
5045 resulting executable can find the beginning of that section and/or make
5046 assumptions based on that.
5048 The default is @option{-fzero-initialized-in-bss}.
5050 @item -fbounds-check
5051 @opindex fbounds-check
5052 For front-ends that support it, generate additional code to check that
5053 indices used to access arrays are within the declared range. This is
5054 currently only supported by the Java and Fortran front-ends, where
5055 this option defaults to true and false respectively.
5057 @item -fmudflap -fmudflapth -fmudflapir
5061 @cindex bounds checking
5063 For front-ends that support it (C and C++), instrument all risky
5064 pointer/array dereferencing operations, some standard library
5065 string/heap functions, and some other associated constructs with
5066 range/validity tests. Modules so instrumented should be immune to
5067 buffer overflows, invalid heap use, and some other classes of C/C++
5068 programming errors. The instrumentation relies on a separate runtime
5069 library (@file{libmudflap}), which will be linked into a program if
5070 @option{-fmudflap} is given at link time. Run-time behavior of the
5071 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5072 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5075 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5076 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5077 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5078 instrumentation should ignore pointer reads. This produces less
5079 instrumentation (and therefore faster execution) and still provides
5080 some protection against outright memory corrupting writes, but allows
5081 erroneously read data to propagate within a program.
5083 @item -fthread-jumps
5084 @opindex fthread-jumps
5085 Perform optimizations where we check to see if a jump branches to a
5086 location where another comparison subsumed by the first is found. If
5087 so, the first branch is redirected to either the destination of the
5088 second branch or a point immediately following it, depending on whether
5089 the condition is known to be true or false.
5091 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5093 @item -fsplit-wide-types
5094 @opindex -fsplit-wide-types
5095 When using a type that occupies multiple registers, such as @code{long
5096 long} on a 32-bit system, split the registers apart and allocate them
5097 independently. This normally generates better code for those types,
5098 but may make debugging more difficult.
5100 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5103 @item -fcse-follow-jumps
5104 @opindex fcse-follow-jumps
5105 In common subexpression elimination, scan through jump instructions
5106 when the target of the jump is not reached by any other path. For
5107 example, when CSE encounters an @code{if} statement with an
5108 @code{else} clause, CSE will follow the jump when the condition
5111 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5113 @item -fcse-skip-blocks
5114 @opindex fcse-skip-blocks
5115 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5116 follow jumps which conditionally skip over blocks. When CSE
5117 encounters a simple @code{if} statement with no else clause,
5118 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5119 body of the @code{if}.
5121 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5123 @item -frerun-cse-after-loop
5124 @opindex frerun-cse-after-loop
5125 Re-run common subexpression elimination after loop optimizations has been
5128 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5132 Perform a global common subexpression elimination pass.
5133 This pass also performs global constant and copy propagation.
5135 @emph{Note:} When compiling a program using computed gotos, a GCC
5136 extension, you may get better runtime performance if you disable
5137 the global common subexpression elimination pass by adding
5138 @option{-fno-gcse} to the command line.
5140 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5144 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5145 attempt to move loads which are only killed by stores into themselves. This
5146 allows a loop containing a load/store sequence to be changed to a load outside
5147 the loop, and a copy/store within the loop.
5149 Enabled by default when gcse is enabled.
5153 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5154 global common subexpression elimination. This pass will attempt to move
5155 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5156 loops containing a load/store sequence can be changed to a load before
5157 the loop and a store after the loop.
5159 Not enabled at any optimization level.
5163 When @option{-fgcse-las} is enabled, the global common subexpression
5164 elimination pass eliminates redundant loads that come after stores to the
5165 same memory location (both partial and full redundancies).
5167 Not enabled at any optimization level.
5169 @item -fgcse-after-reload
5170 @opindex fgcse-after-reload
5171 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5172 pass is performed after reload. The purpose of this pass is to cleanup
5175 @item -funsafe-loop-optimizations
5176 @opindex funsafe-loop-optimizations
5177 If given, the loop optimizer will assume that loop indices do not
5178 overflow, and that the loops with nontrivial exit condition are not
5179 infinite. This enables a wider range of loop optimizations even if
5180 the loop optimizer itself cannot prove that these assumptions are valid.
5181 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5182 if it finds this kind of loop.
5184 @item -fcrossjumping
5185 @opindex crossjumping
5186 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5187 resulting code may or may not perform better than without cross-jumping.
5189 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5191 @item -fif-conversion
5192 @opindex if-conversion
5193 Attempt to transform conditional jumps into branch-less equivalents. This
5194 include use of conditional moves, min, max, set flags and abs instructions, and
5195 some tricks doable by standard arithmetics. The use of conditional execution
5196 on chips where it is available is controlled by @code{if-conversion2}.
5198 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5200 @item -fif-conversion2
5201 @opindex if-conversion2
5202 Use conditional execution (where available) to transform conditional jumps into
5203 branch-less equivalents.
5205 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5207 @item -fdelete-null-pointer-checks
5208 @opindex fdelete-null-pointer-checks
5209 Use global dataflow analysis to identify and eliminate useless checks
5210 for null pointers. The compiler assumes that dereferencing a null
5211 pointer would have halted the program. If a pointer is checked after
5212 it has already been dereferenced, it cannot be null.
5214 In some environments, this assumption is not true, and programs can
5215 safely dereference null pointers. Use
5216 @option{-fno-delete-null-pointer-checks} to disable this optimization
5217 for programs which depend on that behavior.
5219 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5221 @item -fexpensive-optimizations
5222 @opindex fexpensive-optimizations
5223 Perform a number of minor optimizations that are relatively expensive.
5225 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5227 @item -foptimize-register-move
5229 @opindex foptimize-register-move
5231 Attempt to reassign register numbers in move instructions and as
5232 operands of other simple instructions in order to maximize the amount of
5233 register tying. This is especially helpful on machines with two-operand
5236 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5239 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5241 @item -fdelayed-branch
5242 @opindex fdelayed-branch
5243 If supported for the target machine, attempt to reorder instructions
5244 to exploit instruction slots available after delayed branch
5247 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5249 @item -fschedule-insns
5250 @opindex fschedule-insns
5251 If supported for the target machine, attempt to reorder instructions to
5252 eliminate execution stalls due to required data being unavailable. This
5253 helps machines that have slow floating point or memory load instructions
5254 by allowing other instructions to be issued until the result of the load
5255 or floating point instruction is required.
5257 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5259 @item -fschedule-insns2
5260 @opindex fschedule-insns2
5261 Similar to @option{-fschedule-insns}, but requests an additional pass of
5262 instruction scheduling after register allocation has been done. This is
5263 especially useful on machines with a relatively small number of
5264 registers and where memory load instructions take more than one cycle.
5266 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5268 @item -fno-sched-interblock
5269 @opindex fno-sched-interblock
5270 Don't schedule instructions across basic blocks. This is normally
5271 enabled by default when scheduling before register allocation, i.e.@:
5272 with @option{-fschedule-insns} or at @option{-O2} or higher.
5274 @item -fno-sched-spec
5275 @opindex fno-sched-spec
5276 Don't allow speculative motion of non-load instructions. This is normally
5277 enabled by default when scheduling before register allocation, i.e.@:
5278 with @option{-fschedule-insns} or at @option{-O2} or higher.
5280 @item -fsched-spec-load
5281 @opindex fsched-spec-load
5282 Allow speculative motion of some load instructions. This only makes
5283 sense when scheduling before register allocation, i.e.@: with
5284 @option{-fschedule-insns} or at @option{-O2} or higher.
5286 @item -fsched-spec-load-dangerous
5287 @opindex fsched-spec-load-dangerous
5288 Allow speculative motion of more load instructions. This only makes
5289 sense when scheduling before register allocation, i.e.@: with
5290 @option{-fschedule-insns} or at @option{-O2} or higher.
5292 @item -fsched-stalled-insns=@var{n}
5293 @opindex fsched-stalled-insns
5294 Define how many insns (if any) can be moved prematurely from the queue
5295 of stalled insns into the ready list, during the second scheduling pass.
5297 @item -fsched-stalled-insns-dep=@var{n}
5298 @opindex fsched-stalled-insns-dep
5299 Define how many insn groups (cycles) will be examined for a dependency
5300 on a stalled insn that is candidate for premature removal from the queue
5301 of stalled insns. Has an effect only during the second scheduling pass,
5302 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5304 @item -fsched2-use-superblocks
5305 @opindex fsched2-use-superblocks
5306 When scheduling after register allocation, do use superblock scheduling
5307 algorithm. Superblock scheduling allows motion across basic block boundaries
5308 resulting on faster schedules. This option is experimental, as not all machine
5309 descriptions used by GCC model the CPU closely enough to avoid unreliable
5310 results from the algorithm.
5312 This only makes sense when scheduling after register allocation, i.e.@: with
5313 @option{-fschedule-insns2} or at @option{-O2} or higher.
5315 @item -fsched2-use-traces
5316 @opindex fsched2-use-traces
5317 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5318 allocation and additionally perform code duplication in order to increase the
5319 size of superblocks using tracer pass. See @option{-ftracer} for details on
5322 This mode should produce faster but significantly longer programs. Also
5323 without @option{-fbranch-probabilities} the traces constructed may not
5324 match the reality and hurt the performance. This only makes
5325 sense when scheduling after register allocation, i.e.@: with
5326 @option{-fschedule-insns2} or at @option{-O2} or higher.
5330 Eliminates redundant extension instructions and move the non redundant
5331 ones to optimal placement using LCM.
5333 @item -freschedule-modulo-scheduled-loops
5334 @opindex fscheduling-in-modulo-scheduled-loops
5335 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5336 we may want to prevent the later scheduling passes from changing its schedule, we use this
5337 option to control that.
5339 @item -fcaller-saves
5340 @opindex fcaller-saves
5341 Enable values to be allocated in registers that will be clobbered by
5342 function calls, by emitting extra instructions to save and restore the
5343 registers around such calls. Such allocation is done only when it
5344 seems to result in better code than would otherwise be produced.
5346 This option is always enabled by default on certain machines, usually
5347 those which have no call-preserved registers to use instead.
5349 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5352 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5353 enabled by default at @option{-O2} and @option{-O3}.
5356 Perform Full Redundancy Elimination (FRE) on trees. The difference
5357 between FRE and PRE is that FRE only considers expressions
5358 that are computed on all paths leading to the redundant computation.
5359 This analysis faster than PRE, though it exposes fewer redundancies.
5360 This flag is enabled by default at @option{-O} and higher.
5362 @item -ftree-copy-prop
5363 Perform copy propagation on trees. This pass eliminates unnecessary
5364 copy operations. This flag is enabled by default at @option{-O} and
5367 @item -ftree-store-copy-prop
5368 Perform copy propagation of memory loads and stores. This pass
5369 eliminates unnecessary copy operations in memory references
5370 (structures, global variables, arrays, etc). This flag is enabled by
5371 default at @option{-O2} and higher.
5374 Perform structural alias analysis on trees. This flag
5375 is enabled by default at @option{-O} and higher.
5378 Perform interprocedural pointer analysis.
5381 Perform forward store motion on trees. This flag is
5382 enabled by default at @option{-O} and higher.
5385 Perform sparse conditional constant propagation (CCP) on trees. This
5386 pass only operates on local scalar variables and is enabled by default
5387 at @option{-O} and higher.
5389 @item -ftree-store-ccp
5390 Perform sparse conditional constant propagation (CCP) on trees. This
5391 pass operates on both local scalar variables and memory stores and
5392 loads (global variables, structures, arrays, etc). This flag is
5393 enabled by default at @option{-O2} and higher.
5396 Perform dead code elimination (DCE) on trees. This flag is enabled by
5397 default at @option{-O} and higher.
5399 @item -ftree-dominator-opts
5400 Perform a variety of simple scalar cleanups (constant/copy
5401 propagation, redundancy elimination, range propagation and expression
5402 simplification) based on a dominator tree traversal. This also
5403 performs jump threading (to reduce jumps to jumps). This flag is
5404 enabled by default at @option{-O} and higher.
5407 Perform loop header copying on trees. This is beneficial since it increases
5408 effectiveness of code motion optimizations. It also saves one jump. This flag
5409 is enabled by default at @option{-O} and higher. It is not enabled
5410 for @option{-Os}, since it usually increases code size.
5412 @item -ftree-loop-optimize
5413 Perform loop optimizations on trees. This flag is enabled by default
5414 at @option{-O} and higher.
5416 @item -ftree-loop-linear
5417 Perform linear loop transformations on tree. This flag can improve cache
5418 performance and allow further loop optimizations to take place.
5420 @item -ftree-loop-im
5421 Perform loop invariant motion on trees. This pass moves only invariants that
5422 would be hard to handle at RTL level (function calls, operations that expand to
5423 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5424 operands of conditions that are invariant out of the loop, so that we can use
5425 just trivial invariantness analysis in loop unswitching. The pass also includes
5428 @item -ftree-loop-ivcanon
5429 Create a canonical counter for number of iterations in the loop for that
5430 determining number of iterations requires complicated analysis. Later
5431 optimizations then may determine the number easily. Useful especially
5432 in connection with unrolling.
5435 Perform induction variable optimizations (strength reduction, induction
5436 variable merging and induction variable elimination) on trees.
5439 Perform scalar replacement of aggregates. This pass replaces structure
5440 references with scalars to prevent committing structures to memory too
5441 early. This flag is enabled by default at @option{-O} and higher.
5443 @item -ftree-copyrename
5444 Perform copy renaming on trees. This pass attempts to rename compiler
5445 temporaries to other variables at copy locations, usually resulting in
5446 variable names which more closely resemble the original variables. This flag
5447 is enabled by default at @option{-O} and higher.
5450 Perform temporary expression replacement during the SSA->normal phase. Single
5451 use/single def temporaries are replaced at their use location with their
5452 defining expression. This results in non-GIMPLE code, but gives the expanders
5453 much more complex trees to work on resulting in better RTL generation. This is
5454 enabled by default at @option{-O} and higher.
5456 @item -ftree-vectorize
5457 Perform loop vectorization on trees.
5459 @item -ftree-vect-loop-version
5460 @opindex ftree-vect-loop-version
5461 Perform loop versioning when doing loop vectorization on trees. When a loop
5462 appears to be vectorizable except that data alignment or data dependence cannot
5463 be determined at compile time then vectorized and non-vectorized versions of
5464 the loop are generated along with runtime checks for alignment or dependence
5465 to control which version is executed. This option is enabled by default
5466 except at level @option{-Os} where it is disabled.
5469 Perform Value Range Propagation on trees. This is similar to the
5470 constant propagation pass, but instead of values, ranges of values are
5471 propagated. This allows the optimizers to remove unnecessary range
5472 checks like array bound checks and null pointer checks. This is
5473 enabled by default at @option{-O2} and higher. Null pointer check
5474 elimination is only done if @option{-fdelete-null-pointer-checks} is
5479 Perform tail duplication to enlarge superblock size. This transformation
5480 simplifies the control flow of the function allowing other optimizations to do
5483 @item -funroll-loops
5484 @opindex funroll-loops
5485 Unroll loops whose number of iterations can be determined at compile
5486 time or upon entry to the loop. @option{-funroll-loops} implies
5487 @option{-frerun-cse-after-loop}. This option makes code larger,
5488 and may or may not make it run faster.
5490 @item -funroll-all-loops
5491 @opindex funroll-all-loops
5492 Unroll all loops, even if their number of iterations is uncertain when
5493 the loop is entered. This usually makes programs run more slowly.
5494 @option{-funroll-all-loops} implies the same options as
5495 @option{-funroll-loops},
5497 @item -fsplit-ivs-in-unroller
5498 @opindex -fsplit-ivs-in-unroller
5499 Enables expressing of values of induction variables in later iterations
5500 of the unrolled loop using the value in the first iteration. This breaks
5501 long dependency chains, thus improving efficiency of the scheduling passes.
5503 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5504 same effect. However in cases the loop body is more complicated than
5505 a single basic block, this is not reliable. It also does not work at all
5506 on some of the architectures due to restrictions in the CSE pass.
5508 This optimization is enabled by default.
5510 @item -fvariable-expansion-in-unroller
5511 @opindex -fvariable-expansion-in-unroller
5512 With this option, the compiler will create multiple copies of some
5513 local variables when unrolling a loop which can result in superior code.
5515 @item -fprefetch-loop-arrays
5516 @opindex fprefetch-loop-arrays
5517 If supported by the target machine, generate instructions to prefetch
5518 memory to improve the performance of loops that access large arrays.
5520 This option may generate better or worse code; results are highly
5521 dependent on the structure of loops within the source code.
5523 Disabled at level @option{-Os}.
5526 @itemx -fno-peephole2
5527 @opindex fno-peephole
5528 @opindex fno-peephole2
5529 Disable any machine-specific peephole optimizations. The difference
5530 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5531 are implemented in the compiler; some targets use one, some use the
5532 other, a few use both.
5534 @option{-fpeephole} is enabled by default.
5535 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5537 @item -fno-guess-branch-probability
5538 @opindex fno-guess-branch-probability
5539 Do not guess branch probabilities using heuristics.
5541 GCC will use heuristics to guess branch probabilities if they are
5542 not provided by profiling feedback (@option{-fprofile-arcs}). These
5543 heuristics are based on the control flow graph. If some branch probabilities
5544 are specified by @samp{__builtin_expect}, then the heuristics will be
5545 used to guess branch probabilities for the rest of the control flow graph,
5546 taking the @samp{__builtin_expect} info into account. The interactions
5547 between the heuristics and @samp{__builtin_expect} can be complex, and in
5548 some cases, it may be useful to disable the heuristics so that the effects
5549 of @samp{__builtin_expect} are easier to understand.
5551 The default is @option{-fguess-branch-probability} at levels
5552 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5554 @item -freorder-blocks
5555 @opindex freorder-blocks
5556 Reorder basic blocks in the compiled function in order to reduce number of
5557 taken branches and improve code locality.
5559 Enabled at levels @option{-O2}, @option{-O3}.
5561 @item -freorder-blocks-and-partition
5562 @opindex freorder-blocks-and-partition
5563 In addition to reordering basic blocks in the compiled function, in order
5564 to reduce number of taken branches, partitions hot and cold basic blocks
5565 into separate sections of the assembly and .o files, to improve
5566 paging and cache locality performance.
5568 This optimization is automatically turned off in the presence of
5569 exception handling, for linkonce sections, for functions with a user-defined
5570 section attribute and on any architecture that does not support named
5573 @item -freorder-functions
5574 @opindex freorder-functions
5575 Reorder functions in the object file in order to
5576 improve code locality. This is implemented by using special
5577 subsections @code{.text.hot} for most frequently executed functions and
5578 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5579 the linker so object file format must support named sections and linker must
5580 place them in a reasonable way.
5582 Also profile feedback must be available in to make this option effective. See
5583 @option{-fprofile-arcs} for details.
5585 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5587 @item -fstrict-aliasing
5588 @opindex fstrict-aliasing
5589 Allows the compiler to assume the strictest aliasing rules applicable to
5590 the language being compiled. For C (and C++), this activates
5591 optimizations based on the type of expressions. In particular, an
5592 object of one type is assumed never to reside at the same address as an
5593 object of a different type, unless the types are almost the same. For
5594 example, an @code{unsigned int} can alias an @code{int}, but not a
5595 @code{void*} or a @code{double}. A character type may alias any other
5598 Pay special attention to code like this:
5611 The practice of reading from a different union member than the one most
5612 recently written to (called ``type-punning'') is common. Even with
5613 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5614 is accessed through the union type. So, the code above will work as
5615 expected. However, this code might not:
5626 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5628 @item -fstrict-overflow
5629 @opindex fstrict-overflow
5630 Allow the compiler to assume strict signed overflow rules, depending
5631 on the language being compiled. For C (and C++) this means that
5632 overflow when doing arithmetic with signed numbers is undefined, which
5633 means that the compiler may assume that it will not happen. This
5634 permits various optimizations. For example, the compiler will assume
5635 that an expression like @code{i + 10 > i} will always be true for
5636 signed @code{i}. This assumption is only valid if signed overflow is
5637 undefined, as the expression is false if @code{i + 10} overflows when
5638 using twos complement arithmetic. When this option is in effect any
5639 attempt to determine whether an operation on signed numbers will
5640 overflow must be written carefully to not actually involve overflow.
5642 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5643 that signed overflow is fully defined: it wraps. When
5644 @option{-fwrapv} is used, there is no difference between
5645 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5646 @option{-fwrapv} certain types of overflow are permitted. For
5647 example, if the compiler gets an overflow when doing arithmetic on
5648 constants, the overflowed value can still be used with
5649 @option{-fwrapv}, but not otherwise.
5651 The @option{-fstrict-overflow} option is enabled at levels
5652 @option{-O2}, @option{-O3}, @option{-Os}.
5654 @item -falign-functions
5655 @itemx -falign-functions=@var{n}
5656 @opindex falign-functions
5657 Align the start of functions to the next power-of-two greater than
5658 @var{n}, skipping up to @var{n} bytes. For instance,
5659 @option{-falign-functions=32} aligns functions to the next 32-byte
5660 boundary, but @option{-falign-functions=24} would align to the next
5661 32-byte boundary only if this can be done by skipping 23 bytes or less.
5663 @option{-fno-align-functions} and @option{-falign-functions=1} are
5664 equivalent and mean that functions will not be aligned.
5666 Some assemblers only support this flag when @var{n} is a power of two;
5667 in that case, it is rounded up.
5669 If @var{n} is not specified or is zero, use a machine-dependent default.
5671 Enabled at levels @option{-O2}, @option{-O3}.
5673 @item -falign-labels
5674 @itemx -falign-labels=@var{n}
5675 @opindex falign-labels
5676 Align all branch targets to a power-of-two boundary, skipping up to
5677 @var{n} bytes like @option{-falign-functions}. This option can easily
5678 make code slower, because it must insert dummy operations for when the
5679 branch target is reached in the usual flow of the code.
5681 @option{-fno-align-labels} and @option{-falign-labels=1} are
5682 equivalent and mean that labels will not be aligned.
5684 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5685 are greater than this value, then their values are used instead.
5687 If @var{n} is not specified or is zero, use a machine-dependent default
5688 which is very likely to be @samp{1}, meaning no alignment.
5690 Enabled at levels @option{-O2}, @option{-O3}.
5693 @itemx -falign-loops=@var{n}
5694 @opindex falign-loops
5695 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5696 like @option{-falign-functions}. The hope is that the loop will be
5697 executed many times, which will make up for any execution of the dummy
5700 @option{-fno-align-loops} and @option{-falign-loops=1} are
5701 equivalent and mean that loops will not be aligned.
5703 If @var{n} is not specified or is zero, use a machine-dependent default.
5705 Enabled at levels @option{-O2}, @option{-O3}.
5708 @itemx -falign-jumps=@var{n}
5709 @opindex falign-jumps
5710 Align branch targets to a power-of-two boundary, for branch targets
5711 where the targets can only be reached by jumping, skipping up to @var{n}
5712 bytes like @option{-falign-functions}. In this case, no dummy operations
5715 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5716 equivalent and mean that loops will not be aligned.
5718 If @var{n} is not specified or is zero, use a machine-dependent default.
5720 Enabled at levels @option{-O2}, @option{-O3}.
5722 @item -funit-at-a-time
5723 @opindex funit-at-a-time
5724 Parse the whole compilation unit before starting to produce code.
5725 This allows some extra optimizations to take place but consumes
5726 more memory (in general). There are some compatibility issues
5727 with @emph{unit-at-a-time} mode:
5730 enabling @emph{unit-at-a-time} mode may change the order
5731 in which functions, variables, and top-level @code{asm} statements
5732 are emitted, and will likely break code relying on some particular
5733 ordering. The majority of such top-level @code{asm} statements,
5734 though, can be replaced by @code{section} attributes. The
5735 @option{fno-toplevel-reorder} option may be used to keep the ordering
5736 used in the input file, at the cost of some optimizations.
5739 @emph{unit-at-a-time} mode removes unreferenced static variables
5740 and functions. This may result in undefined references
5741 when an @code{asm} statement refers directly to variables or functions
5742 that are otherwise unused. In that case either the variable/function
5743 shall be listed as an operand of the @code{asm} statement operand or,
5744 in the case of top-level @code{asm} statements the attribute @code{used}
5745 shall be used on the declaration.
5748 Static functions now can use non-standard passing conventions that
5749 may break @code{asm} statements calling functions directly. Again,
5750 attribute @code{used} will prevent this behavior.
5753 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5754 but this scheme may not be supported by future releases of GCC@.
5756 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5758 @item -fno-toplevel-reorder
5759 Do not reorder top-level functions, variables, and @code{asm}
5760 statements. Output them in the same order that they appear in the
5761 input file. When this option is used, unreferenced static variables
5762 will not be removed. This option is intended to support existing code
5763 which relies on a particular ordering. For new code, it is better to
5768 Constructs webs as commonly used for register allocation purposes and assign
5769 each web individual pseudo register. This allows the register allocation pass
5770 to operate on pseudos directly, but also strengthens several other optimization
5771 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5772 however, make debugging impossible, since variables will no longer stay in a
5775 Enabled by default with @option{-funroll-loops}.
5777 @item -fwhole-program
5778 @opindex fwhole-program
5779 Assume that the current compilation unit represents whole program being
5780 compiled. All public functions and variables with the exception of @code{main}
5781 and those merged by attribute @code{externally_visible} become static functions
5782 and in a affect gets more aggressively optimized by interprocedural optimizers.
5783 While this option is equivalent to proper use of @code{static} keyword for
5784 programs consisting of single file, in combination with option
5785 @option{--combine} this flag can be used to compile most of smaller scale C
5786 programs since the functions and variables become local for the whole combined
5787 compilation unit, not for the single source file itself.
5790 @item -fno-cprop-registers
5791 @opindex fno-cprop-registers
5792 After register allocation and post-register allocation instruction splitting,
5793 we perform a copy-propagation pass to try to reduce scheduling dependencies
5794 and occasionally eliminate the copy.
5796 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5798 @item -fprofile-generate
5799 @opindex fprofile-generate
5801 Enable options usually used for instrumenting application to produce
5802 profile useful for later recompilation with profile feedback based
5803 optimization. You must use @option{-fprofile-generate} both when
5804 compiling and when linking your program.
5806 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5809 @opindex fprofile-use
5810 Enable profile feedback directed optimizations, and optimizations
5811 generally profitable only with profile feedback available.
5813 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5814 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5816 By default, GCC emits an error message if the feedback profiles do not
5817 match the source code. This error can be turned into a warning by using
5818 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
5822 The following options control compiler behavior regarding floating
5823 point arithmetic. These options trade off between speed and
5824 correctness. All must be specifically enabled.
5828 @opindex ffloat-store
5829 Do not store floating point variables in registers, and inhibit other
5830 options that might change whether a floating point value is taken from a
5833 @cindex floating point precision
5834 This option prevents undesirable excess precision on machines such as
5835 the 68000 where the floating registers (of the 68881) keep more
5836 precision than a @code{double} is supposed to have. Similarly for the
5837 x86 architecture. For most programs, the excess precision does only
5838 good, but a few programs rely on the precise definition of IEEE floating
5839 point. Use @option{-ffloat-store} for such programs, after modifying
5840 them to store all pertinent intermediate computations into variables.
5844 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5845 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5846 @option{-fno-rounding-math}, @option{-fno-signaling-nans},
5847 @option{-fno-signed-zeros} and @option{fcx-limited-range}.
5849 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5851 This option should never be turned on by any @option{-O} option since
5852 it can result in incorrect output for programs which depend on
5853 an exact implementation of IEEE or ISO rules/specifications for
5856 @item -fno-math-errno
5857 @opindex fno-math-errno
5858 Do not set ERRNO after calling math functions that are executed
5859 with a single instruction, e.g., sqrt. A program that relies on
5860 IEEE exceptions for math error handling may want to use this flag
5861 for speed while maintaining IEEE arithmetic compatibility.
5863 This option should never be turned on by any @option{-O} option since
5864 it can result in incorrect output for programs which depend on
5865 an exact implementation of IEEE or ISO rules/specifications for
5868 The default is @option{-fmath-errno}.
5870 On Darwin systems, the math library never sets @code{errno}. There is
5871 therefore no reason for the compiler to consider the possibility that
5872 it might, and @option{-fno-math-errno} is the default.
5874 @item -funsafe-math-optimizations
5875 @opindex funsafe-math-optimizations
5876 Allow optimizations for floating-point arithmetic that (a) assume
5877 that arguments and results are valid and (b) may violate IEEE or
5878 ANSI standards. When used at link-time, it may include libraries
5879 or startup files that change the default FPU control word or other
5880 similar optimizations.
5882 This option should never be turned on by any @option{-O} option since
5883 it can result in incorrect output for programs which depend on
5884 an exact implementation of IEEE or ISO rules/specifications for
5887 The default is @option{-fno-unsafe-math-optimizations}.
5889 @item -ffinite-math-only
5890 @opindex ffinite-math-only
5891 Allow optimizations for floating-point arithmetic that assume
5892 that arguments and results are not NaNs or +-Infs.
5894 This option should never be turned on by any @option{-O} option since
5895 it can result in incorrect output for programs which depend on
5896 an exact implementation of IEEE or ISO rules/specifications.
5898 The default is @option{-fno-finite-math-only}.
5900 @item -fno-signed-zeros
5901 @opindex fno-signed-zeros
5902 Allow optimizations for floating point arithmetic that ignore the
5903 signedness of zero. IEEE arithmetic specifies the behavior of
5904 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
5905 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
5906 This option implies that the sign of a zero result isn't significant.
5908 The default is @option{-fsigned-zeros}.
5910 @item -fno-trapping-math
5911 @opindex fno-trapping-math
5912 Compile code assuming that floating-point operations cannot generate
5913 user-visible traps. These traps include division by zero, overflow,
5914 underflow, inexact result and invalid operation. This option implies
5915 @option{-fno-signaling-nans}. Setting this option may allow faster
5916 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5918 This option should never be turned on by any @option{-O} option since
5919 it can result in incorrect output for programs which depend on
5920 an exact implementation of IEEE or ISO rules/specifications for
5923 The default is @option{-ftrapping-math}.
5925 @item -frounding-math
5926 @opindex frounding-math
5927 Disable transformations and optimizations that assume default floating
5928 point rounding behavior. This is round-to-zero for all floating point
5929 to integer conversions, and round-to-nearest for all other arithmetic
5930 truncations. This option should be specified for programs that change
5931 the FP rounding mode dynamically, or that may be executed with a
5932 non-default rounding mode. This option disables constant folding of
5933 floating point expressions at compile-time (which may be affected by
5934 rounding mode) and arithmetic transformations that are unsafe in the
5935 presence of sign-dependent rounding modes.
5937 The default is @option{-fno-rounding-math}.
5939 This option is experimental and does not currently guarantee to
5940 disable all GCC optimizations that are affected by rounding mode.
5941 Future versions of GCC may provide finer control of this setting
5942 using C99's @code{FENV_ACCESS} pragma. This command line option
5943 will be used to specify the default state for @code{FENV_ACCESS}.
5945 @item -frtl-abstract-sequences
5946 @opindex frtl-abstract-sequences
5947 It is a size optimization method. This option is to find identical
5948 sequences of code, which can be turned into pseudo-procedures and
5949 then replace all occurrences with calls to the newly created
5950 subroutine. It is kind of an opposite of @option{-finline-functions}.
5951 This optimization runs at RTL level.
5953 @item -fsignaling-nans
5954 @opindex fsignaling-nans
5955 Compile code assuming that IEEE signaling NaNs may generate user-visible
5956 traps during floating-point operations. Setting this option disables
5957 optimizations that may change the number of exceptions visible with
5958 signaling NaNs. This option implies @option{-ftrapping-math}.
5960 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5963 The default is @option{-fno-signaling-nans}.
5965 This option is experimental and does not currently guarantee to
5966 disable all GCC optimizations that affect signaling NaN behavior.
5968 @item -fsingle-precision-constant
5969 @opindex fsingle-precision-constant
5970 Treat floating point constant as single precision constant instead of
5971 implicitly converting it to double precision constant.
5973 @item -fcx-limited-range
5974 @itemx -fno-cx-limited-range
5975 @opindex fcx-limited-range
5976 @opindex fno-cx-limited-range
5977 When enabled, this option states that a range reduction step is not
5978 needed when performing complex division. The default is
5979 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5981 This option controls the default setting of the ISO C99
5982 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5987 The following options control optimizations that may improve
5988 performance, but are not enabled by any @option{-O} options. This
5989 section includes experimental options that may produce broken code.
5992 @item -fbranch-probabilities
5993 @opindex fbranch-probabilities
5994 After running a program compiled with @option{-fprofile-arcs}
5995 (@pxref{Debugging Options,, Options for Debugging Your Program or
5996 @command{gcc}}), you can compile it a second time using
5997 @option{-fbranch-probabilities}, to improve optimizations based on
5998 the number of times each branch was taken. When the program
5999 compiled with @option{-fprofile-arcs} exits it saves arc execution
6000 counts to a file called @file{@var{sourcename}.gcda} for each source
6001 file. The information in this data file is very dependent on the
6002 structure of the generated code, so you must use the same source code
6003 and the same optimization options for both compilations.
6005 With @option{-fbranch-probabilities}, GCC puts a
6006 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6007 These can be used to improve optimization. Currently, they are only
6008 used in one place: in @file{reorg.c}, instead of guessing which path a
6009 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6010 exactly determine which path is taken more often.
6012 @item -fprofile-values
6013 @opindex fprofile-values
6014 If combined with @option{-fprofile-arcs}, it adds code so that some
6015 data about values of expressions in the program is gathered.
6017 With @option{-fbranch-probabilities}, it reads back the data gathered
6018 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6019 notes to instructions for their later usage in optimizations.
6021 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6025 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6026 a code to gather information about values of expressions.
6028 With @option{-fbranch-probabilities}, it reads back the data gathered
6029 and actually performs the optimizations based on them.
6030 Currently the optimizations include specialization of division operation
6031 using the knowledge about the value of the denominator.
6033 @item -frename-registers
6034 @opindex frename-registers
6035 Attempt to avoid false dependencies in scheduled code by making use
6036 of registers left over after register allocation. This optimization
6037 will most benefit processors with lots of registers. Depending on the
6038 debug information format adopted by the target, however, it can
6039 make debugging impossible, since variables will no longer stay in
6040 a ``home register''.
6042 Enabled by default with @option{-funroll-loops}.
6046 Perform tail duplication to enlarge superblock size. This transformation
6047 simplifies the control flow of the function allowing other optimizations to do
6050 Enabled with @option{-fprofile-use}.
6052 @item -funroll-loops
6053 @opindex funroll-loops
6054 Unroll loops whose number of iterations can be determined at compile time or
6055 upon entry to the loop. @option{-funroll-loops} implies
6056 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6057 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6058 small constant number of iterations). This option makes code larger, and may
6059 or may not make it run faster.
6061 Enabled with @option{-fprofile-use}.
6063 @item -funroll-all-loops
6064 @opindex funroll-all-loops
6065 Unroll all loops, even if their number of iterations is uncertain when
6066 the loop is entered. This usually makes programs run more slowly.
6067 @option{-funroll-all-loops} implies the same options as
6068 @option{-funroll-loops}.
6071 @opindex fpeel-loops
6072 Peels the loops for that there is enough information that they do not
6073 roll much (from profile feedback). It also turns on complete loop peeling
6074 (i.e.@: complete removal of loops with small constant number of iterations).
6076 Enabled with @option{-fprofile-use}.
6078 @item -fmove-loop-invariants
6079 @opindex fmove-loop-invariants
6080 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6081 at level @option{-O1}
6083 @item -funswitch-loops
6084 @opindex funswitch-loops
6085 Move branches with loop invariant conditions out of the loop, with duplicates
6086 of the loop on both branches (modified according to result of the condition).
6088 @item -ffunction-sections
6089 @itemx -fdata-sections
6090 @opindex ffunction-sections
6091 @opindex fdata-sections
6092 Place each function or data item into its own section in the output
6093 file if the target supports arbitrary sections. The name of the
6094 function or the name of the data item determines the section's name
6097 Use these options on systems where the linker can perform optimizations
6098 to improve locality of reference in the instruction space. Most systems
6099 using the ELF object format and SPARC processors running Solaris 2 have
6100 linkers with such optimizations. AIX may have these optimizations in
6103 Only use these options when there are significant benefits from doing
6104 so. When you specify these options, the assembler and linker will
6105 create larger object and executable files and will also be slower.
6106 You will not be able to use @code{gprof} on all systems if you
6107 specify this option and you may have problems with debugging if
6108 you specify both this option and @option{-g}.
6110 @item -fbranch-target-load-optimize
6111 @opindex fbranch-target-load-optimize
6112 Perform branch target register load optimization before prologue / epilogue
6114 The use of target registers can typically be exposed only during reload,
6115 thus hoisting loads out of loops and doing inter-block scheduling needs
6116 a separate optimization pass.
6118 @item -fbranch-target-load-optimize2
6119 @opindex fbranch-target-load-optimize2
6120 Perform branch target register load optimization after prologue / epilogue
6123 @item -fbtr-bb-exclusive
6124 @opindex fbtr-bb-exclusive
6125 When performing branch target register load optimization, don't reuse
6126 branch target registers in within any basic block.
6128 @item -fstack-protector
6129 Emit extra code to check for buffer overflows, such as stack smashing
6130 attacks. This is done by adding a guard variable to functions with
6131 vulnerable objects. This includes functions that call alloca, and
6132 functions with buffers larger than 8 bytes. The guards are initialized
6133 when a function is entered and then checked when the function exits.
6134 If a guard check fails, an error message is printed and the program exits.
6136 @item -fstack-protector-all
6137 Like @option{-fstack-protector} except that all functions are protected.
6139 @item -fsection-anchors
6140 @opindex fsection-anchors
6141 Try to reduce the number of symbolic address calculations by using
6142 shared ``anchor'' symbols to address nearby objects. This transformation
6143 can help to reduce the number of GOT entries and GOT accesses on some
6146 For example, the implementation of the following function @code{foo}:
6150 int foo (void) @{ return a + b + c; @}
6153 would usually calculate the addresses of all three variables, but if you
6154 compile it with @option{-fsection-anchors}, it will access the variables
6155 from a common anchor point instead. The effect is similar to the
6156 following pseudocode (which isn't valid C):
6161 register int *xr = &x;
6162 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6166 Not all targets support this option.
6168 @item --param @var{name}=@var{value}
6170 In some places, GCC uses various constants to control the amount of
6171 optimization that is done. For example, GCC will not inline functions
6172 that contain more that a certain number of instructions. You can
6173 control some of these constants on the command-line using the
6174 @option{--param} option.
6176 The names of specific parameters, and the meaning of the values, are
6177 tied to the internals of the compiler, and are subject to change
6178 without notice in future releases.
6180 In each case, the @var{value} is an integer. The allowable choices for
6181 @var{name} are given in the following table:
6184 @item salias-max-implicit-fields
6185 The maximum number of fields in a variable without direct
6186 structure accesses for which structure aliasing will consider trying
6187 to track each field. The default is 5
6189 @item salias-max-array-elements
6190 The maximum number of elements an array can have and its elements
6191 still be tracked individually by structure aliasing. The default is 4
6193 @item sra-max-structure-size
6194 The maximum structure size, in bytes, at which the scalar replacement
6195 of aggregates (SRA) optimization will perform block copies. The
6196 default value, 0, implies that GCC will select the most appropriate
6199 @item sra-field-structure-ratio
6200 The threshold ratio (as a percentage) between instantiated fields and
6201 the complete structure size. We say that if the ratio of the number
6202 of bytes in instantiated fields to the number of bytes in the complete
6203 structure exceeds this parameter, then block copies are not used. The
6206 @item max-crossjump-edges
6207 The maximum number of incoming edges to consider for crossjumping.
6208 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6209 the number of edges incoming to each block. Increasing values mean
6210 more aggressive optimization, making the compile time increase with
6211 probably small improvement in executable size.
6213 @item min-crossjump-insns
6214 The minimum number of instructions which must be matched at the end
6215 of two blocks before crossjumping will be performed on them. This
6216 value is ignored in the case where all instructions in the block being
6217 crossjumped from are matched. The default value is 5.
6219 @item max-grow-copy-bb-insns
6220 The maximum code size expansion factor when copying basic blocks
6221 instead of jumping. The expansion is relative to a jump instruction.
6222 The default value is 8.
6224 @item max-goto-duplication-insns
6225 The maximum number of instructions to duplicate to a block that jumps
6226 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6227 passes, GCC factors computed gotos early in the compilation process,
6228 and unfactors them as late as possible. Only computed jumps at the
6229 end of a basic blocks with no more than max-goto-duplication-insns are
6230 unfactored. The default value is 8.
6232 @item max-delay-slot-insn-search
6233 The maximum number of instructions to consider when looking for an
6234 instruction to fill a delay slot. If more than this arbitrary number of
6235 instructions is searched, the time savings from filling the delay slot
6236 will be minimal so stop searching. Increasing values mean more
6237 aggressive optimization, making the compile time increase with probably
6238 small improvement in executable run time.
6240 @item max-delay-slot-live-search
6241 When trying to fill delay slots, the maximum number of instructions to
6242 consider when searching for a block with valid live register
6243 information. Increasing this arbitrarily chosen value means more
6244 aggressive optimization, increasing the compile time. This parameter
6245 should be removed when the delay slot code is rewritten to maintain the
6248 @item max-gcse-memory
6249 The approximate maximum amount of memory that will be allocated in
6250 order to perform the global common subexpression elimination
6251 optimization. If more memory than specified is required, the
6252 optimization will not be done.
6254 @item max-gcse-passes
6255 The maximum number of passes of GCSE to run. The default is 1.
6257 @item max-pending-list-length
6258 The maximum number of pending dependencies scheduling will allow
6259 before flushing the current state and starting over. Large functions
6260 with few branches or calls can create excessively large lists which
6261 needlessly consume memory and resources.
6263 @item max-inline-insns-single
6264 Several parameters control the tree inliner used in gcc.
6265 This number sets the maximum number of instructions (counted in GCC's
6266 internal representation) in a single function that the tree inliner
6267 will consider for inlining. This only affects functions declared
6268 inline and methods implemented in a class declaration (C++).
6269 The default value is 450.
6271 @item max-inline-insns-auto
6272 When you use @option{-finline-functions} (included in @option{-O3}),
6273 a lot of functions that would otherwise not be considered for inlining
6274 by the compiler will be investigated. To those functions, a different
6275 (more restrictive) limit compared to functions declared inline can
6277 The default value is 90.
6279 @item large-function-insns
6280 The limit specifying really large functions. For functions larger than this
6281 limit after inlining inlining is constrained by
6282 @option{--param large-function-growth}. This parameter is useful primarily
6283 to avoid extreme compilation time caused by non-linear algorithms used by the
6285 This parameter is ignored when @option{-funit-at-a-time} is not used.
6286 The default value is 2700.
6288 @item large-function-growth
6289 Specifies maximal growth of large function caused by inlining in percents.
6290 This parameter is ignored when @option{-funit-at-a-time} is not used.
6291 The default value is 100 which limits large function growth to 2.0 times
6294 @item large-unit-insns
6295 The limit specifying large translation unit. Growth caused by inlining of
6296 units larger than this limit is limited by @option{--param inline-unit-growth}.
6297 For small units this might be too tight (consider unit consisting of function A
6298 that is inline and B that just calls A three time. If B is small relative to
6299 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6300 large units consisting of small inlininable functions however the overall unit
6301 growth limit is needed to avoid exponential explosion of code size. Thus for
6302 smaller units, the size is increased to @option{--param large-unit-insns}
6303 before applying @option{--param inline-unit-growth}. The default is 10000
6305 @item inline-unit-growth
6306 Specifies maximal overall growth of the compilation unit caused by inlining.
6307 This parameter is ignored when @option{-funit-at-a-time} is not used.
6308 The default value is 30 which limits unit growth to 1.3 times the original
6311 @item large-stack-frame
6312 The limit specifying large stack frames. While inlining the algorithm is trying
6313 to not grow past this limit too much. Default value is 256 bytes.
6315 @item large-stack-frame-growth
6316 Specifies maximal growth of large stack frames caused by inlining in percents.
6317 The default value is 1000 which limits large stack frame growth to 11 times
6320 @item max-inline-insns-recursive
6321 @itemx max-inline-insns-recursive-auto
6322 Specifies maximum number of instructions out-of-line copy of self recursive inline
6323 function can grow into by performing recursive inlining.
6325 For functions declared inline @option{--param max-inline-insns-recursive} is
6326 taken into account. For function not declared inline, recursive inlining
6327 happens only when @option{-finline-functions} (included in @option{-O3}) is
6328 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6329 default value is 450.
6331 @item max-inline-recursive-depth
6332 @itemx max-inline-recursive-depth-auto
6333 Specifies maximum recursion depth used by the recursive inlining.
6335 For functions declared inline @option{--param max-inline-recursive-depth} is
6336 taken into account. For function not declared inline, recursive inlining
6337 happens only when @option{-finline-functions} (included in @option{-O3}) is
6338 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6339 default value is 450.
6341 @item min-inline-recursive-probability
6342 Recursive inlining is profitable only for function having deep recursion
6343 in average and can hurt for function having little recursion depth by
6344 increasing the prologue size or complexity of function body to other
6347 When profile feedback is available (see @option{-fprofile-generate}) the actual
6348 recursion depth can be guessed from probability that function will recurse via
6349 given call expression. This parameter limits inlining only to call expression
6350 whose probability exceeds given threshold (in percents). The default value is
6353 @item inline-call-cost
6354 Specify cost of call instruction relative to simple arithmetics operations
6355 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6356 functions and at the same time increases size of leaf function that is believed to
6357 reduce function size by being inlined. In effect it increases amount of
6358 inlining for code having large abstraction penalty (many functions that just
6359 pass the arguments to other functions) and decrease inlining for code with low
6360 abstraction penalty. The default value is 16.
6362 @item min-vect-loop-bound
6363 The minimum number of iterations under which a loop will not get vectorized
6364 when @option{-ftree-vectorize} is used. The number of iterations after
6365 vectorization needs to be greater than the value specified by this option
6366 to allow vectorization. The default value is 0.
6368 @item max-unrolled-insns
6369 The maximum number of instructions that a loop should have if that loop
6370 is unrolled, and if the loop is unrolled, it determines how many times
6371 the loop code is unrolled.
6373 @item max-average-unrolled-insns
6374 The maximum number of instructions biased by probabilities of their execution
6375 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6376 it determines how many times the loop code is unrolled.
6378 @item max-unroll-times
6379 The maximum number of unrollings of a single loop.
6381 @item max-peeled-insns
6382 The maximum number of instructions that a loop should have if that loop
6383 is peeled, and if the loop is peeled, it determines how many times
6384 the loop code is peeled.
6386 @item max-peel-times
6387 The maximum number of peelings of a single loop.
6389 @item max-completely-peeled-insns
6390 The maximum number of insns of a completely peeled loop.
6392 @item max-completely-peel-times
6393 The maximum number of iterations of a loop to be suitable for complete peeling.
6395 @item max-unswitch-insns
6396 The maximum number of insns of an unswitched loop.
6398 @item max-unswitch-level
6399 The maximum number of branches unswitched in a single loop.
6402 The minimum cost of an expensive expression in the loop invariant motion.
6404 @item iv-consider-all-candidates-bound
6405 Bound on number of candidates for induction variables below that
6406 all candidates are considered for each use in induction variable
6407 optimizations. Only the most relevant candidates are considered
6408 if there are more candidates, to avoid quadratic time complexity.
6410 @item iv-max-considered-uses
6411 The induction variable optimizations give up on loops that contain more
6412 induction variable uses.
6414 @item iv-always-prune-cand-set-bound
6415 If number of candidates in the set is smaller than this value,
6416 we always try to remove unnecessary ivs from the set during its
6417 optimization when a new iv is added to the set.
6419 @item scev-max-expr-size
6420 Bound on size of expressions used in the scalar evolutions analyzer.
6421 Large expressions slow the analyzer.
6423 @item vect-max-version-checks
6424 The maximum number of runtime checks that can be performed when doing
6425 loop versioning in the vectorizer. See option ftree-vect-loop-version
6426 for more information.
6428 @item max-iterations-to-track
6430 The maximum number of iterations of a loop the brute force algorithm
6431 for analysis of # of iterations of the loop tries to evaluate.
6433 @item hot-bb-count-fraction
6434 Select fraction of the maximal count of repetitions of basic block in program
6435 given basic block needs to have to be considered hot.
6437 @item hot-bb-frequency-fraction
6438 Select fraction of the maximal frequency of executions of basic block in
6439 function given basic block needs to have to be considered hot
6441 @item max-predicted-iterations
6442 The maximum number of loop iterations we predict statically. This is useful
6443 in cases where function contain single loop with known bound and other loop
6444 with unknown. We predict the known number of iterations correctly, while
6445 the unknown number of iterations average to roughly 10. This means that the
6446 loop without bounds would appear artificially cold relative to the other one.
6448 @item tracer-dynamic-coverage
6449 @itemx tracer-dynamic-coverage-feedback
6451 This value is used to limit superblock formation once the given percentage of
6452 executed instructions is covered. This limits unnecessary code size
6455 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6456 feedback is available. The real profiles (as opposed to statically estimated
6457 ones) are much less balanced allowing the threshold to be larger value.
6459 @item tracer-max-code-growth
6460 Stop tail duplication once code growth has reached given percentage. This is
6461 rather hokey argument, as most of the duplicates will be eliminated later in
6462 cross jumping, so it may be set to much higher values than is the desired code
6465 @item tracer-min-branch-ratio
6467 Stop reverse growth when the reverse probability of best edge is less than this
6468 threshold (in percent).
6470 @item tracer-min-branch-ratio
6471 @itemx tracer-min-branch-ratio-feedback
6473 Stop forward growth if the best edge do have probability lower than this
6476 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6477 compilation for profile feedback and one for compilation without. The value
6478 for compilation with profile feedback needs to be more conservative (higher) in
6479 order to make tracer effective.
6481 @item max-cse-path-length
6483 Maximum number of basic blocks on path that cse considers. The default is 10.
6486 The maximum instructions CSE process before flushing. The default is 1000.
6488 @item max-aliased-vops
6490 Maximum number of virtual operands per statement allowed to represent
6491 aliases before triggering the alias grouping heuristic. Alias
6492 grouping reduces compile times and memory consumption needed for
6493 aliasing at the expense of precision loss in alias information.
6495 @item ggc-min-expand
6497 GCC uses a garbage collector to manage its own memory allocation. This
6498 parameter specifies the minimum percentage by which the garbage
6499 collector's heap should be allowed to expand between collections.
6500 Tuning this may improve compilation speed; it has no effect on code
6503 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6504 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6505 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6506 GCC is not able to calculate RAM on a particular platform, the lower
6507 bound of 30% is used. Setting this parameter and
6508 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6509 every opportunity. This is extremely slow, but can be useful for
6512 @item ggc-min-heapsize
6514 Minimum size of the garbage collector's heap before it begins bothering
6515 to collect garbage. The first collection occurs after the heap expands
6516 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6517 tuning this may improve compilation speed, and has no effect on code
6520 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6521 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6522 with a lower bound of 4096 (four megabytes) and an upper bound of
6523 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6524 particular platform, the lower bound is used. Setting this parameter
6525 very large effectively disables garbage collection. Setting this
6526 parameter and @option{ggc-min-expand} to zero causes a full collection
6527 to occur at every opportunity.
6529 @item max-reload-search-insns
6530 The maximum number of instruction reload should look backward for equivalent
6531 register. Increasing values mean more aggressive optimization, making the
6532 compile time increase with probably slightly better performance. The default
6535 @item max-cselib-memory-locations
6536 The maximum number of memory locations cselib should take into account.
6537 Increasing values mean more aggressive optimization, making the compile time
6538 increase with probably slightly better performance. The default value is 500.
6540 @item max-flow-memory-locations
6541 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6542 The default value is 100.
6544 @item reorder-blocks-duplicate
6545 @itemx reorder-blocks-duplicate-feedback
6547 Used by basic block reordering pass to decide whether to use unconditional
6548 branch or duplicate the code on its destination. Code is duplicated when its
6549 estimated size is smaller than this value multiplied by the estimated size of
6550 unconditional jump in the hot spots of the program.
6552 The @option{reorder-block-duplicate-feedback} is used only when profile
6553 feedback is available and may be set to higher values than
6554 @option{reorder-block-duplicate} since information about the hot spots is more
6557 @item max-sched-ready-insns
6558 The maximum number of instructions ready to be issued the scheduler should
6559 consider at any given time during the first scheduling pass. Increasing
6560 values mean more thorough searches, making the compilation time increase
6561 with probably little benefit. The default value is 100.
6563 @item max-sched-region-blocks
6564 The maximum number of blocks in a region to be considered for
6565 interblock scheduling. The default value is 10.
6567 @item max-sched-region-insns
6568 The maximum number of insns in a region to be considered for
6569 interblock scheduling. The default value is 100.
6572 The minimum probability (in percents) of reaching a source block
6573 for interblock speculative scheduling. The default value is 40.
6575 @item max-sched-extend-regions-iters
6576 The maximum number of iterations through CFG to extend regions.
6577 0 - disable region extension,
6578 N - do at most N iterations.
6579 The default value is 0.
6581 @item max-sched-insn-conflict-delay
6582 The maximum conflict delay for an insn to be considered for speculative motion.
6583 The default value is 3.
6585 @item sched-spec-prob-cutoff
6586 The minimal probability of speculation success (in percents), so that
6587 speculative insn will be scheduled.
6588 The default value is 40.
6590 @item max-last-value-rtl
6592 The maximum size measured as number of RTLs that can be recorded in an expression
6593 in combiner for a pseudo register as last known value of that register. The default
6596 @item integer-share-limit
6597 Small integer constants can use a shared data structure, reducing the
6598 compiler's memory usage and increasing its speed. This sets the maximum
6599 value of a shared integer constant's. The default value is 256.
6601 @item min-virtual-mappings
6602 Specifies the minimum number of virtual mappings in the incremental
6603 SSA updater that should be registered to trigger the virtual mappings
6604 heuristic defined by virtual-mappings-ratio. The default value is
6607 @item virtual-mappings-ratio
6608 If the number of virtual mappings is virtual-mappings-ratio bigger
6609 than the number of virtual symbols to be updated, then the incremental
6610 SSA updater switches to a full update for those symbols. The default
6613 @item ssp-buffer-size
6614 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6615 protection when @option{-fstack-protection} is used.
6617 @item max-jump-thread-duplication-stmts
6618 Maximum number of statements allowed in a block that needs to be
6619 duplicated when threading jumps.
6621 @item max-fields-for-field-sensitive
6622 Maximum number of fields in a structure we will treat in
6623 a field sensitive manner during pointer analysis.
6625 @item prefetch-latency
6626 Estimate on average number of instructions that are executed before
6627 prefetch finishes. The distance we prefetch ahead is proportional
6628 to this constant. Increasing this number may also lead to less
6629 streams being prefetched (see @option{simultaneous-prefetches}).
6631 @item simultaneous-prefetches
6632 Maximum number of prefetches that can run at the same time.
6634 @item l1-cache-line-size
6635 The size of cache line in L1 cache, in bytes.
6638 The number of cache lines in L1 cache.
6640 @item verify-canonical-types
6641 Whether the compiler should verify the ``canonical'' types used for
6642 type equality comparisons within the C++ and Objective-C++ front
6643 ends. Set to 1 (the default when GCC is configured with
6644 --enable-checking) to enable verification, 0 to disable verification
6645 (the default when GCC is configured with --disable-checking).
6650 @node Preprocessor Options
6651 @section Options Controlling the Preprocessor
6652 @cindex preprocessor options
6653 @cindex options, preprocessor
6655 These options control the C preprocessor, which is run on each C source
6656 file before actual compilation.
6658 If you use the @option{-E} option, nothing is done except preprocessing.
6659 Some of these options make sense only together with @option{-E} because
6660 they cause the preprocessor output to be unsuitable for actual
6665 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6666 and pass @var{option} directly through to the preprocessor. If
6667 @var{option} contains commas, it is split into multiple options at the
6668 commas. However, many options are modified, translated or interpreted
6669 by the compiler driver before being passed to the preprocessor, and
6670 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6671 interface is undocumented and subject to change, so whenever possible
6672 you should avoid using @option{-Wp} and let the driver handle the
6675 @item -Xpreprocessor @var{option}
6676 @opindex preprocessor
6677 Pass @var{option} as an option to the preprocessor. You can use this to
6678 supply system-specific preprocessor options which GCC does not know how to
6681 If you want to pass an option that takes an argument, you must use
6682 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6685 @include cppopts.texi
6687 @node Assembler Options
6688 @section Passing Options to the Assembler
6690 @c prevent bad page break with this line
6691 You can pass options to the assembler.
6694 @item -Wa,@var{option}
6696 Pass @var{option} as an option to the assembler. If @var{option}
6697 contains commas, it is split into multiple options at the commas.
6699 @item -Xassembler @var{option}
6701 Pass @var{option} as an option to the assembler. You can use this to
6702 supply system-specific assembler options which GCC does not know how to
6705 If you want to pass an option that takes an argument, you must use
6706 @option{-Xassembler} twice, once for the option and once for the argument.
6711 @section Options for Linking
6712 @cindex link options
6713 @cindex options, linking
6715 These options come into play when the compiler links object files into
6716 an executable output file. They are meaningless if the compiler is
6717 not doing a link step.
6721 @item @var{object-file-name}
6722 A file name that does not end in a special recognized suffix is
6723 considered to name an object file or library. (Object files are
6724 distinguished from libraries by the linker according to the file
6725 contents.) If linking is done, these object files are used as input
6734 If any of these options is used, then the linker is not run, and
6735 object file names should not be used as arguments. @xref{Overall
6739 @item -l@var{library}
6740 @itemx -l @var{library}
6742 Search the library named @var{library} when linking. (The second
6743 alternative with the library as a separate argument is only for
6744 POSIX compliance and is not recommended.)
6746 It makes a difference where in the command you write this option; the
6747 linker searches and processes libraries and object files in the order they
6748 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6749 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6750 to functions in @samp{z}, those functions may not be loaded.
6752 The linker searches a standard list of directories for the library,
6753 which is actually a file named @file{lib@var{library}.a}. The linker
6754 then uses this file as if it had been specified precisely by name.
6756 The directories searched include several standard system directories
6757 plus any that you specify with @option{-L}.
6759 Normally the files found this way are library files---archive files
6760 whose members are object files. The linker handles an archive file by
6761 scanning through it for members which define symbols that have so far
6762 been referenced but not defined. But if the file that is found is an
6763 ordinary object file, it is linked in the usual fashion. The only
6764 difference between using an @option{-l} option and specifying a file name
6765 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6766 and searches several directories.
6770 You need this special case of the @option{-l} option in order to
6771 link an Objective-C or Objective-C++ program.
6774 @opindex nostartfiles
6775 Do not use the standard system startup files when linking.
6776 The standard system libraries are used normally, unless @option{-nostdlib}
6777 or @option{-nodefaultlibs} is used.
6779 @item -nodefaultlibs
6780 @opindex nodefaultlibs
6781 Do not use the standard system libraries when linking.
6782 Only the libraries you specify will be passed to the linker.
6783 The standard startup files are used normally, unless @option{-nostartfiles}
6784 is used. The compiler may generate calls to @code{memcmp},
6785 @code{memset}, @code{memcpy} and @code{memmove}.
6786 These entries are usually resolved by entries in
6787 libc. These entry points should be supplied through some other
6788 mechanism when this option is specified.
6792 Do not use the standard system startup files or libraries when linking.
6793 No startup files and only the libraries you specify will be passed to
6794 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6795 @code{memcpy} and @code{memmove}.
6796 These entries are usually resolved by entries in
6797 libc. These entry points should be supplied through some other
6798 mechanism when this option is specified.
6800 @cindex @option{-lgcc}, use with @option{-nostdlib}
6801 @cindex @option{-nostdlib} and unresolved references
6802 @cindex unresolved references and @option{-nostdlib}
6803 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6804 @cindex @option{-nodefaultlibs} and unresolved references
6805 @cindex unresolved references and @option{-nodefaultlibs}
6806 One of the standard libraries bypassed by @option{-nostdlib} and
6807 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6808 that GCC uses to overcome shortcomings of particular machines, or special
6809 needs for some languages.
6810 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6811 Collection (GCC) Internals},
6812 for more discussion of @file{libgcc.a}.)
6813 In most cases, you need @file{libgcc.a} even when you want to avoid
6814 other standard libraries. In other words, when you specify @option{-nostdlib}
6815 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6816 This ensures that you have no unresolved references to internal GCC
6817 library subroutines. (For example, @samp{__main}, used to ensure C++
6818 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6819 GNU Compiler Collection (GCC) Internals}.)
6823 Produce a position independent executable on targets which support it.
6824 For predictable results, you must also specify the same set of options
6825 that were used to generate code (@option{-fpie}, @option{-fPIE},
6826 or model suboptions) when you specify this option.
6830 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6831 that support it. This instructs the linker to add all symbols, not
6832 only used ones, to the dynamic symbol table. This option is needed
6833 for some uses of @code{dlopen} or to allow obtaining backtraces
6834 from within a program.
6838 Remove all symbol table and relocation information from the executable.
6842 On systems that support dynamic linking, this prevents linking with the shared
6843 libraries. On other systems, this option has no effect.
6847 Produce a shared object which can then be linked with other objects to
6848 form an executable. Not all systems support this option. For predictable
6849 results, you must also specify the same set of options that were used to
6850 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6851 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6852 needs to build supplementary stub code for constructors to work. On
6853 multi-libbed systems, @samp{gcc -shared} must select the correct support
6854 libraries to link against. Failing to supply the correct flags may lead
6855 to subtle defects. Supplying them in cases where they are not necessary
6858 @item -shared-libgcc
6859 @itemx -static-libgcc
6860 @opindex shared-libgcc
6861 @opindex static-libgcc
6862 On systems that provide @file{libgcc} as a shared library, these options
6863 force the use of either the shared or static version respectively.
6864 If no shared version of @file{libgcc} was built when the compiler was
6865 configured, these options have no effect.
6867 There are several situations in which an application should use the
6868 shared @file{libgcc} instead of the static version. The most common
6869 of these is when the application wishes to throw and catch exceptions
6870 across different shared libraries. In that case, each of the libraries
6871 as well as the application itself should use the shared @file{libgcc}.
6873 Therefore, the G++ and GCJ drivers automatically add
6874 @option{-shared-libgcc} whenever you build a shared library or a main
6875 executable, because C++ and Java programs typically use exceptions, so
6876 this is the right thing to do.
6878 If, instead, you use the GCC driver to create shared libraries, you may
6879 find that they will not always be linked with the shared @file{libgcc}.
6880 If GCC finds, at its configuration time, that you have a non-GNU linker
6881 or a GNU linker that does not support option @option{--eh-frame-hdr},
6882 it will link the shared version of @file{libgcc} into shared libraries
6883 by default. Otherwise, it will take advantage of the linker and optimize
6884 away the linking with the shared version of @file{libgcc}, linking with
6885 the static version of libgcc by default. This allows exceptions to
6886 propagate through such shared libraries, without incurring relocation
6887 costs at library load time.
6889 However, if a library or main executable is supposed to throw or catch
6890 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6891 for the languages used in the program, or using the option
6892 @option{-shared-libgcc}, such that it is linked with the shared
6897 Bind references to global symbols when building a shared object. Warn
6898 about any unresolved references (unless overridden by the link editor
6899 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6902 @item -Xlinker @var{option}
6904 Pass @var{option} as an option to the linker. You can use this to
6905 supply system-specific linker options which GCC does not know how to
6908 If you want to pass an option that takes an argument, you must use
6909 @option{-Xlinker} twice, once for the option and once for the argument.
6910 For example, to pass @option{-assert definitions}, you must write
6911 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6912 @option{-Xlinker "-assert definitions"}, because this passes the entire
6913 string as a single argument, which is not what the linker expects.
6915 @item -Wl,@var{option}
6917 Pass @var{option} as an option to the linker. If @var{option} contains
6918 commas, it is split into multiple options at the commas.
6920 @item -u @var{symbol}
6922 Pretend the symbol @var{symbol} is undefined, to force linking of
6923 library modules to define it. You can use @option{-u} multiple times with
6924 different symbols to force loading of additional library modules.
6927 @node Directory Options
6928 @section Options for Directory Search
6929 @cindex directory options
6930 @cindex options, directory search
6933 These options specify directories to search for header files, for
6934 libraries and for parts of the compiler:
6939 Add the directory @var{dir} to the head of the list of directories to be
6940 searched for header files. This can be used to override a system header
6941 file, substituting your own version, since these directories are
6942 searched before the system header file directories. However, you should
6943 not use this option to add directories that contain vendor-supplied
6944 system header files (use @option{-isystem} for that). If you use more than
6945 one @option{-I} option, the directories are scanned in left-to-right
6946 order; the standard system directories come after.
6948 If a standard system include directory, or a directory specified with
6949 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6950 option will be ignored. The directory will still be searched but as a
6951 system directory at its normal position in the system include chain.
6952 This is to ensure that GCC's procedure to fix buggy system headers and
6953 the ordering for the include_next directive are not inadvertently changed.
6954 If you really need to change the search order for system directories,
6955 use the @option{-nostdinc} and/or @option{-isystem} options.
6957 @item -iquote@var{dir}
6959 Add the directory @var{dir} to the head of the list of directories to
6960 be searched for header files only for the case of @samp{#include
6961 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6962 otherwise just like @option{-I}.
6966 Add directory @var{dir} to the list of directories to be searched
6969 @item -B@var{prefix}
6971 This option specifies where to find the executables, libraries,
6972 include files, and data files of the compiler itself.
6974 The compiler driver program runs one or more of the subprograms
6975 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6976 @var{prefix} as a prefix for each program it tries to run, both with and
6977 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6979 For each subprogram to be run, the compiler driver first tries the
6980 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6981 was not specified, the driver tries two standard prefixes, which are
6982 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6983 those results in a file name that is found, the unmodified program
6984 name is searched for using the directories specified in your
6985 @env{PATH} environment variable.
6987 The compiler will check to see if the path provided by the @option{-B}
6988 refers to a directory, and if necessary it will add a directory
6989 separator character at the end of the path.
6991 @option{-B} prefixes that effectively specify directory names also apply
6992 to libraries in the linker, because the compiler translates these
6993 options into @option{-L} options for the linker. They also apply to
6994 includes files in the preprocessor, because the compiler translates these
6995 options into @option{-isystem} options for the preprocessor. In this case,
6996 the compiler appends @samp{include} to the prefix.
6998 The run-time support file @file{libgcc.a} can also be searched for using
6999 the @option{-B} prefix, if needed. If it is not found there, the two
7000 standard prefixes above are tried, and that is all. The file is left
7001 out of the link if it is not found by those means.
7003 Another way to specify a prefix much like the @option{-B} prefix is to use
7004 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7007 As a special kludge, if the path provided by @option{-B} is
7008 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7009 9, then it will be replaced by @file{[dir/]include}. This is to help
7010 with boot-strapping the compiler.
7012 @item -specs=@var{file}
7014 Process @var{file} after the compiler reads in the standard @file{specs}
7015 file, in order to override the defaults that the @file{gcc} driver
7016 program uses when determining what switches to pass to @file{cc1},
7017 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7018 @option{-specs=@var{file}} can be specified on the command line, and they
7019 are processed in order, from left to right.
7021 @item --sysroot=@var{dir}
7023 Use @var{dir} as the logical root directory for headers and libraries.
7024 For example, if the compiler would normally search for headers in
7025 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7026 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7028 If you use both this option and the @option{-isysroot} option, then
7029 the @option{--sysroot} option will apply to libraries, but the
7030 @option{-isysroot} option will apply to header files.
7032 The GNU linker (beginning with version 2.16) has the necessary support
7033 for this option. If your linker does not support this option, the
7034 header file aspect of @option{--sysroot} will still work, but the
7035 library aspect will not.
7039 This option has been deprecated. Please use @option{-iquote} instead for
7040 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7041 Any directories you specify with @option{-I} options before the @option{-I-}
7042 option are searched only for the case of @samp{#include "@var{file}"};
7043 they are not searched for @samp{#include <@var{file}>}.
7045 If additional directories are specified with @option{-I} options after
7046 the @option{-I-}, these directories are searched for all @samp{#include}
7047 directives. (Ordinarily @emph{all} @option{-I} directories are used
7050 In addition, the @option{-I-} option inhibits the use of the current
7051 directory (where the current input file came from) as the first search
7052 directory for @samp{#include "@var{file}"}. There is no way to
7053 override this effect of @option{-I-}. With @option{-I.} you can specify
7054 searching the directory which was current when the compiler was
7055 invoked. That is not exactly the same as what the preprocessor does
7056 by default, but it is often satisfactory.
7058 @option{-I-} does not inhibit the use of the standard system directories
7059 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7066 @section Specifying subprocesses and the switches to pass to them
7069 @command{gcc} is a driver program. It performs its job by invoking a
7070 sequence of other programs to do the work of compiling, assembling and
7071 linking. GCC interprets its command-line parameters and uses these to
7072 deduce which programs it should invoke, and which command-line options
7073 it ought to place on their command lines. This behavior is controlled
7074 by @dfn{spec strings}. In most cases there is one spec string for each
7075 program that GCC can invoke, but a few programs have multiple spec
7076 strings to control their behavior. The spec strings built into GCC can
7077 be overridden by using the @option{-specs=} command-line switch to specify
7080 @dfn{Spec files} are plaintext files that are used to construct spec
7081 strings. They consist of a sequence of directives separated by blank
7082 lines. The type of directive is determined by the first non-whitespace
7083 character on the line and it can be one of the following:
7086 @item %@var{command}
7087 Issues a @var{command} to the spec file processor. The commands that can
7091 @item %include <@var{file}>
7093 Search for @var{file} and insert its text at the current point in the
7096 @item %include_noerr <@var{file}>
7097 @cindex %include_noerr
7098 Just like @samp{%include}, but do not generate an error message if the include
7099 file cannot be found.
7101 @item %rename @var{old_name} @var{new_name}
7103 Rename the spec string @var{old_name} to @var{new_name}.
7107 @item *[@var{spec_name}]:
7108 This tells the compiler to create, override or delete the named spec
7109 string. All lines after this directive up to the next directive or
7110 blank line are considered to be the text for the spec string. If this
7111 results in an empty string then the spec will be deleted. (Or, if the
7112 spec did not exist, then nothing will happened.) Otherwise, if the spec
7113 does not currently exist a new spec will be created. If the spec does
7114 exist then its contents will be overridden by the text of this
7115 directive, unless the first character of that text is the @samp{+}
7116 character, in which case the text will be appended to the spec.
7118 @item [@var{suffix}]:
7119 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7120 and up to the next directive or blank line are considered to make up the
7121 spec string for the indicated suffix. When the compiler encounters an
7122 input file with the named suffix, it will processes the spec string in
7123 order to work out how to compile that file. For example:
7130 This says that any input file whose name ends in @samp{.ZZ} should be
7131 passed to the program @samp{z-compile}, which should be invoked with the
7132 command-line switch @option{-input} and with the result of performing the
7133 @samp{%i} substitution. (See below.)
7135 As an alternative to providing a spec string, the text that follows a
7136 suffix directive can be one of the following:
7139 @item @@@var{language}
7140 This says that the suffix is an alias for a known @var{language}. This is
7141 similar to using the @option{-x} command-line switch to GCC to specify a
7142 language explicitly. For example:
7149 Says that .ZZ files are, in fact, C++ source files.
7152 This causes an error messages saying:
7155 @var{name} compiler not installed on this system.
7159 GCC already has an extensive list of suffixes built into it.
7160 This directive will add an entry to the end of the list of suffixes, but
7161 since the list is searched from the end backwards, it is effectively
7162 possible to override earlier entries using this technique.
7166 GCC has the following spec strings built into it. Spec files can
7167 override these strings or create their own. Note that individual
7168 targets can also add their own spec strings to this list.
7171 asm Options to pass to the assembler
7172 asm_final Options to pass to the assembler post-processor
7173 cpp Options to pass to the C preprocessor
7174 cc1 Options to pass to the C compiler
7175 cc1plus Options to pass to the C++ compiler
7176 endfile Object files to include at the end of the link
7177 link Options to pass to the linker
7178 lib Libraries to include on the command line to the linker
7179 libgcc Decides which GCC support library to pass to the linker
7180 linker Sets the name of the linker
7181 predefines Defines to be passed to the C preprocessor
7182 signed_char Defines to pass to CPP to say whether @code{char} is signed
7184 startfile Object files to include at the start of the link
7187 Here is a small example of a spec file:
7193 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7196 This example renames the spec called @samp{lib} to @samp{old_lib} and
7197 then overrides the previous definition of @samp{lib} with a new one.
7198 The new definition adds in some extra command-line options before
7199 including the text of the old definition.
7201 @dfn{Spec strings} are a list of command-line options to be passed to their
7202 corresponding program. In addition, the spec strings can contain
7203 @samp{%}-prefixed sequences to substitute variable text or to
7204 conditionally insert text into the command line. Using these constructs
7205 it is possible to generate quite complex command lines.
7207 Here is a table of all defined @samp{%}-sequences for spec
7208 strings. Note that spaces are not generated automatically around the
7209 results of expanding these sequences. Therefore you can concatenate them
7210 together or combine them with constant text in a single argument.
7214 Substitute one @samp{%} into the program name or argument.
7217 Substitute the name of the input file being processed.
7220 Substitute the basename of the input file being processed.
7221 This is the substring up to (and not including) the last period
7222 and not including the directory.
7225 This is the same as @samp{%b}, but include the file suffix (text after
7229 Marks the argument containing or following the @samp{%d} as a
7230 temporary file name, so that that file will be deleted if GCC exits
7231 successfully. Unlike @samp{%g}, this contributes no text to the
7234 @item %g@var{suffix}
7235 Substitute a file name that has suffix @var{suffix} and is chosen
7236 once per compilation, and mark the argument in the same way as
7237 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7238 name is now chosen in a way that is hard to predict even when previously
7239 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7240 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7241 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7242 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7243 was simply substituted with a file name chosen once per compilation,
7244 without regard to any appended suffix (which was therefore treated
7245 just like ordinary text), making such attacks more likely to succeed.
7247 @item %u@var{suffix}
7248 Like @samp{%g}, but generates a new temporary file name even if
7249 @samp{%u@var{suffix}} was already seen.
7251 @item %U@var{suffix}
7252 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7253 new one if there is no such last file name. In the absence of any
7254 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7255 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7256 would involve the generation of two distinct file names, one
7257 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7258 simply substituted with a file name chosen for the previous @samp{%u},
7259 without regard to any appended suffix.
7261 @item %j@var{suffix}
7262 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7263 writable, and if save-temps is off; otherwise, substitute the name
7264 of a temporary file, just like @samp{%u}. This temporary file is not
7265 meant for communication between processes, but rather as a junk
7268 @item %|@var{suffix}
7269 @itemx %m@var{suffix}
7270 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7271 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7272 all. These are the two most common ways to instruct a program that it
7273 should read from standard input or write to standard output. If you
7274 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7275 construct: see for example @file{f/lang-specs.h}.
7277 @item %.@var{SUFFIX}
7278 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7279 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7280 terminated by the next space or %.
7283 Marks the argument containing or following the @samp{%w} as the
7284 designated output file of this compilation. This puts the argument
7285 into the sequence of arguments that @samp{%o} will substitute later.
7288 Substitutes the names of all the output files, with spaces
7289 automatically placed around them. You should write spaces
7290 around the @samp{%o} as well or the results are undefined.
7291 @samp{%o} is for use in the specs for running the linker.
7292 Input files whose names have no recognized suffix are not compiled
7293 at all, but they are included among the output files, so they will
7297 Substitutes the suffix for object files. Note that this is
7298 handled specially when it immediately follows @samp{%g, %u, or %U},
7299 because of the need for those to form complete file names. The
7300 handling is such that @samp{%O} is treated exactly as if it had already
7301 been substituted, except that @samp{%g, %u, and %U} do not currently
7302 support additional @var{suffix} characters following @samp{%O} as they would
7303 following, for example, @samp{.o}.
7306 Substitutes the standard macro predefinitions for the
7307 current target machine. Use this when running @code{cpp}.
7310 Like @samp{%p}, but puts @samp{__} before and after the name of each
7311 predefined macro, except for macros that start with @samp{__} or with
7312 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7316 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7317 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7318 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7319 and @option{-imultilib} as necessary.
7322 Current argument is the name of a library or startup file of some sort.
7323 Search for that file in a standard list of directories and substitute
7324 the full name found.
7327 Print @var{str} as an error message. @var{str} is terminated by a newline.
7328 Use this when inconsistent options are detected.
7331 Substitute the contents of spec string @var{name} at this point.
7334 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7336 @item %x@{@var{option}@}
7337 Accumulate an option for @samp{%X}.
7340 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7344 Output the accumulated assembler options specified by @option{-Wa}.
7347 Output the accumulated preprocessor options specified by @option{-Wp}.
7350 Process the @code{asm} spec. This is used to compute the
7351 switches to be passed to the assembler.
7354 Process the @code{asm_final} spec. This is a spec string for
7355 passing switches to an assembler post-processor, if such a program is
7359 Process the @code{link} spec. This is the spec for computing the
7360 command line passed to the linker. Typically it will make use of the
7361 @samp{%L %G %S %D and %E} sequences.
7364 Dump out a @option{-L} option for each directory that GCC believes might
7365 contain startup files. If the target supports multilibs then the
7366 current multilib directory will be prepended to each of these paths.
7369 Process the @code{lib} spec. This is a spec string for deciding which
7370 libraries should be included on the command line to the linker.
7373 Process the @code{libgcc} spec. This is a spec string for deciding
7374 which GCC support library should be included on the command line to the linker.
7377 Process the @code{startfile} spec. This is a spec for deciding which
7378 object files should be the first ones passed to the linker. Typically
7379 this might be a file named @file{crt0.o}.
7382 Process the @code{endfile} spec. This is a spec string that specifies
7383 the last object files that will be passed to the linker.
7386 Process the @code{cpp} spec. This is used to construct the arguments
7387 to be passed to the C preprocessor.
7390 Process the @code{cc1} spec. This is used to construct the options to be
7391 passed to the actual C compiler (@samp{cc1}).
7394 Process the @code{cc1plus} spec. This is used to construct the options to be
7395 passed to the actual C++ compiler (@samp{cc1plus}).
7398 Substitute the variable part of a matched option. See below.
7399 Note that each comma in the substituted string is replaced by
7403 Remove all occurrences of @code{-S} from the command line. Note---this
7404 command is position dependent. @samp{%} commands in the spec string
7405 before this one will see @code{-S}, @samp{%} commands in the spec string
7406 after this one will not.
7408 @item %:@var{function}(@var{args})
7409 Call the named function @var{function}, passing it @var{args}.
7410 @var{args} is first processed as a nested spec string, then split
7411 into an argument vector in the usual fashion. The function returns
7412 a string which is processed as if it had appeared literally as part
7413 of the current spec.
7415 The following built-in spec functions are provided:
7418 @item @code{if-exists}
7419 The @code{if-exists} spec function takes one argument, an absolute
7420 pathname to a file. If the file exists, @code{if-exists} returns the
7421 pathname. Here is a small example of its usage:
7425 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7428 @item @code{if-exists-else}
7429 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7430 spec function, except that it takes two arguments. The first argument is
7431 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7432 returns the pathname. If it does not exist, it returns the second argument.
7433 This way, @code{if-exists-else} can be used to select one file or another,
7434 based on the existence of the first. Here is a small example of its usage:
7438 crt0%O%s %:if-exists(crti%O%s) \
7439 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7442 @item @code{replace-outfile}
7443 The @code{replace-outfile} spec function takes two arguments. It looks for the
7444 first argument in the outfiles array and replaces it with the second argument. Here
7445 is a small example of its usage:
7448 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7454 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7455 If that switch was not specified, this substitutes nothing. Note that
7456 the leading dash is omitted when specifying this option, and it is
7457 automatically inserted if the substitution is performed. Thus the spec
7458 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7459 and would output the command line option @option{-foo}.
7461 @item %W@{@code{S}@}
7462 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7465 @item %@{@code{S}*@}
7466 Substitutes all the switches specified to GCC whose names start
7467 with @code{-S}, but which also take an argument. This is used for
7468 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7469 GCC considers @option{-o foo} as being
7470 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7471 text, including the space. Thus two arguments would be generated.
7473 @item %@{@code{S}*&@code{T}*@}
7474 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7475 (the order of @code{S} and @code{T} in the spec is not significant).
7476 There can be any number of ampersand-separated variables; for each the
7477 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7479 @item %@{@code{S}:@code{X}@}
7480 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7482 @item %@{!@code{S}:@code{X}@}
7483 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7485 @item %@{@code{S}*:@code{X}@}
7486 Substitutes @code{X} if one or more switches whose names start with
7487 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7488 once, no matter how many such switches appeared. However, if @code{%*}
7489 appears somewhere in @code{X}, then @code{X} will be substituted once
7490 for each matching switch, with the @code{%*} replaced by the part of
7491 that switch that matched the @code{*}.
7493 @item %@{.@code{S}:@code{X}@}
7494 Substitutes @code{X}, if processing a file with suffix @code{S}.
7496 @item %@{!.@code{S}:@code{X}@}
7497 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7499 @item %@{@code{S}|@code{P}:@code{X}@}
7500 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7501 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7502 although they have a stronger binding than the @samp{|}. If @code{%*}
7503 appears in @code{X}, all of the alternatives must be starred, and only
7504 the first matching alternative is substituted.
7506 For example, a spec string like this:
7509 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7512 will output the following command-line options from the following input
7513 command-line options:
7518 -d fred.c -foo -baz -boggle
7519 -d jim.d -bar -baz -boggle
7522 @item %@{S:X; T:Y; :D@}
7524 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7525 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7526 be as many clauses as you need. This may be combined with @code{.},
7527 @code{!}, @code{|}, and @code{*} as needed.
7532 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7533 construct may contain other nested @samp{%} constructs or spaces, or
7534 even newlines. They are processed as usual, as described above.
7535 Trailing white space in @code{X} is ignored. White space may also
7536 appear anywhere on the left side of the colon in these constructs,
7537 except between @code{.} or @code{*} and the corresponding word.
7539 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7540 handled specifically in these constructs. If another value of
7541 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7542 @option{-W} switch is found later in the command line, the earlier
7543 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7544 just one letter, which passes all matching options.
7546 The character @samp{|} at the beginning of the predicate text is used to
7547 indicate that a command should be piped to the following command, but
7548 only if @option{-pipe} is specified.
7550 It is built into GCC which switches take arguments and which do not.
7551 (You might think it would be useful to generalize this to allow each
7552 compiler's spec to say which switches take arguments. But this cannot
7553 be done in a consistent fashion. GCC cannot even decide which input
7554 files have been specified without knowing which switches take arguments,
7555 and it must know which input files to compile in order to tell which
7558 GCC also knows implicitly that arguments starting in @option{-l} are to be
7559 treated as compiler output files, and passed to the linker in their
7560 proper position among the other output files.
7562 @c man begin OPTIONS
7564 @node Target Options
7565 @section Specifying Target Machine and Compiler Version
7566 @cindex target options
7567 @cindex cross compiling
7568 @cindex specifying machine version
7569 @cindex specifying compiler version and target machine
7570 @cindex compiler version, specifying
7571 @cindex target machine, specifying
7573 The usual way to run GCC is to run the executable called @file{gcc}, or
7574 @file{<machine>-gcc} when cross-compiling, or
7575 @file{<machine>-gcc-<version>} to run a version other than the one that
7576 was installed last. Sometimes this is inconvenient, so GCC provides
7577 options that will switch to another cross-compiler or version.
7580 @item -b @var{machine}
7582 The argument @var{machine} specifies the target machine for compilation.
7584 The value to use for @var{machine} is the same as was specified as the
7585 machine type when configuring GCC as a cross-compiler. For
7586 example, if a cross-compiler was configured with @samp{configure
7587 arm-elf}, meaning to compile for an arm processor with elf binaries,
7588 then you would specify @option{-b arm-elf} to run that cross compiler.
7589 Because there are other options beginning with @option{-b}, the
7590 configuration must contain a hyphen.
7592 @item -V @var{version}
7594 The argument @var{version} specifies which version of GCC to run.
7595 This is useful when multiple versions are installed. For example,
7596 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7599 The @option{-V} and @option{-b} options work by running the
7600 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7601 use them if you can just run that directly.
7603 @node Submodel Options
7604 @section Hardware Models and Configurations
7605 @cindex submodel options
7606 @cindex specifying hardware config
7607 @cindex hardware models and configurations, specifying
7608 @cindex machine dependent options
7610 Earlier we discussed the standard option @option{-b} which chooses among
7611 different installed compilers for completely different target
7612 machines, such as VAX vs.@: 68000 vs.@: 80386.
7614 In addition, each of these target machine types can have its own
7615 special options, starting with @samp{-m}, to choose among various
7616 hardware models or configurations---for example, 68010 vs 68020,
7617 floating coprocessor or none. A single installed version of the
7618 compiler can compile for any model or configuration, according to the
7621 Some configurations of the compiler also support additional special
7622 options, usually for compatibility with other compilers on the same
7625 @c This list is ordered alphanumerically by subsection name.
7626 @c It should be the same order and spelling as these options are listed
7627 @c in Machine Dependent Options
7633 * Blackfin Options::
7637 * DEC Alpha Options::
7638 * DEC Alpha/VMS Options::
7640 * GNU/Linux Options::
7643 * i386 and x86-64 Options::
7656 * RS/6000 and PowerPC Options::
7657 * S/390 and zSeries Options::
7662 * System V Options::
7663 * TMS320C3x/C4x Options::
7667 * Xstormy16 Options::
7673 @subsection ARC Options
7676 These options are defined for ARC implementations:
7681 Compile code for little endian mode. This is the default.
7685 Compile code for big endian mode.
7688 @opindex mmangle-cpu
7689 Prepend the name of the cpu to all public symbol names.
7690 In multiple-processor systems, there are many ARC variants with different
7691 instruction and register set characteristics. This flag prevents code
7692 compiled for one cpu to be linked with code compiled for another.
7693 No facility exists for handling variants that are ``almost identical''.
7694 This is an all or nothing option.
7696 @item -mcpu=@var{cpu}
7698 Compile code for ARC variant @var{cpu}.
7699 Which variants are supported depend on the configuration.
7700 All variants support @option{-mcpu=base}, this is the default.
7702 @item -mtext=@var{text-section}
7703 @itemx -mdata=@var{data-section}
7704 @itemx -mrodata=@var{readonly-data-section}
7708 Put functions, data, and readonly data in @var{text-section},
7709 @var{data-section}, and @var{readonly-data-section} respectively
7710 by default. This can be overridden with the @code{section} attribute.
7711 @xref{Variable Attributes}.
7716 @subsection ARM Options
7719 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7723 @item -mabi=@var{name}
7725 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7726 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7729 @opindex mapcs-frame
7730 Generate a stack frame that is compliant with the ARM Procedure Call
7731 Standard for all functions, even if this is not strictly necessary for
7732 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7733 with this option will cause the stack frames not to be generated for
7734 leaf functions. The default is @option{-mno-apcs-frame}.
7738 This is a synonym for @option{-mapcs-frame}.
7741 @c not currently implemented
7742 @item -mapcs-stack-check
7743 @opindex mapcs-stack-check
7744 Generate code to check the amount of stack space available upon entry to
7745 every function (that actually uses some stack space). If there is
7746 insufficient space available then either the function
7747 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7748 called, depending upon the amount of stack space required. The run time
7749 system is required to provide these functions. The default is
7750 @option{-mno-apcs-stack-check}, since this produces smaller code.
7752 @c not currently implemented
7754 @opindex mapcs-float
7755 Pass floating point arguments using the float point registers. This is
7756 one of the variants of the APCS@. This option is recommended if the
7757 target hardware has a floating point unit or if a lot of floating point
7758 arithmetic is going to be performed by the code. The default is
7759 @option{-mno-apcs-float}, since integer only code is slightly increased in
7760 size if @option{-mapcs-float} is used.
7762 @c not currently implemented
7763 @item -mapcs-reentrant
7764 @opindex mapcs-reentrant
7765 Generate reentrant, position independent code. The default is
7766 @option{-mno-apcs-reentrant}.
7769 @item -mthumb-interwork
7770 @opindex mthumb-interwork
7771 Generate code which supports calling between the ARM and Thumb
7772 instruction sets. Without this option the two instruction sets cannot
7773 be reliably used inside one program. The default is
7774 @option{-mno-thumb-interwork}, since slightly larger code is generated
7775 when @option{-mthumb-interwork} is specified.
7777 @item -mno-sched-prolog
7778 @opindex mno-sched-prolog
7779 Prevent the reordering of instructions in the function prolog, or the
7780 merging of those instruction with the instructions in the function's
7781 body. This means that all functions will start with a recognizable set
7782 of instructions (or in fact one of a choice from a small set of
7783 different function prologues), and this information can be used to
7784 locate the start if functions inside an executable piece of code. The
7785 default is @option{-msched-prolog}.
7788 @opindex mhard-float
7789 Generate output containing floating point instructions. This is the
7793 @opindex msoft-float
7794 Generate output containing library calls for floating point.
7795 @strong{Warning:} the requisite libraries are not available for all ARM
7796 targets. Normally the facilities of the machine's usual C compiler are
7797 used, but this cannot be done directly in cross-compilation. You must make
7798 your own arrangements to provide suitable library functions for
7801 @option{-msoft-float} changes the calling convention in the output file;
7802 therefore, it is only useful if you compile @emph{all} of a program with
7803 this option. In particular, you need to compile @file{libgcc.a}, the
7804 library that comes with GCC, with @option{-msoft-float} in order for
7807 @item -mfloat-abi=@var{name}
7809 Specifies which ABI to use for floating point values. Permissible values
7810 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7812 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7813 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7814 of floating point instructions, but still uses the soft-float calling
7817 @item -mlittle-endian
7818 @opindex mlittle-endian
7819 Generate code for a processor running in little-endian mode. This is
7820 the default for all standard configurations.
7823 @opindex mbig-endian
7824 Generate code for a processor running in big-endian mode; the default is
7825 to compile code for a little-endian processor.
7827 @item -mwords-little-endian
7828 @opindex mwords-little-endian
7829 This option only applies when generating code for big-endian processors.
7830 Generate code for a little-endian word order but a big-endian byte
7831 order. That is, a byte order of the form @samp{32107654}. Note: this
7832 option should only be used if you require compatibility with code for
7833 big-endian ARM processors generated by versions of the compiler prior to
7836 @item -mcpu=@var{name}
7838 This specifies the name of the target ARM processor. GCC uses this name
7839 to determine what kind of instructions it can emit when generating
7840 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7841 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7842 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7843 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7844 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7845 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7846 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7847 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7848 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7849 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7850 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7851 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7852 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7853 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
7854 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
7855 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
7857 @itemx -mtune=@var{name}
7859 This option is very similar to the @option{-mcpu=} option, except that
7860 instead of specifying the actual target processor type, and hence
7861 restricting which instructions can be used, it specifies that GCC should
7862 tune the performance of the code as if the target were of the type
7863 specified in this option, but still choosing the instructions that it
7864 will generate based on the cpu specified by a @option{-mcpu=} option.
7865 For some ARM implementations better performance can be obtained by using
7868 @item -march=@var{name}
7870 This specifies the name of the target ARM architecture. GCC uses this
7871 name to determine what kind of instructions it can emit when generating
7872 assembly code. This option can be used in conjunction with or instead
7873 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7874 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7875 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7876 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
7877 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
7879 @item -mfpu=@var{name}
7880 @itemx -mfpe=@var{number}
7881 @itemx -mfp=@var{number}
7885 This specifies what floating point hardware (or hardware emulation) is
7886 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7887 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7888 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7889 with older versions of GCC@.
7891 If @option{-msoft-float} is specified this specifies the format of
7892 floating point values.
7894 @item -mstructure-size-boundary=@var{n}
7895 @opindex mstructure-size-boundary
7896 The size of all structures and unions will be rounded up to a multiple
7897 of the number of bits set by this option. Permissible values are 8, 32
7898 and 64. The default value varies for different toolchains. For the COFF
7899 targeted toolchain the default value is 8. A value of 64 is only allowed
7900 if the underlying ABI supports it.
7902 Specifying the larger number can produce faster, more efficient code, but
7903 can also increase the size of the program. Different values are potentially
7904 incompatible. Code compiled with one value cannot necessarily expect to
7905 work with code or libraries compiled with another value, if they exchange
7906 information using structures or unions.
7908 @item -mabort-on-noreturn
7909 @opindex mabort-on-noreturn
7910 Generate a call to the function @code{abort} at the end of a
7911 @code{noreturn} function. It will be executed if the function tries to
7915 @itemx -mno-long-calls
7916 @opindex mlong-calls
7917 @opindex mno-long-calls
7918 Tells the compiler to perform function calls by first loading the
7919 address of the function into a register and then performing a subroutine
7920 call on this register. This switch is needed if the target function
7921 will lie outside of the 64 megabyte addressing range of the offset based
7922 version of subroutine call instruction.
7924 Even if this switch is enabled, not all function calls will be turned
7925 into long calls. The heuristic is that static functions, functions
7926 which have the @samp{short-call} attribute, functions that are inside
7927 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7928 definitions have already been compiled within the current compilation
7929 unit, will not be turned into long calls. The exception to this rule is
7930 that weak function definitions, functions with the @samp{long-call}
7931 attribute or the @samp{section} attribute, and functions that are within
7932 the scope of a @samp{#pragma long_calls} directive, will always be
7933 turned into long calls.
7935 This feature is not enabled by default. Specifying
7936 @option{-mno-long-calls} will restore the default behavior, as will
7937 placing the function calls within the scope of a @samp{#pragma
7938 long_calls_off} directive. Note these switches have no effect on how
7939 the compiler generates code to handle function calls via function
7942 @item -mnop-fun-dllimport
7943 @opindex mnop-fun-dllimport
7944 Disable support for the @code{dllimport} attribute.
7946 @item -msingle-pic-base
7947 @opindex msingle-pic-base
7948 Treat the register used for PIC addressing as read-only, rather than
7949 loading it in the prologue for each function. The run-time system is
7950 responsible for initializing this register with an appropriate value
7951 before execution begins.
7953 @item -mpic-register=@var{reg}
7954 @opindex mpic-register
7955 Specify the register to be used for PIC addressing. The default is R10
7956 unless stack-checking is enabled, when R9 is used.
7958 @item -mcirrus-fix-invalid-insns
7959 @opindex mcirrus-fix-invalid-insns
7960 @opindex mno-cirrus-fix-invalid-insns
7961 Insert NOPs into the instruction stream to in order to work around
7962 problems with invalid Maverick instruction combinations. This option
7963 is only valid if the @option{-mcpu=ep9312} option has been used to
7964 enable generation of instructions for the Cirrus Maverick floating
7965 point co-processor. This option is not enabled by default, since the
7966 problem is only present in older Maverick implementations. The default
7967 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7970 @item -mpoke-function-name
7971 @opindex mpoke-function-name
7972 Write the name of each function into the text section, directly
7973 preceding the function prologue. The generated code is similar to this:
7977 .ascii "arm_poke_function_name", 0
7980 .word 0xff000000 + (t1 - t0)
7981 arm_poke_function_name
7983 stmfd sp!, @{fp, ip, lr, pc@}
7987 When performing a stack backtrace, code can inspect the value of
7988 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7989 location @code{pc - 12} and the top 8 bits are set, then we know that
7990 there is a function name embedded immediately preceding this location
7991 and has length @code{((pc[-3]) & 0xff000000)}.
7995 Generate code for the Thumb instruction set. The default is to
7996 use the 32-bit ARM instruction set.
7997 This option automatically enables either 16-bit Thumb-1 or
7998 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
7999 and @option{-march=@var{name}} options.
8002 @opindex mtpcs-frame
8003 Generate a stack frame that is compliant with the Thumb Procedure Call
8004 Standard for all non-leaf functions. (A leaf function is one that does
8005 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8007 @item -mtpcs-leaf-frame
8008 @opindex mtpcs-leaf-frame
8009 Generate a stack frame that is compliant with the Thumb Procedure Call
8010 Standard for all leaf functions. (A leaf function is one that does
8011 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8013 @item -mcallee-super-interworking
8014 @opindex mcallee-super-interworking
8015 Gives all externally visible functions in the file being compiled an ARM
8016 instruction set header which switches to Thumb mode before executing the
8017 rest of the function. This allows these functions to be called from
8018 non-interworking code.
8020 @item -mcaller-super-interworking
8021 @opindex mcaller-super-interworking
8022 Allows calls via function pointers (including virtual functions) to
8023 execute correctly regardless of whether the target code has been
8024 compiled for interworking or not. There is a small overhead in the cost
8025 of executing a function pointer if this option is enabled.
8027 @item -mtp=@var{name}
8029 Specify the access model for the thread local storage pointer. The valid
8030 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8031 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8032 (supported in the arm6k architecture), and @option{auto}, which uses the
8033 best available method for the selected processor. The default setting is
8039 @subsection AVR Options
8042 These options are defined for AVR implementations:
8045 @item -mmcu=@var{mcu}
8047 Specify ATMEL AVR instruction set or MCU type.
8049 Instruction set avr1 is for the minimal AVR core, not supported by the C
8050 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8051 attiny11, attiny12, attiny15, attiny28).
8053 Instruction set avr2 (default) is for the classic AVR core with up to
8054 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8055 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8056 at90c8534, at90s8535).
8058 Instruction set avr3 is for the classic AVR core with up to 128K program
8059 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8061 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8062 memory space (MCU types: atmega8, atmega83, atmega85).
8064 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8065 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8066 atmega64, atmega128, at43usb355, at94k).
8070 Output instruction sizes to the asm file.
8072 @item -minit-stack=@var{N}
8073 @opindex minit-stack
8074 Specify the initial stack address, which may be a symbol or numeric value,
8075 @samp{__stack} is the default.
8077 @item -mno-interrupts
8078 @opindex mno-interrupts
8079 Generated code is not compatible with hardware interrupts.
8080 Code size will be smaller.
8082 @item -mcall-prologues
8083 @opindex mcall-prologues
8084 Functions prologues/epilogues expanded as call to appropriate
8085 subroutines. Code size will be smaller.
8087 @item -mno-tablejump
8088 @opindex mno-tablejump
8089 Do not generate tablejump insns which sometimes increase code size.
8092 @opindex mtiny-stack
8093 Change only the low 8 bits of the stack pointer.
8097 Assume int to be 8 bit integer. This affects the sizes of all types: A
8098 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8099 and long long will be 4 bytes. Please note that this option does not
8100 comply to the C standards, but it will provide you with smaller code
8104 @node Blackfin Options
8105 @subsection Blackfin Options
8106 @cindex Blackfin Options
8109 @item -momit-leaf-frame-pointer
8110 @opindex momit-leaf-frame-pointer
8111 Don't keep the frame pointer in a register for leaf functions. This
8112 avoids the instructions to save, set up and restore frame pointers and
8113 makes an extra register available in leaf functions. The option
8114 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8115 which might make debugging harder.
8117 @item -mspecld-anomaly
8118 @opindex mspecld-anomaly
8119 When enabled, the compiler will ensure that the generated code does not
8120 contain speculative loads after jump instructions. This option is enabled
8123 @item -mno-specld-anomaly
8124 @opindex mno-specld-anomaly
8125 Don't generate extra code to prevent speculative loads from occurring.
8127 @item -mcsync-anomaly
8128 @opindex mcsync-anomaly
8129 When enabled, the compiler will ensure that the generated code does not
8130 contain CSYNC or SSYNC instructions too soon after conditional branches.
8131 This option is enabled by default.
8133 @item -mno-csync-anomaly
8134 @opindex mno-csync-anomaly
8135 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8136 occurring too soon after a conditional branch.
8140 When enabled, the compiler is free to take advantage of the knowledge that
8141 the entire program fits into the low 64k of memory.
8144 @opindex mno-low-64k
8145 Assume that the program is arbitrarily large. This is the default.
8147 @item -mstack-check-l1
8148 @opindex mstack-check-l1
8149 Do stack checking using information placed into L1 scratchpad memory by the
8152 @item -mid-shared-library
8153 @opindex mid-shared-library
8154 Generate code that supports shared libraries via the library ID method.
8155 This allows for execute in place and shared libraries in an environment
8156 without virtual memory management. This option implies @option{-fPIC}.
8158 @item -mno-id-shared-library
8159 @opindex mno-id-shared-library
8160 Generate code that doesn't assume ID based shared libraries are being used.
8161 This is the default.
8163 @item -mleaf-id-shared-library
8164 @opindex mleaf-id-shared-library
8165 Generate code that supports shared libraries via the library ID method,
8166 but assumes that this library or executable won't link against any other
8167 ID shared libraries. That allows the compiler to use faster code for jumps
8170 @item -mno-leaf-id-shared-library
8171 @opindex mno-leaf-id-shared-library
8172 Do not assume that the code being compiled won't link against any ID shared
8173 libraries. Slower code will be generated for jump and call insns.
8175 @item -mshared-library-id=n
8176 @opindex mshared-library-id
8177 Specified the identification number of the ID based shared library being
8178 compiled. Specifying a value of 0 will generate more compact code, specifying
8179 other values will force the allocation of that number to the current
8180 library but is no more space or time efficient than omitting this option.
8184 Generate code that allows the data segment to be located in a different
8185 area of memory from the text segment. This allows for execute in place in
8186 an environment without virtual memory management by eliminating relocations
8187 against the text section.
8190 @opindex mno-sep-data
8191 Generate code that assumes that the data segment follows the text segment.
8192 This is the default.
8195 @itemx -mno-long-calls
8196 @opindex mlong-calls
8197 @opindex mno-long-calls
8198 Tells the compiler to perform function calls by first loading the
8199 address of the function into a register and then performing a subroutine
8200 call on this register. This switch is needed if the target function
8201 will lie outside of the 24 bit addressing range of the offset based
8202 version of subroutine call instruction.
8204 This feature is not enabled by default. Specifying
8205 @option{-mno-long-calls} will restore the default behavior. Note these
8206 switches have no effect on how the compiler generates code to handle
8207 function calls via function pointers.
8211 @subsection CRIS Options
8212 @cindex CRIS Options
8214 These options are defined specifically for the CRIS ports.
8217 @item -march=@var{architecture-type}
8218 @itemx -mcpu=@var{architecture-type}
8221 Generate code for the specified architecture. The choices for
8222 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8223 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8224 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8227 @item -mtune=@var{architecture-type}
8229 Tune to @var{architecture-type} everything applicable about the generated
8230 code, except for the ABI and the set of available instructions. The
8231 choices for @var{architecture-type} are the same as for
8232 @option{-march=@var{architecture-type}}.
8234 @item -mmax-stack-frame=@var{n}
8235 @opindex mmax-stack-frame
8236 Warn when the stack frame of a function exceeds @var{n} bytes.
8238 @item -melinux-stacksize=@var{n}
8239 @opindex melinux-stacksize
8240 Only available with the @samp{cris-axis-aout} target. Arranges for
8241 indications in the program to the kernel loader that the stack of the
8242 program should be set to @var{n} bytes.
8248 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8249 @option{-march=v3} and @option{-march=v8} respectively.
8251 @item -mmul-bug-workaround
8252 @itemx -mno-mul-bug-workaround
8253 @opindex mmul-bug-workaround
8254 @opindex mno-mul-bug-workaround
8255 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8256 models where it applies. This option is active by default.
8260 Enable CRIS-specific verbose debug-related information in the assembly
8261 code. This option also has the effect to turn off the @samp{#NO_APP}
8262 formatted-code indicator to the assembler at the beginning of the
8267 Do not use condition-code results from previous instruction; always emit
8268 compare and test instructions before use of condition codes.
8270 @item -mno-side-effects
8271 @opindex mno-side-effects
8272 Do not emit instructions with side-effects in addressing modes other than
8276 @itemx -mno-stack-align
8278 @itemx -mno-data-align
8279 @itemx -mconst-align
8280 @itemx -mno-const-align
8281 @opindex mstack-align
8282 @opindex mno-stack-align
8283 @opindex mdata-align
8284 @opindex mno-data-align
8285 @opindex mconst-align
8286 @opindex mno-const-align
8287 These options (no-options) arranges (eliminate arrangements) for the
8288 stack-frame, individual data and constants to be aligned for the maximum
8289 single data access size for the chosen CPU model. The default is to
8290 arrange for 32-bit alignment. ABI details such as structure layout are
8291 not affected by these options.
8299 Similar to the stack- data- and const-align options above, these options
8300 arrange for stack-frame, writable data and constants to all be 32-bit,
8301 16-bit or 8-bit aligned. The default is 32-bit alignment.
8303 @item -mno-prologue-epilogue
8304 @itemx -mprologue-epilogue
8305 @opindex mno-prologue-epilogue
8306 @opindex mprologue-epilogue
8307 With @option{-mno-prologue-epilogue}, the normal function prologue and
8308 epilogue that sets up the stack-frame are omitted and no return
8309 instructions or return sequences are generated in the code. Use this
8310 option only together with visual inspection of the compiled code: no
8311 warnings or errors are generated when call-saved registers must be saved,
8312 or storage for local variable needs to be allocated.
8318 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8319 instruction sequences that load addresses for functions from the PLT part
8320 of the GOT rather than (traditional on other architectures) calls to the
8321 PLT@. The default is @option{-mgotplt}.
8325 Legacy no-op option only recognized with the cris-axis-aout target.
8329 Legacy no-op option only recognized with the cris-axis-elf and
8330 cris-axis-linux-gnu targets.
8334 Only recognized with the cris-axis-aout target, where it selects a
8335 GNU/linux-like multilib, include files and instruction set for
8340 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8344 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8345 to link with input-output functions from a simulator library. Code,
8346 initialized data and zero-initialized data are allocated consecutively.
8350 Like @option{-sim}, but pass linker options to locate initialized data at
8351 0x40000000 and zero-initialized data at 0x80000000.
8355 @subsection CRX Options
8358 These options are defined specifically for the CRX ports.
8364 Enable the use of multiply-accumulate instructions. Disabled by default.
8368 Push instructions will be used to pass outgoing arguments when functions
8369 are called. Enabled by default.
8372 @node Darwin Options
8373 @subsection Darwin Options
8374 @cindex Darwin options
8376 These options are defined for all architectures running the Darwin operating
8379 FSF GCC on Darwin does not create ``fat'' object files; it will create
8380 an object file for the single architecture that it was built to
8381 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8382 @option{-arch} options are used; it does so by running the compiler or
8383 linker multiple times and joining the results together with
8386 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8387 @samp{i686}) is determined by the flags that specify the ISA
8388 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8389 @option{-force_cpusubtype_ALL} option can be used to override this.
8391 The Darwin tools vary in their behavior when presented with an ISA
8392 mismatch. The assembler, @file{as}, will only permit instructions to
8393 be used that are valid for the subtype of the file it is generating,
8394 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8395 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8396 and print an error if asked to create a shared library with a less
8397 restrictive subtype than its input files (for instance, trying to put
8398 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8399 for executables, @file{ld}, will quietly give the executable the most
8400 restrictive subtype of any of its input files.
8405 Add the framework directory @var{dir} to the head of the list of
8406 directories to be searched for header files. These directories are
8407 interleaved with those specified by @option{-I} options and are
8408 scanned in a left-to-right order.
8410 A framework directory is a directory with frameworks in it. A
8411 framework is a directory with a @samp{"Headers"} and/or
8412 @samp{"PrivateHeaders"} directory contained directly in it that ends
8413 in @samp{".framework"}. The name of a framework is the name of this
8414 directory excluding the @samp{".framework"}. Headers associated with
8415 the framework are found in one of those two directories, with
8416 @samp{"Headers"} being searched first. A subframework is a framework
8417 directory that is in a framework's @samp{"Frameworks"} directory.
8418 Includes of subframework headers can only appear in a header of a
8419 framework that contains the subframework, or in a sibling subframework
8420 header. Two subframeworks are siblings if they occur in the same
8421 framework. A subframework should not have the same name as a
8422 framework, a warning will be issued if this is violated. Currently a
8423 subframework cannot have subframeworks, in the future, the mechanism
8424 may be extended to support this. The standard frameworks can be found
8425 in @samp{"/System/Library/Frameworks"} and
8426 @samp{"/Library/Frameworks"}. An example include looks like
8427 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8428 the name of the framework and header.h is found in the
8429 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8433 Emit debugging information for symbols that are used. For STABS
8434 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8435 This is by default ON@.
8439 Emit debugging information for all symbols and types.
8441 @item -mmacosx-version-min=@var{version}
8442 The earliest version of MacOS X that this executable will run on
8443 is @var{version}. Typical values of @var{version} include @code{10.1},
8444 @code{10.2}, and @code{10.3.9}.
8446 The default for this option is to make choices that seem to be most
8451 Enable kernel development mode. The @option{-mkernel} option sets
8452 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8453 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8454 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8455 applicable. This mode also sets @option{-mno-altivec},
8456 @option{-msoft-float}, @option{-fno-builtin} and
8457 @option{-mlong-branch} for PowerPC targets.
8459 @item -mone-byte-bool
8460 @opindex -mone-byte-bool
8461 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8462 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8463 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8464 option has no effect on x86.
8466 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8467 to generate code that is not binary compatible with code generated
8468 without that switch. Using this switch may require recompiling all
8469 other modules in a program, including system libraries. Use this
8470 switch to conform to a non-default data model.
8472 @item -mfix-and-continue
8473 @itemx -ffix-and-continue
8474 @itemx -findirect-data
8475 @opindex mfix-and-continue
8476 @opindex ffix-and-continue
8477 @opindex findirect-data
8478 Generate code suitable for fast turn around development. Needed to
8479 enable gdb to dynamically load @code{.o} files into already running
8480 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8481 are provided for backwards compatibility.
8485 Loads all members of static archive libraries.
8486 See man ld(1) for more information.
8488 @item -arch_errors_fatal
8489 @opindex arch_errors_fatal
8490 Cause the errors having to do with files that have the wrong architecture
8494 @opindex bind_at_load
8495 Causes the output file to be marked such that the dynamic linker will
8496 bind all undefined references when the file is loaded or launched.
8500 Produce a Mach-o bundle format file.
8501 See man ld(1) for more information.
8503 @item -bundle_loader @var{executable}
8504 @opindex bundle_loader
8505 This option specifies the @var{executable} that will be loading the build
8506 output file being linked. See man ld(1) for more information.
8509 @opindex -dynamiclib
8510 When passed this option, GCC will produce a dynamic library instead of
8511 an executable when linking, using the Darwin @file{libtool} command.
8513 @item -force_cpusubtype_ALL
8514 @opindex -force_cpusubtype_ALL
8515 This causes GCC's output file to have the @var{ALL} subtype, instead of
8516 one controlled by the @option{-mcpu} or @option{-march} option.
8518 @item -allowable_client @var{client_name}
8520 @itemx -compatibility_version
8521 @itemx -current_version
8523 @itemx -dependency-file
8525 @itemx -dylinker_install_name
8527 @itemx -exported_symbols_list
8529 @itemx -flat_namespace
8530 @itemx -force_flat_namespace
8531 @itemx -headerpad_max_install_names
8534 @itemx -install_name
8535 @itemx -keep_private_externs
8536 @itemx -multi_module
8537 @itemx -multiply_defined
8538 @itemx -multiply_defined_unused
8540 @itemx -no_dead_strip_inits_and_terms
8541 @itemx -nofixprebinding
8544 @itemx -noseglinkedit
8545 @itemx -pagezero_size
8547 @itemx -prebind_all_twolevel_modules
8548 @itemx -private_bundle
8549 @itemx -read_only_relocs
8551 @itemx -sectobjectsymbols
8555 @itemx -sectobjectsymbols
8558 @itemx -segs_read_only_addr
8559 @itemx -segs_read_write_addr
8560 @itemx -seg_addr_table
8561 @itemx -seg_addr_table_filename
8564 @itemx -segs_read_only_addr
8565 @itemx -segs_read_write_addr
8566 @itemx -single_module
8569 @itemx -sub_umbrella
8570 @itemx -twolevel_namespace
8573 @itemx -unexported_symbols_list
8574 @itemx -weak_reference_mismatches
8577 @opindex allowable_client
8578 @opindex client_name
8579 @opindex compatibility_version
8580 @opindex current_version
8582 @opindex dependency-file
8584 @opindex dylinker_install_name
8586 @opindex exported_symbols_list
8588 @opindex flat_namespace
8589 @opindex force_flat_namespace
8590 @opindex headerpad_max_install_names
8593 @opindex install_name
8594 @opindex keep_private_externs
8595 @opindex multi_module
8596 @opindex multiply_defined
8597 @opindex multiply_defined_unused
8599 @opindex no_dead_strip_inits_and_terms
8600 @opindex nofixprebinding
8601 @opindex nomultidefs
8603 @opindex noseglinkedit
8604 @opindex pagezero_size
8606 @opindex prebind_all_twolevel_modules
8607 @opindex private_bundle
8608 @opindex read_only_relocs
8610 @opindex sectobjectsymbols
8614 @opindex sectobjectsymbols
8617 @opindex segs_read_only_addr
8618 @opindex segs_read_write_addr
8619 @opindex seg_addr_table
8620 @opindex seg_addr_table_filename
8621 @opindex seglinkedit
8623 @opindex segs_read_only_addr
8624 @opindex segs_read_write_addr
8625 @opindex single_module
8627 @opindex sub_library
8628 @opindex sub_umbrella
8629 @opindex twolevel_namespace
8632 @opindex unexported_symbols_list
8633 @opindex weak_reference_mismatches
8634 @opindex whatsloaded
8636 These options are passed to the Darwin linker. The Darwin linker man page
8637 describes them in detail.
8640 @node DEC Alpha Options
8641 @subsection DEC Alpha Options
8643 These @samp{-m} options are defined for the DEC Alpha implementations:
8646 @item -mno-soft-float
8648 @opindex mno-soft-float
8649 @opindex msoft-float
8650 Use (do not use) the hardware floating-point instructions for
8651 floating-point operations. When @option{-msoft-float} is specified,
8652 functions in @file{libgcc.a} will be used to perform floating-point
8653 operations. Unless they are replaced by routines that emulate the
8654 floating-point operations, or compiled in such a way as to call such
8655 emulations routines, these routines will issue floating-point
8656 operations. If you are compiling for an Alpha without floating-point
8657 operations, you must ensure that the library is built so as not to call
8660 Note that Alpha implementations without floating-point operations are
8661 required to have floating-point registers.
8666 @opindex mno-fp-regs
8667 Generate code that uses (does not use) the floating-point register set.
8668 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8669 register set is not used, floating point operands are passed in integer
8670 registers as if they were integers and floating-point results are passed
8671 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8672 so any function with a floating-point argument or return value called by code
8673 compiled with @option{-mno-fp-regs} must also be compiled with that
8676 A typical use of this option is building a kernel that does not use,
8677 and hence need not save and restore, any floating-point registers.
8681 The Alpha architecture implements floating-point hardware optimized for
8682 maximum performance. It is mostly compliant with the IEEE floating
8683 point standard. However, for full compliance, software assistance is
8684 required. This option generates code fully IEEE compliant code
8685 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8686 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8687 defined during compilation. The resulting code is less efficient but is
8688 able to correctly support denormalized numbers and exceptional IEEE
8689 values such as not-a-number and plus/minus infinity. Other Alpha
8690 compilers call this option @option{-ieee_with_no_inexact}.
8692 @item -mieee-with-inexact
8693 @opindex mieee-with-inexact
8694 This is like @option{-mieee} except the generated code also maintains
8695 the IEEE @var{inexact-flag}. Turning on this option causes the
8696 generated code to implement fully-compliant IEEE math. In addition to
8697 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8698 macro. On some Alpha implementations the resulting code may execute
8699 significantly slower than the code generated by default. Since there is
8700 very little code that depends on the @var{inexact-flag}, you should
8701 normally not specify this option. Other Alpha compilers call this
8702 option @option{-ieee_with_inexact}.
8704 @item -mfp-trap-mode=@var{trap-mode}
8705 @opindex mfp-trap-mode
8706 This option controls what floating-point related traps are enabled.
8707 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8708 The trap mode can be set to one of four values:
8712 This is the default (normal) setting. The only traps that are enabled
8713 are the ones that cannot be disabled in software (e.g., division by zero
8717 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8721 Like @samp{u}, but the instructions are marked to be safe for software
8722 completion (see Alpha architecture manual for details).
8725 Like @samp{su}, but inexact traps are enabled as well.
8728 @item -mfp-rounding-mode=@var{rounding-mode}
8729 @opindex mfp-rounding-mode
8730 Selects the IEEE rounding mode. Other Alpha compilers call this option
8731 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8736 Normal IEEE rounding mode. Floating point numbers are rounded towards
8737 the nearest machine number or towards the even machine number in case
8741 Round towards minus infinity.
8744 Chopped rounding mode. Floating point numbers are rounded towards zero.
8747 Dynamic rounding mode. A field in the floating point control register
8748 (@var{fpcr}, see Alpha architecture reference manual) controls the
8749 rounding mode in effect. The C library initializes this register for
8750 rounding towards plus infinity. Thus, unless your program modifies the
8751 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8754 @item -mtrap-precision=@var{trap-precision}
8755 @opindex mtrap-precision
8756 In the Alpha architecture, floating point traps are imprecise. This
8757 means without software assistance it is impossible to recover from a
8758 floating trap and program execution normally needs to be terminated.
8759 GCC can generate code that can assist operating system trap handlers
8760 in determining the exact location that caused a floating point trap.
8761 Depending on the requirements of an application, different levels of
8762 precisions can be selected:
8766 Program precision. This option is the default and means a trap handler
8767 can only identify which program caused a floating point exception.
8770 Function precision. The trap handler can determine the function that
8771 caused a floating point exception.
8774 Instruction precision. The trap handler can determine the exact
8775 instruction that caused a floating point exception.
8778 Other Alpha compilers provide the equivalent options called
8779 @option{-scope_safe} and @option{-resumption_safe}.
8781 @item -mieee-conformant
8782 @opindex mieee-conformant
8783 This option marks the generated code as IEEE conformant. You must not
8784 use this option unless you also specify @option{-mtrap-precision=i} and either
8785 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8786 is to emit the line @samp{.eflag 48} in the function prologue of the
8787 generated assembly file. Under DEC Unix, this has the effect that
8788 IEEE-conformant math library routines will be linked in.
8790 @item -mbuild-constants
8791 @opindex mbuild-constants
8792 Normally GCC examines a 32- or 64-bit integer constant to
8793 see if it can construct it from smaller constants in two or three
8794 instructions. If it cannot, it will output the constant as a literal and
8795 generate code to load it from the data segment at runtime.
8797 Use this option to require GCC to construct @emph{all} integer constants
8798 using code, even if it takes more instructions (the maximum is six).
8800 You would typically use this option to build a shared library dynamic
8801 loader. Itself a shared library, it must relocate itself in memory
8802 before it can find the variables and constants in its own data segment.
8808 Select whether to generate code to be assembled by the vendor-supplied
8809 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8827 Indicate whether GCC should generate code to use the optional BWX,
8828 CIX, FIX and MAX instruction sets. The default is to use the instruction
8829 sets supported by the CPU type specified via @option{-mcpu=} option or that
8830 of the CPU on which GCC was built if none was specified.
8835 @opindex mfloat-ieee
8836 Generate code that uses (does not use) VAX F and G floating point
8837 arithmetic instead of IEEE single and double precision.
8839 @item -mexplicit-relocs
8840 @itemx -mno-explicit-relocs
8841 @opindex mexplicit-relocs
8842 @opindex mno-explicit-relocs
8843 Older Alpha assemblers provided no way to generate symbol relocations
8844 except via assembler macros. Use of these macros does not allow
8845 optimal instruction scheduling. GNU binutils as of version 2.12
8846 supports a new syntax that allows the compiler to explicitly mark
8847 which relocations should apply to which instructions. This option
8848 is mostly useful for debugging, as GCC detects the capabilities of
8849 the assembler when it is built and sets the default accordingly.
8853 @opindex msmall-data
8854 @opindex mlarge-data
8855 When @option{-mexplicit-relocs} is in effect, static data is
8856 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8857 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8858 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8859 16-bit relocations off of the @code{$gp} register. This limits the
8860 size of the small data area to 64KB, but allows the variables to be
8861 directly accessed via a single instruction.
8863 The default is @option{-mlarge-data}. With this option the data area
8864 is limited to just below 2GB@. Programs that require more than 2GB of
8865 data must use @code{malloc} or @code{mmap} to allocate the data in the
8866 heap instead of in the program's data segment.
8868 When generating code for shared libraries, @option{-fpic} implies
8869 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8873 @opindex msmall-text
8874 @opindex mlarge-text
8875 When @option{-msmall-text} is used, the compiler assumes that the
8876 code of the entire program (or shared library) fits in 4MB, and is
8877 thus reachable with a branch instruction. When @option{-msmall-data}
8878 is used, the compiler can assume that all local symbols share the
8879 same @code{$gp} value, and thus reduce the number of instructions
8880 required for a function call from 4 to 1.
8882 The default is @option{-mlarge-text}.
8884 @item -mcpu=@var{cpu_type}
8886 Set the instruction set and instruction scheduling parameters for
8887 machine type @var{cpu_type}. You can specify either the @samp{EV}
8888 style name or the corresponding chip number. GCC supports scheduling
8889 parameters for the EV4, EV5 and EV6 family of processors and will
8890 choose the default values for the instruction set from the processor
8891 you specify. If you do not specify a processor type, GCC will default
8892 to the processor on which the compiler was built.
8894 Supported values for @var{cpu_type} are
8900 Schedules as an EV4 and has no instruction set extensions.
8904 Schedules as an EV5 and has no instruction set extensions.
8908 Schedules as an EV5 and supports the BWX extension.
8913 Schedules as an EV5 and supports the BWX and MAX extensions.
8917 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8921 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8924 @item -mtune=@var{cpu_type}
8926 Set only the instruction scheduling parameters for machine type
8927 @var{cpu_type}. The instruction set is not changed.
8929 @item -mmemory-latency=@var{time}
8930 @opindex mmemory-latency
8931 Sets the latency the scheduler should assume for typical memory
8932 references as seen by the application. This number is highly
8933 dependent on the memory access patterns used by the application
8934 and the size of the external cache on the machine.
8936 Valid options for @var{time} are
8940 A decimal number representing clock cycles.
8946 The compiler contains estimates of the number of clock cycles for
8947 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8948 (also called Dcache, Scache, and Bcache), as well as to main memory.
8949 Note that L3 is only valid for EV5.
8954 @node DEC Alpha/VMS Options
8955 @subsection DEC Alpha/VMS Options
8957 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8960 @item -mvms-return-codes
8961 @opindex mvms-return-codes
8962 Return VMS condition codes from main. The default is to return POSIX
8963 style condition (e.g.@ error) codes.
8967 @subsection FRV Options
8974 Only use the first 32 general purpose registers.
8979 Use all 64 general purpose registers.
8984 Use only the first 32 floating point registers.
8989 Use all 64 floating point registers
8992 @opindex mhard-float
8994 Use hardware instructions for floating point operations.
8997 @opindex msoft-float
8999 Use library routines for floating point operations.
9004 Dynamically allocate condition code registers.
9009 Do not try to dynamically allocate condition code registers, only
9010 use @code{icc0} and @code{fcc0}.
9015 Change ABI to use double word insns.
9020 Do not use double word instructions.
9025 Use floating point double instructions.
9030 Do not use floating point double instructions.
9035 Use media instructions.
9040 Do not use media instructions.
9045 Use multiply and add/subtract instructions.
9050 Do not use multiply and add/subtract instructions.
9055 Select the FDPIC ABI, that uses function descriptors to represent
9056 pointers to functions. Without any PIC/PIE-related options, it
9057 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9058 assumes GOT entries and small data are within a 12-bit range from the
9059 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9060 are computed with 32 bits.
9063 @opindex minline-plt
9065 Enable inlining of PLT entries in function calls to functions that are
9066 not known to bind locally. It has no effect without @option{-mfdpic}.
9067 It's enabled by default if optimizing for speed and compiling for
9068 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9069 optimization option such as @option{-O3} or above is present in the
9075 Assume a large TLS segment when generating thread-local code.
9080 Do not assume a large TLS segment when generating thread-local code.
9085 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9086 that is known to be in read-only sections. It's enabled by default,
9087 except for @option{-fpic} or @option{-fpie}: even though it may help
9088 make the global offset table smaller, it trades 1 instruction for 4.
9089 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9090 one of which may be shared by multiple symbols, and it avoids the need
9091 for a GOT entry for the referenced symbol, so it's more likely to be a
9092 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9094 @item -multilib-library-pic
9095 @opindex multilib-library-pic
9097 Link with the (library, not FD) pic libraries. It's implied by
9098 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9099 @option{-fpic} without @option{-mfdpic}. You should never have to use
9105 Follow the EABI requirement of always creating a frame pointer whenever
9106 a stack frame is allocated. This option is enabled by default and can
9107 be disabled with @option{-mno-linked-fp}.
9110 @opindex mlong-calls
9112 Use indirect addressing to call functions outside the current
9113 compilation unit. This allows the functions to be placed anywhere
9114 within the 32-bit address space.
9116 @item -malign-labels
9117 @opindex malign-labels
9119 Try to align labels to an 8-byte boundary by inserting nops into the
9120 previous packet. This option only has an effect when VLIW packing
9121 is enabled. It doesn't create new packets; it merely adds nops to
9125 @opindex mlibrary-pic
9127 Generate position-independent EABI code.
9132 Use only the first four media accumulator registers.
9137 Use all eight media accumulator registers.
9142 Pack VLIW instructions.
9147 Do not pack VLIW instructions.
9152 Do not mark ABI switches in e_flags.
9157 Enable the use of conditional-move instructions (default).
9159 This switch is mainly for debugging the compiler and will likely be removed
9160 in a future version.
9162 @item -mno-cond-move
9163 @opindex mno-cond-move
9165 Disable the use of conditional-move instructions.
9167 This switch is mainly for debugging the compiler and will likely be removed
9168 in a future version.
9173 Enable the use of conditional set instructions (default).
9175 This switch is mainly for debugging the compiler and will likely be removed
9176 in a future version.
9181 Disable the use of conditional set instructions.
9183 This switch is mainly for debugging the compiler and will likely be removed
9184 in a future version.
9189 Enable the use of conditional execution (default).
9191 This switch is mainly for debugging the compiler and will likely be removed
9192 in a future version.
9194 @item -mno-cond-exec
9195 @opindex mno-cond-exec
9197 Disable the use of conditional execution.
9199 This switch is mainly for debugging the compiler and will likely be removed
9200 in a future version.
9203 @opindex mvliw-branch
9205 Run a pass to pack branches into VLIW instructions (default).
9207 This switch is mainly for debugging the compiler and will likely be removed
9208 in a future version.
9210 @item -mno-vliw-branch
9211 @opindex mno-vliw-branch
9213 Do not run a pass to pack branches into VLIW instructions.
9215 This switch is mainly for debugging the compiler and will likely be removed
9216 in a future version.
9218 @item -mmulti-cond-exec
9219 @opindex mmulti-cond-exec
9221 Enable optimization of @code{&&} and @code{||} in conditional execution
9224 This switch is mainly for debugging the compiler and will likely be removed
9225 in a future version.
9227 @item -mno-multi-cond-exec
9228 @opindex mno-multi-cond-exec
9230 Disable optimization of @code{&&} and @code{||} in conditional execution.
9232 This switch is mainly for debugging the compiler and will likely be removed
9233 in a future version.
9235 @item -mnested-cond-exec
9236 @opindex mnested-cond-exec
9238 Enable nested conditional execution optimizations (default).
9240 This switch is mainly for debugging the compiler and will likely be removed
9241 in a future version.
9243 @item -mno-nested-cond-exec
9244 @opindex mno-nested-cond-exec
9246 Disable nested conditional execution optimizations.
9248 This switch is mainly for debugging the compiler and will likely be removed
9249 in a future version.
9251 @item -moptimize-membar
9252 @opindex moptimize-membar
9254 This switch removes redundant @code{membar} instructions from the
9255 compiler generated code. It is enabled by default.
9257 @item -mno-optimize-membar
9258 @opindex mno-optimize-membar
9260 This switch disables the automatic removal of redundant @code{membar}
9261 instructions from the generated code.
9263 @item -mtomcat-stats
9264 @opindex mtomcat-stats
9266 Cause gas to print out tomcat statistics.
9268 @item -mcpu=@var{cpu}
9271 Select the processor type for which to generate code. Possible values are
9272 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9273 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9277 @node GNU/Linux Options
9278 @subsection GNU/Linux Options
9280 These @samp{-m} options are defined for GNU/Linux targets:
9285 Use the GNU C library instead of uClibc. This is the default except
9286 on @samp{*-*-linux-*uclibc*} targets.
9290 Use uClibc instead of the GNU C library. This is the default on
9291 @samp{*-*-linux-*uclibc*} targets.
9294 @node H8/300 Options
9295 @subsection H8/300 Options
9297 These @samp{-m} options are defined for the H8/300 implementations:
9302 Shorten some address references at link time, when possible; uses the
9303 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9304 ld, Using ld}, for a fuller description.
9308 Generate code for the H8/300H@.
9312 Generate code for the H8S@.
9316 Generate code for the H8S and H8/300H in the normal mode. This switch
9317 must be used either with @option{-mh} or @option{-ms}.
9321 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9325 Make @code{int} data 32 bits by default.
9329 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9330 The default for the H8/300H and H8S is to align longs and floats on 4
9332 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9333 This option has no effect on the H8/300.
9337 @subsection HPPA Options
9338 @cindex HPPA Options
9340 These @samp{-m} options are defined for the HPPA family of computers:
9343 @item -march=@var{architecture-type}
9345 Generate code for the specified architecture. The choices for
9346 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9347 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9348 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9349 architecture option for your machine. Code compiled for lower numbered
9350 architectures will run on higher numbered architectures, but not the
9354 @itemx -mpa-risc-1-1
9355 @itemx -mpa-risc-2-0
9356 @opindex mpa-risc-1-0
9357 @opindex mpa-risc-1-1
9358 @opindex mpa-risc-2-0
9359 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9362 @opindex mbig-switch
9363 Generate code suitable for big switch tables. Use this option only if
9364 the assembler/linker complain about out of range branches within a switch
9367 @item -mjump-in-delay
9368 @opindex mjump-in-delay
9369 Fill delay slots of function calls with unconditional jump instructions
9370 by modifying the return pointer for the function call to be the target
9371 of the conditional jump.
9373 @item -mdisable-fpregs
9374 @opindex mdisable-fpregs
9375 Prevent floating point registers from being used in any manner. This is
9376 necessary for compiling kernels which perform lazy context switching of
9377 floating point registers. If you use this option and attempt to perform
9378 floating point operations, the compiler will abort.
9380 @item -mdisable-indexing
9381 @opindex mdisable-indexing
9382 Prevent the compiler from using indexing address modes. This avoids some
9383 rather obscure problems when compiling MIG generated code under MACH@.
9385 @item -mno-space-regs
9386 @opindex mno-space-regs
9387 Generate code that assumes the target has no space registers. This allows
9388 GCC to generate faster indirect calls and use unscaled index address modes.
9390 Such code is suitable for level 0 PA systems and kernels.
9392 @item -mfast-indirect-calls
9393 @opindex mfast-indirect-calls
9394 Generate code that assumes calls never cross space boundaries. This
9395 allows GCC to emit code which performs faster indirect calls.
9397 This option will not work in the presence of shared libraries or nested
9400 @item -mfixed-range=@var{register-range}
9401 @opindex mfixed-range
9402 Generate code treating the given register range as fixed registers.
9403 A fixed register is one that the register allocator can not use. This is
9404 useful when compiling kernel code. A register range is specified as
9405 two registers separated by a dash. Multiple register ranges can be
9406 specified separated by a comma.
9408 @item -mlong-load-store
9409 @opindex mlong-load-store
9410 Generate 3-instruction load and store sequences as sometimes required by
9411 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9414 @item -mportable-runtime
9415 @opindex mportable-runtime
9416 Use the portable calling conventions proposed by HP for ELF systems.
9420 Enable the use of assembler directives only GAS understands.
9422 @item -mschedule=@var{cpu-type}
9424 Schedule code according to the constraints for the machine type
9425 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9426 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9427 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9428 proper scheduling option for your machine. The default scheduling is
9432 @opindex mlinker-opt
9433 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9434 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9435 linkers in which they give bogus error messages when linking some programs.
9438 @opindex msoft-float
9439 Generate output containing library calls for floating point.
9440 @strong{Warning:} the requisite libraries are not available for all HPPA
9441 targets. Normally the facilities of the machine's usual C compiler are
9442 used, but this cannot be done directly in cross-compilation. You must make
9443 your own arrangements to provide suitable library functions for
9444 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9445 does provide software floating point support.
9447 @option{-msoft-float} changes the calling convention in the output file;
9448 therefore, it is only useful if you compile @emph{all} of a program with
9449 this option. In particular, you need to compile @file{libgcc.a}, the
9450 library that comes with GCC, with @option{-msoft-float} in order for
9455 Generate the predefine, @code{_SIO}, for server IO@. The default is
9456 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9457 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9458 options are available under HP-UX and HI-UX@.
9462 Use GNU ld specific options. This passes @option{-shared} to ld when
9463 building a shared library. It is the default when GCC is configured,
9464 explicitly or implicitly, with the GNU linker. This option does not
9465 have any affect on which ld is called, it only changes what parameters
9466 are passed to that ld. The ld that is called is determined by the
9467 @option{--with-ld} configure option, GCC's program search path, and
9468 finally by the user's @env{PATH}. The linker used by GCC can be printed
9469 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9470 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9474 Use HP ld specific options. This passes @option{-b} to ld when building
9475 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9476 links. It is the default when GCC is configured, explicitly or
9477 implicitly, with the HP linker. This option does not have any affect on
9478 which ld is called, it only changes what parameters are passed to that
9479 ld. The ld that is called is determined by the @option{--with-ld}
9480 configure option, GCC's program search path, and finally by the user's
9481 @env{PATH}. The linker used by GCC can be printed using @samp{which
9482 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9483 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9486 @opindex mno-long-calls
9487 Generate code that uses long call sequences. This ensures that a call
9488 is always able to reach linker generated stubs. The default is to generate
9489 long calls only when the distance from the call site to the beginning
9490 of the function or translation unit, as the case may be, exceeds a
9491 predefined limit set by the branch type being used. The limits for
9492 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9493 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9496 Distances are measured from the beginning of functions when using the
9497 @option{-ffunction-sections} option, or when using the @option{-mgas}
9498 and @option{-mno-portable-runtime} options together under HP-UX with
9501 It is normally not desirable to use this option as it will degrade
9502 performance. However, it may be useful in large applications,
9503 particularly when partial linking is used to build the application.
9505 The types of long calls used depends on the capabilities of the
9506 assembler and linker, and the type of code being generated. The
9507 impact on systems that support long absolute calls, and long pic
9508 symbol-difference or pc-relative calls should be relatively small.
9509 However, an indirect call is used on 32-bit ELF systems in pic code
9510 and it is quite long.
9512 @item -munix=@var{unix-std}
9514 Generate compiler predefines and select a startfile for the specified
9515 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9516 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9517 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9518 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9519 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9522 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9523 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9524 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9525 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9526 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9527 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9529 It is @emph{important} to note that this option changes the interfaces
9530 for various library routines. It also affects the operational behavior
9531 of the C library. Thus, @emph{extreme} care is needed in using this
9534 Library code that is intended to operate with more than one UNIX
9535 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9536 as appropriate. Most GNU software doesn't provide this capability.
9540 Suppress the generation of link options to search libdld.sl when the
9541 @option{-static} option is specified on HP-UX 10 and later.
9545 The HP-UX implementation of setlocale in libc has a dependency on
9546 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9547 when the @option{-static} option is specified, special link options
9548 are needed to resolve this dependency.
9550 On HP-UX 10 and later, the GCC driver adds the necessary options to
9551 link with libdld.sl when the @option{-static} option is specified.
9552 This causes the resulting binary to be dynamic. On the 64-bit port,
9553 the linkers generate dynamic binaries by default in any case. The
9554 @option{-nolibdld} option can be used to prevent the GCC driver from
9555 adding these link options.
9559 Add support for multithreading with the @dfn{dce thread} library
9560 under HP-UX@. This option sets flags for both the preprocessor and
9564 @node i386 and x86-64 Options
9565 @subsection Intel 386 and AMD x86-64 Options
9566 @cindex i386 Options
9567 @cindex x86-64 Options
9568 @cindex Intel 386 Options
9569 @cindex AMD x86-64 Options
9571 These @samp{-m} options are defined for the i386 and x86-64 family of
9575 @item -mtune=@var{cpu-type}
9577 Tune to @var{cpu-type} everything applicable about the generated code, except
9578 for the ABI and the set of available instructions. The choices for
9582 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9583 If you know the CPU on which your code will run, then you should use
9584 the corresponding @option{-mtune} option instead of
9585 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9586 of your application will have, then you should use this option.
9588 As new processors are deployed in the marketplace, the behavior of this
9589 option will change. Therefore, if you upgrade to a newer version of
9590 GCC, the code generated option will change to reflect the processors
9591 that were most common when that version of GCC was released.
9593 There is no @option{-march=generic} option because @option{-march}
9594 indicates the instruction set the compiler can use, and there is no
9595 generic instruction set applicable to all processors. In contrast,
9596 @option{-mtune} indicates the processor (or, in this case, collection of
9597 processors) for which the code is optimized.
9599 This selects the CPU to tune for at compilation time by determining
9600 the processor type of the compiling machine. Using @option{-mtune=native}
9601 will produce code optimized for the local machine under the constraints
9602 of the selected instruction set. Using @option{-march=native} will
9603 enable all instruction subsets supported by the local machine (hence
9604 the result might not run on different machines).
9606 Original Intel's i386 CPU@.
9608 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9610 Intel Pentium CPU with no MMX support.
9612 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9614 Intel PentiumPro CPU@.
9616 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9617 instruction set will be used, so the code will run on all i686 family chips.
9619 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9620 @item pentium3, pentium3m
9621 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9624 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9625 support. Used by Centrino notebooks.
9626 @item pentium4, pentium4m
9627 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9629 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9632 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9633 SSE2 and SSE3 instruction set support.
9635 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9636 instruction set support.
9638 AMD K6 CPU with MMX instruction set support.
9640 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9641 @item athlon, athlon-tbird
9642 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9644 @item athlon-4, athlon-xp, athlon-mp
9645 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9646 instruction set support.
9647 @item k8, opteron, athlon64, athlon-fx
9648 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9649 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9651 AMD Family 10 core based CPUs with x86-64 instruction set support. (This
9652 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
9653 instruction set extensions.)
9655 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9658 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9659 instruction set support.
9661 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9662 implemented for this chip.)
9664 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9665 implemented for this chip.)
9667 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9670 While picking a specific @var{cpu-type} will schedule things appropriately
9671 for that particular chip, the compiler will not generate any code that
9672 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9675 @item -march=@var{cpu-type}
9677 Generate instructions for the machine type @var{cpu-type}. The choices
9678 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9679 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9681 @item -mcpu=@var{cpu-type}
9683 A deprecated synonym for @option{-mtune}.
9692 @opindex mpentiumpro
9693 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9694 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9695 These synonyms are deprecated.
9697 @item -mfpmath=@var{unit}
9699 Generate floating point arithmetics for selected unit @var{unit}. The choices
9704 Use the standard 387 floating point coprocessor present majority of chips and
9705 emulated otherwise. Code compiled with this option will run almost everywhere.
9706 The temporary results are computed in 80bit precision instead of precision
9707 specified by the type resulting in slightly different results compared to most
9708 of other chips. See @option{-ffloat-store} for more detailed description.
9710 This is the default choice for i386 compiler.
9713 Use scalar floating point instructions present in the SSE instruction set.
9714 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9715 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9716 instruction set supports only single precision arithmetics, thus the double and
9717 extended precision arithmetics is still done using 387. Later version, present
9718 only in Pentium4 and the future AMD x86-64 chips supports double precision
9721 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9722 or @option{-msse2} switches to enable SSE extensions and make this option
9723 effective. For the x86-64 compiler, these extensions are enabled by default.
9725 The resulting code should be considerably faster in the majority of cases and avoid
9726 the numerical instability problems of 387 code, but may break some existing
9727 code that expects temporaries to be 80bit.
9729 This is the default choice for the x86-64 compiler.
9732 Attempt to utilize both instruction sets at once. This effectively double the
9733 amount of available registers and on chips with separate execution units for
9734 387 and SSE the execution resources too. Use this option with care, as it is
9735 still experimental, because the GCC register allocator does not model separate
9736 functional units well resulting in instable performance.
9739 @item -masm=@var{dialect}
9740 @opindex masm=@var{dialect}
9741 Output asm instructions using selected @var{dialect}. Supported
9742 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9743 not support @samp{intel}.
9748 @opindex mno-ieee-fp
9749 Control whether or not the compiler uses IEEE floating point
9750 comparisons. These handle correctly the case where the result of a
9751 comparison is unordered.
9754 @opindex msoft-float
9755 Generate output containing library calls for floating point.
9756 @strong{Warning:} the requisite libraries are not part of GCC@.
9757 Normally the facilities of the machine's usual C compiler are used, but
9758 this can't be done directly in cross-compilation. You must make your
9759 own arrangements to provide suitable library functions for
9762 On machines where a function returns floating point results in the 80387
9763 register stack, some floating point opcodes may be emitted even if
9764 @option{-msoft-float} is used.
9766 @item -mno-fp-ret-in-387
9767 @opindex mno-fp-ret-in-387
9768 Do not use the FPU registers for return values of functions.
9770 The usual calling convention has functions return values of types
9771 @code{float} and @code{double} in an FPU register, even if there
9772 is no FPU@. The idea is that the operating system should emulate
9775 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9776 in ordinary CPU registers instead.
9778 @item -mno-fancy-math-387
9779 @opindex mno-fancy-math-387
9780 Some 387 emulators do not support the @code{sin}, @code{cos} and
9781 @code{sqrt} instructions for the 387. Specify this option to avoid
9782 generating those instructions. This option is the default on FreeBSD,
9783 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9784 indicates that the target cpu will always have an FPU and so the
9785 instruction will not need emulation. As of revision 2.6.1, these
9786 instructions are not generated unless you also use the
9787 @option{-funsafe-math-optimizations} switch.
9789 @item -malign-double
9790 @itemx -mno-align-double
9791 @opindex malign-double
9792 @opindex mno-align-double
9793 Control whether GCC aligns @code{double}, @code{long double}, and
9794 @code{long long} variables on a two word boundary or a one word
9795 boundary. Aligning @code{double} variables on a two word boundary will
9796 produce code that runs somewhat faster on a @samp{Pentium} at the
9797 expense of more memory.
9799 On x86-64, @option{-malign-double} is enabled by default.
9801 @strong{Warning:} if you use the @option{-malign-double} switch,
9802 structures containing the above types will be aligned differently than
9803 the published application binary interface specifications for the 386
9804 and will not be binary compatible with structures in code compiled
9805 without that switch.
9807 @item -m96bit-long-double
9808 @itemx -m128bit-long-double
9809 @opindex m96bit-long-double
9810 @opindex m128bit-long-double
9811 These switches control the size of @code{long double} type. The i386
9812 application binary interface specifies the size to be 96 bits,
9813 so @option{-m96bit-long-double} is the default in 32 bit mode.
9815 Modern architectures (Pentium and newer) would prefer @code{long double}
9816 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9817 conforming to the ABI, this would not be possible. So specifying a
9818 @option{-m128bit-long-double} will align @code{long double}
9819 to a 16 byte boundary by padding the @code{long double} with an additional
9822 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9823 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9825 Notice that neither of these options enable any extra precision over the x87
9826 standard of 80 bits for a @code{long double}.
9828 @strong{Warning:} if you override the default value for your target ABI, the
9829 structures and arrays containing @code{long double} variables will change
9830 their size as well as function calling convention for function taking
9831 @code{long double} will be modified. Hence they will not be binary
9832 compatible with arrays or structures in code compiled without that switch.
9834 @item -mmlarge-data-threshold=@var{number}
9835 @opindex mlarge-data-threshold=@var{number}
9836 When @option{-mcmodel=medium} is specified, the data greater than
9837 @var{threshold} are placed in large data section. This value must be the
9838 same across all object linked into the binary and defaults to 65535.
9841 @itemx -mno-svr3-shlib
9842 @opindex msvr3-shlib
9843 @opindex mno-svr3-shlib
9844 Control whether GCC places uninitialized local variables into the
9845 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9846 into @code{bss}. These options are meaningful only on System V Release 3.
9850 Use a different function-calling convention, in which functions that
9851 take a fixed number of arguments return with the @code{ret} @var{num}
9852 instruction, which pops their arguments while returning. This saves one
9853 instruction in the caller since there is no need to pop the arguments
9856 You can specify that an individual function is called with this calling
9857 sequence with the function attribute @samp{stdcall}. You can also
9858 override the @option{-mrtd} option by using the function attribute
9859 @samp{cdecl}. @xref{Function Attributes}.
9861 @strong{Warning:} this calling convention is incompatible with the one
9862 normally used on Unix, so you cannot use it if you need to call
9863 libraries compiled with the Unix compiler.
9865 Also, you must provide function prototypes for all functions that
9866 take variable numbers of arguments (including @code{printf});
9867 otherwise incorrect code will be generated for calls to those
9870 In addition, seriously incorrect code will result if you call a
9871 function with too many arguments. (Normally, extra arguments are
9872 harmlessly ignored.)
9874 @item -mregparm=@var{num}
9876 Control how many registers are used to pass integer arguments. By
9877 default, no registers are used to pass arguments, and at most 3
9878 registers can be used. You can control this behavior for a specific
9879 function by using the function attribute @samp{regparm}.
9880 @xref{Function Attributes}.
9882 @strong{Warning:} if you use this switch, and
9883 @var{num} is nonzero, then you must build all modules with the same
9884 value, including any libraries. This includes the system libraries and
9888 @opindex msseregparm
9889 Use SSE register passing conventions for float and double arguments
9890 and return values. You can control this behavior for a specific
9891 function by using the function attribute @samp{sseregparm}.
9892 @xref{Function Attributes}.
9894 @strong{Warning:} if you use this switch then you must build all
9895 modules with the same value, including any libraries. This includes
9896 the system libraries and startup modules.
9898 @item -mstackrealign
9899 @opindex mstackrealign
9900 Realign the stack at entry. On the Intel x86, the
9901 @option{-mstackrealign} option will generate an alternate prologue and
9902 epilogue that realigns the runtime stack. This supports mixing legacy
9903 codes that keep a 4-byte aligned stack with modern codes that keep a
9904 16-byte stack for SSE compatibility. The alternate prologue and
9905 epilogue are slower and bigger than the regular ones, and the
9906 alternate prologue requires an extra scratch register; this lowers the
9907 number of registers available if used in conjunction with the
9908 @code{regparm} attribute. The @option{-mstackrealign} option is
9909 incompatible with the nested function prologue; this is considered a
9910 hard error. See also the attribute @code{force_align_arg_pointer},
9911 applicable to individual functions.
9913 @item -mpreferred-stack-boundary=@var{num}
9914 @opindex mpreferred-stack-boundary
9915 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9916 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9917 the default is 4 (16 bytes or 128 bits).
9919 On Pentium and PentiumPro, @code{double} and @code{long double} values
9920 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9921 suffer significant run time performance penalties. On Pentium III, the
9922 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9923 properly if it is not 16 byte aligned.
9925 To ensure proper alignment of this values on the stack, the stack boundary
9926 must be as aligned as that required by any value stored on the stack.
9927 Further, every function must be generated such that it keeps the stack
9928 aligned. Thus calling a function compiled with a higher preferred
9929 stack boundary from a function compiled with a lower preferred stack
9930 boundary will most likely misalign the stack. It is recommended that
9931 libraries that use callbacks always use the default setting.
9933 This extra alignment does consume extra stack space, and generally
9934 increases code size. Code that is sensitive to stack space usage, such
9935 as embedded systems and operating system kernels, may want to reduce the
9936 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9962 These switches enable or disable the use of instructions in the MMX,
9963 SSE, SSE2, SSE3, SSSE3, SSE4A, ABM or 3DNow! extended instruction sets.
9964 These extensions are also available as built-in functions: see
9965 @ref{X86 Built-in Functions}, for details of the functions enabled and
9966 disabled by these switches.
9968 To have SSE/SSE2 instructions generated automatically from floating-point
9969 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9971 These options will enable GCC to use these extended instructions in
9972 generated code, even without @option{-mfpmath=sse}. Applications which
9973 perform runtime CPU detection must compile separate files for each
9974 supported architecture, using the appropriate flags. In particular,
9975 the file containing the CPU detection code should be compiled without
9979 @itemx -mno-push-args
9981 @opindex mno-push-args
9982 Use PUSH operations to store outgoing parameters. This method is shorter
9983 and usually equally fast as method using SUB/MOV operations and is enabled
9984 by default. In some cases disabling it may improve performance because of
9985 improved scheduling and reduced dependencies.
9987 @item -maccumulate-outgoing-args
9988 @opindex maccumulate-outgoing-args
9989 If enabled, the maximum amount of space required for outgoing arguments will be
9990 computed in the function prologue. This is faster on most modern CPUs
9991 because of reduced dependencies, improved scheduling and reduced stack usage
9992 when preferred stack boundary is not equal to 2. The drawback is a notable
9993 increase in code size. This switch implies @option{-mno-push-args}.
9997 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9998 on thread-safe exception handling must compile and link all code with the
9999 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10000 @option{-D_MT}; when linking, it links in a special thread helper library
10001 @option{-lmingwthrd} which cleans up per thread exception handling data.
10003 @item -mno-align-stringops
10004 @opindex mno-align-stringops
10005 Do not align destination of inlined string operations. This switch reduces
10006 code size and improves performance in case the destination is already aligned,
10007 but GCC doesn't know about it.
10009 @item -minline-all-stringops
10010 @opindex minline-all-stringops
10011 By default GCC inlines string operations only when destination is known to be
10012 aligned at least to 4 byte boundary. This enables more inlining, increase code
10013 size, but may improve performance of code that depends on fast memcpy, strlen
10014 and memset for short lengths.
10016 @item -minline-stringops-dynamically
10017 @opindex minline-stringops-dynamically
10018 For string operation of unknown size, inline runtime checks so for small
10019 blocks inline code is used, while for large blocks library call is used.
10021 @item -mstringop-strategy=@var{alg}
10022 @opindex mstringop-strategy=@var{alg}
10023 Overwrite internal decision heuristic about particular algorithm to inline
10024 string operation with. The allowed values are @code{rep_byte},
10025 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10026 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10027 expanding inline loop, @code{libcall} for always expanding library call.
10029 @item -momit-leaf-frame-pointer
10030 @opindex momit-leaf-frame-pointer
10031 Don't keep the frame pointer in a register for leaf functions. This
10032 avoids the instructions to save, set up and restore frame pointers and
10033 makes an extra register available in leaf functions. The option
10034 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10035 which might make debugging harder.
10037 @item -mtls-direct-seg-refs
10038 @itemx -mno-tls-direct-seg-refs
10039 @opindex mtls-direct-seg-refs
10040 Controls whether TLS variables may be accessed with offsets from the
10041 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10042 or whether the thread base pointer must be added. Whether or not this
10043 is legal depends on the operating system, and whether it maps the
10044 segment to cover the entire TLS area.
10046 For systems that use GNU libc, the default is on.
10049 These @samp{-m} switches are supported in addition to the above
10050 on AMD x86-64 processors in 64-bit environments.
10057 Generate code for a 32-bit or 64-bit environment.
10058 The 32-bit environment sets int, long and pointer to 32 bits and
10059 generates code that runs on any i386 system.
10060 The 64-bit environment sets int to 32 bits and long and pointer
10061 to 64 bits and generates code for AMD's x86-64 architecture.
10063 @item -mno-red-zone
10064 @opindex no-red-zone
10065 Do not use a so called red zone for x86-64 code. The red zone is mandated
10066 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10067 stack pointer that will not be modified by signal or interrupt handlers
10068 and therefore can be used for temporary data without adjusting the stack
10069 pointer. The flag @option{-mno-red-zone} disables this red zone.
10071 @item -mcmodel=small
10072 @opindex mcmodel=small
10073 Generate code for the small code model: the program and its symbols must
10074 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10075 Programs can be statically or dynamically linked. This is the default
10078 @item -mcmodel=kernel
10079 @opindex mcmodel=kernel
10080 Generate code for the kernel code model. The kernel runs in the
10081 negative 2 GB of the address space.
10082 This model has to be used for Linux kernel code.
10084 @item -mcmodel=medium
10085 @opindex mcmodel=medium
10086 Generate code for the medium model: The program is linked in the lower 2
10087 GB of the address space but symbols can be located anywhere in the
10088 address space. Programs can be statically or dynamically linked, but
10089 building of shared libraries are not supported with the medium model.
10091 @item -mcmodel=large
10092 @opindex mcmodel=large
10093 Generate code for the large model: This model makes no assumptions
10094 about addresses and sizes of sections. Currently GCC does not implement
10098 @node IA-64 Options
10099 @subsection IA-64 Options
10100 @cindex IA-64 Options
10102 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10106 @opindex mbig-endian
10107 Generate code for a big endian target. This is the default for HP-UX@.
10109 @item -mlittle-endian
10110 @opindex mlittle-endian
10111 Generate code for a little endian target. This is the default for AIX5
10117 @opindex mno-gnu-as
10118 Generate (or don't) code for the GNU assembler. This is the default.
10119 @c Also, this is the default if the configure option @option{--with-gnu-as}
10125 @opindex mno-gnu-ld
10126 Generate (or don't) code for the GNU linker. This is the default.
10127 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10132 Generate code that does not use a global pointer register. The result
10133 is not position independent code, and violates the IA-64 ABI@.
10135 @item -mvolatile-asm-stop
10136 @itemx -mno-volatile-asm-stop
10137 @opindex mvolatile-asm-stop
10138 @opindex mno-volatile-asm-stop
10139 Generate (or don't) a stop bit immediately before and after volatile asm
10142 @item -mregister-names
10143 @itemx -mno-register-names
10144 @opindex mregister-names
10145 @opindex mno-register-names
10146 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10147 the stacked registers. This may make assembler output more readable.
10153 Disable (or enable) optimizations that use the small data section. This may
10154 be useful for working around optimizer bugs.
10156 @item -mconstant-gp
10157 @opindex mconstant-gp
10158 Generate code that uses a single constant global pointer value. This is
10159 useful when compiling kernel code.
10163 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10164 This is useful when compiling firmware code.
10166 @item -minline-float-divide-min-latency
10167 @opindex minline-float-divide-min-latency
10168 Generate code for inline divides of floating point values
10169 using the minimum latency algorithm.
10171 @item -minline-float-divide-max-throughput
10172 @opindex minline-float-divide-max-throughput
10173 Generate code for inline divides of floating point values
10174 using the maximum throughput algorithm.
10176 @item -minline-int-divide-min-latency
10177 @opindex minline-int-divide-min-latency
10178 Generate code for inline divides of integer values
10179 using the minimum latency algorithm.
10181 @item -minline-int-divide-max-throughput
10182 @opindex minline-int-divide-max-throughput
10183 Generate code for inline divides of integer values
10184 using the maximum throughput algorithm.
10186 @item -minline-sqrt-min-latency
10187 @opindex minline-sqrt-min-latency
10188 Generate code for inline square roots
10189 using the minimum latency algorithm.
10191 @item -minline-sqrt-max-throughput
10192 @opindex minline-sqrt-max-throughput
10193 Generate code for inline square roots
10194 using the maximum throughput algorithm.
10196 @item -mno-dwarf2-asm
10197 @itemx -mdwarf2-asm
10198 @opindex mno-dwarf2-asm
10199 @opindex mdwarf2-asm
10200 Don't (or do) generate assembler code for the DWARF2 line number debugging
10201 info. This may be useful when not using the GNU assembler.
10203 @item -mearly-stop-bits
10204 @itemx -mno-early-stop-bits
10205 @opindex mearly-stop-bits
10206 @opindex mno-early-stop-bits
10207 Allow stop bits to be placed earlier than immediately preceding the
10208 instruction that triggered the stop bit. This can improve instruction
10209 scheduling, but does not always do so.
10211 @item -mfixed-range=@var{register-range}
10212 @opindex mfixed-range
10213 Generate code treating the given register range as fixed registers.
10214 A fixed register is one that the register allocator can not use. This is
10215 useful when compiling kernel code. A register range is specified as
10216 two registers separated by a dash. Multiple register ranges can be
10217 specified separated by a comma.
10219 @item -mtls-size=@var{tls-size}
10221 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
10224 @item -mtune=@var{cpu-type}
10226 Tune the instruction scheduling for a particular CPU, Valid values are
10227 itanium, itanium1, merced, itanium2, and mckinley.
10233 Add support for multithreading using the POSIX threads library. This
10234 option sets flags for both the preprocessor and linker. It does
10235 not affect the thread safety of object code produced by the compiler or
10236 that of libraries supplied with it. These are HP-UX specific flags.
10242 Generate code for a 32-bit or 64-bit environment.
10243 The 32-bit environment sets int, long and pointer to 32 bits.
10244 The 64-bit environment sets int to 32 bits and long and pointer
10245 to 64 bits. These are HP-UX specific flags.
10247 @item -mno-sched-br-data-spec
10248 @itemx -msched-br-data-spec
10249 @opindex -mno-sched-br-data-spec
10250 @opindex -msched-br-data-spec
10251 (Dis/En)able data speculative scheduling before reload.
10252 This will result in generation of the ld.a instructions and
10253 the corresponding check instructions (ld.c / chk.a).
10254 The default is 'disable'.
10256 @item -msched-ar-data-spec
10257 @itemx -mno-sched-ar-data-spec
10258 @opindex -msched-ar-data-spec
10259 @opindex -mno-sched-ar-data-spec
10260 (En/Dis)able data speculative scheduling after reload.
10261 This will result in generation of the ld.a instructions and
10262 the corresponding check instructions (ld.c / chk.a).
10263 The default is 'enable'.
10265 @item -mno-sched-control-spec
10266 @itemx -msched-control-spec
10267 @opindex -mno-sched-control-spec
10268 @opindex -msched-control-spec
10269 (Dis/En)able control speculative scheduling. This feature is
10270 available only during region scheduling (i.e. before reload).
10271 This will result in generation of the ld.s instructions and
10272 the corresponding check instructions chk.s .
10273 The default is 'disable'.
10275 @item -msched-br-in-data-spec
10276 @itemx -mno-sched-br-in-data-spec
10277 @opindex -msched-br-in-data-spec
10278 @opindex -mno-sched-br-in-data-spec
10279 (En/Dis)able speculative scheduling of the instructions that
10280 are dependent on the data speculative loads before reload.
10281 This is effective only with @option{-msched-br-data-spec} enabled.
10282 The default is 'enable'.
10284 @item -msched-ar-in-data-spec
10285 @itemx -mno-sched-ar-in-data-spec
10286 @opindex -msched-ar-in-data-spec
10287 @opindex -mno-sched-ar-in-data-spec
10288 (En/Dis)able speculative scheduling of the instructions that
10289 are dependent on the data speculative loads after reload.
10290 This is effective only with @option{-msched-ar-data-spec} enabled.
10291 The default is 'enable'.
10293 @item -msched-in-control-spec
10294 @itemx -mno-sched-in-control-spec
10295 @opindex -msched-in-control-spec
10296 @opindex -mno-sched-in-control-spec
10297 (En/Dis)able speculative scheduling of the instructions that
10298 are dependent on the control speculative loads.
10299 This is effective only with @option{-msched-control-spec} enabled.
10300 The default is 'enable'.
10303 @itemx -mno-sched-ldc
10304 @opindex -msched-ldc
10305 @opindex -mno-sched-ldc
10306 (En/Dis)able use of simple data speculation checks ld.c .
10307 If disabled, only chk.a instructions will be emitted to check
10308 data speculative loads.
10309 The default is 'enable'.
10311 @item -mno-sched-control-ldc
10312 @itemx -msched-control-ldc
10313 @opindex -mno-sched-control-ldc
10314 @opindex -msched-control-ldc
10315 (Dis/En)able use of ld.c instructions to check control speculative loads.
10316 If enabled, in case of control speculative load with no speculatively
10317 scheduled dependent instructions this load will be emitted as ld.sa and
10318 ld.c will be used to check it.
10319 The default is 'disable'.
10321 @item -mno-sched-spec-verbose
10322 @itemx -msched-spec-verbose
10323 @opindex -mno-sched-spec-verbose
10324 @opindex -msched-spec-verbose
10325 (Dis/En)able printing of the information about speculative motions.
10327 @item -mno-sched-prefer-non-data-spec-insns
10328 @itemx -msched-prefer-non-data-spec-insns
10329 @opindex -mno-sched-prefer-non-data-spec-insns
10330 @opindex -msched-prefer-non-data-spec-insns
10331 If enabled, data speculative instructions will be chosen for schedule
10332 only if there are no other choices at the moment. This will make
10333 the use of the data speculation much more conservative.
10334 The default is 'disable'.
10336 @item -mno-sched-prefer-non-control-spec-insns
10337 @itemx -msched-prefer-non-control-spec-insns
10338 @opindex -mno-sched-prefer-non-control-spec-insns
10339 @opindex -msched-prefer-non-control-spec-insns
10340 If enabled, control speculative instructions will be chosen for schedule
10341 only if there are no other choices at the moment. This will make
10342 the use of the control speculation much more conservative.
10343 The default is 'disable'.
10345 @item -mno-sched-count-spec-in-critical-path
10346 @itemx -msched-count-spec-in-critical-path
10347 @opindex -mno-sched-count-spec-in-critical-path
10348 @opindex -msched-count-spec-in-critical-path
10349 If enabled, speculative dependencies will be considered during
10350 computation of the instructions priorities. This will make the use of the
10351 speculation a bit more conservative.
10352 The default is 'disable'.
10357 @subsection M32C Options
10358 @cindex M32C options
10361 @item -mcpu=@var{name}
10363 Select the CPU for which code is generated. @var{name} may be one of
10364 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10365 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10366 the M32C/80 series.
10370 Specifies that the program will be run on the simulator. This causes
10371 an alternate runtime library to be linked in which supports, for
10372 example, file I/O. You must not use this option when generating
10373 programs that will run on real hardware; you must provide your own
10374 runtime library for whatever I/O functions are needed.
10376 @item -memregs=@var{number}
10378 Specifies the number of memory-based pseudo-registers GCC will use
10379 during code generation. These pseudo-registers will be used like real
10380 registers, so there is a tradeoff between GCC's ability to fit the
10381 code into available registers, and the performance penalty of using
10382 memory instead of registers. Note that all modules in a program must
10383 be compiled with the same value for this option. Because of that, you
10384 must not use this option with the default runtime libraries gcc
10389 @node M32R/D Options
10390 @subsection M32R/D Options
10391 @cindex M32R/D options
10393 These @option{-m} options are defined for Renesas M32R/D architectures:
10398 Generate code for the M32R/2@.
10402 Generate code for the M32R/X@.
10406 Generate code for the M32R@. This is the default.
10408 @item -mmodel=small
10409 @opindex mmodel=small
10410 Assume all objects live in the lower 16MB of memory (so that their addresses
10411 can be loaded with the @code{ld24} instruction), and assume all subroutines
10412 are reachable with the @code{bl} instruction.
10413 This is the default.
10415 The addressability of a particular object can be set with the
10416 @code{model} attribute.
10418 @item -mmodel=medium
10419 @opindex mmodel=medium
10420 Assume objects may be anywhere in the 32-bit address space (the compiler
10421 will generate @code{seth/add3} instructions to load their addresses), and
10422 assume all subroutines are reachable with the @code{bl} instruction.
10424 @item -mmodel=large
10425 @opindex mmodel=large
10426 Assume objects may be anywhere in the 32-bit address space (the compiler
10427 will generate @code{seth/add3} instructions to load their addresses), and
10428 assume subroutines may not be reachable with the @code{bl} instruction
10429 (the compiler will generate the much slower @code{seth/add3/jl}
10430 instruction sequence).
10433 @opindex msdata=none
10434 Disable use of the small data area. Variables will be put into
10435 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10436 @code{section} attribute has been specified).
10437 This is the default.
10439 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10440 Objects may be explicitly put in the small data area with the
10441 @code{section} attribute using one of these sections.
10443 @item -msdata=sdata
10444 @opindex msdata=sdata
10445 Put small global and static data in the small data area, but do not
10446 generate special code to reference them.
10449 @opindex msdata=use
10450 Put small global and static data in the small data area, and generate
10451 special instructions to reference them.
10455 @cindex smaller data references
10456 Put global and static objects less than or equal to @var{num} bytes
10457 into the small data or bss sections instead of the normal data or bss
10458 sections. The default value of @var{num} is 8.
10459 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10460 for this option to have any effect.
10462 All modules should be compiled with the same @option{-G @var{num}} value.
10463 Compiling with different values of @var{num} may or may not work; if it
10464 doesn't the linker will give an error message---incorrect code will not be
10469 Makes the M32R specific code in the compiler display some statistics
10470 that might help in debugging programs.
10472 @item -malign-loops
10473 @opindex malign-loops
10474 Align all loops to a 32-byte boundary.
10476 @item -mno-align-loops
10477 @opindex mno-align-loops
10478 Do not enforce a 32-byte alignment for loops. This is the default.
10480 @item -missue-rate=@var{number}
10481 @opindex missue-rate=@var{number}
10482 Issue @var{number} instructions per cycle. @var{number} can only be 1
10485 @item -mbranch-cost=@var{number}
10486 @opindex mbranch-cost=@var{number}
10487 @var{number} can only be 1 or 2. If it is 1 then branches will be
10488 preferred over conditional code, if it is 2, then the opposite will
10491 @item -mflush-trap=@var{number}
10492 @opindex mflush-trap=@var{number}
10493 Specifies the trap number to use to flush the cache. The default is
10494 12. Valid numbers are between 0 and 15 inclusive.
10496 @item -mno-flush-trap
10497 @opindex mno-flush-trap
10498 Specifies that the cache cannot be flushed by using a trap.
10500 @item -mflush-func=@var{name}
10501 @opindex mflush-func=@var{name}
10502 Specifies the name of the operating system function to call to flush
10503 the cache. The default is @emph{_flush_cache}, but a function call
10504 will only be used if a trap is not available.
10506 @item -mno-flush-func
10507 @opindex mno-flush-func
10508 Indicates that there is no OS function for flushing the cache.
10512 @node M680x0 Options
10513 @subsection M680x0 Options
10514 @cindex M680x0 options
10516 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
10517 The default settings depend on which architecture was selected when
10518 the compiler was configured; the defaults for the most common choices
10522 @item -march=@var{arch}
10524 Generate code for a specific M680x0 or ColdFire instruction set
10525 architecture. Permissible values of @var{arch} for M680x0
10526 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
10527 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
10528 architectures are selected according to Freescale's ISA classification
10529 and the permissible values are: @samp{isaa}, @samp{isaaplus},
10530 @samp{isab} and @samp{isac}.
10532 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
10533 code for a ColdFire target. The @var{arch} in this macro is one of the
10534 @option{-march} arguments given above.
10536 When used together, @option{-march} and @option{-mtune} select code
10537 that runs on a family of similar processors but that is optimized
10538 for a particular microarchitecture.
10540 @item -mcpu=@var{cpu}
10542 Generate code for a specific M680x0 or ColdFire processor.
10543 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
10544 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
10545 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
10546 below, which also classifies the CPUs into families:
10548 @multitable @columnfractions 0.20 0.80
10549 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
10550 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
10551 @item @samp{5206e} @tab @samp{5206e}
10552 @item @samp{5208} @tab @samp{5207} @samp{5208}
10553 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
10554 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
10555 @item @samp{5216} @tab @samp{5214} @samp{5216}
10556 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
10557 @item @samp{5225} @tab @samp{5224} @samp{5225}
10558 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
10559 @item @samp{5249} @tab @samp{5249}
10560 @item @samp{5250} @tab @samp{5250}
10561 @item @samp{5271} @tab @samp{5270} @samp{5271}
10562 @item @samp{5272} @tab @samp{5272}
10563 @item @samp{5275} @tab @samp{5274} @samp{5275}
10564 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
10565 @item @samp{5307} @tab @samp{5307}
10566 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
10567 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
10568 @item @samp{5407} @tab @samp{5407}
10569 @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}
10572 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
10573 @var{arch} is compatible with @var{cpu}. Other combinations of
10574 @option{-mcpu} and @option{-march} are rejected.
10576 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
10577 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
10578 where the value of @var{family} is given by the table above.
10580 @item -mtune=@var{tune}
10582 Tune the code for a particular microarchitecture, within the
10583 constraints set by @option{-march} and @option{-mcpu}.
10584 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
10585 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
10586 and @samp{cpu32}. The ColdFire microarchitectures
10587 are: @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
10589 You can also use @option{-mtune=68020-40} for code that needs
10590 to run relatively well on 68020, 68030 and 68040 targets.
10591 @option{-mtune=68020-60} is similar but includes 68060 targets
10592 as well. These two options select the same tuning decisions as
10593 @option{-m68020-40} and @option{-m68020-60} respectively.
10595 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
10596 when tuning for 680x0 architecture @var{arch}. It also defines
10597 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
10598 option is used. If gcc is tuning for a range of architectures,
10599 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
10600 it defines the macros for every architecture in the range.
10602 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
10603 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
10604 of the arguments given above.
10610 Generate output for a 68000. This is the default
10611 when the compiler is configured for 68000-based systems.
10612 It is equivalent to @option{-march=68000}.
10614 Use this option for microcontrollers with a 68000 or EC000 core,
10615 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10619 Generate output for a 68010. This is the default
10620 when the compiler is configured for 68010-based systems.
10621 It is equivalent to @option{-march=68010}.
10627 Generate output for a 68020. This is the default
10628 when the compiler is configured for 68020-based systems.
10629 It is equivalent to @option{-march=68020}.
10633 Generate output for a 68030. This is the default when the compiler is
10634 configured for 68030-based systems. It is equivalent to
10635 @option{-march=68030}.
10639 Generate output for a 68040. This is the default when the compiler is
10640 configured for 68040-based systems. It is equivalent to
10641 @option{-march=68040}.
10643 This option inhibits the use of 68881/68882 instructions that have to be
10644 emulated by software on the 68040. Use this option if your 68040 does not
10645 have code to emulate those instructions.
10649 Generate output for a 68060. This is the default when the compiler is
10650 configured for 68060-based systems. It is equivalent to
10651 @option{-march=68060}.
10653 This option inhibits the use of 68020 and 68881/68882 instructions that
10654 have to be emulated by software on the 68060. Use this option if your 68060
10655 does not have code to emulate those instructions.
10659 Generate output for a CPU32. This is the default
10660 when the compiler is configured for CPU32-based systems.
10661 It is equivalent to @option{-march=cpu32}.
10663 Use this option for microcontrollers with a
10664 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10665 68336, 68340, 68341, 68349 and 68360.
10669 Generate output for a 520X ColdFire CPU. This is the default
10670 when the compiler is configured for 520X-based systems.
10671 It is equivalent to @option{-mcpu=5206}, and is now deprecated
10672 in favor of that option.
10674 Use this option for microcontroller with a 5200 core, including
10675 the MCF5202, MCF5203, MCF5204 and MCF5206.
10679 Generate output for a 5206e ColdFire CPU. The option is now
10680 deprecated in favor of the equivalent @option{-mcpu=5206e}.
10684 Generate output for a member of the ColdFire 528X family.
10685 The option is now deprecated in favor of the equivalent
10686 @option{-mcpu=528x}.
10690 Generate output for a ColdFire 5307 CPU. The option is now deprecated
10691 in favor of the equivalent @option{-mcpu=5307}.
10695 Generate output for a ColdFire 5407 CPU. The option is now deprecated
10696 in favor of the equivalent @option{-mcpu=5407}.
10700 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
10701 This includes use of hardware floating point instructions.
10702 The option is equivalent to @option{-mcpu=547x}, and is now
10703 deprecated in favor of that option.
10707 Generate output for a 68040, without using any of the new instructions.
10708 This results in code which can run relatively efficiently on either a
10709 68020/68881 or a 68030 or a 68040. The generated code does use the
10710 68881 instructions that are emulated on the 68040.
10712 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
10716 Generate output for a 68060, without using any of the new instructions.
10717 This results in code which can run relatively efficiently on either a
10718 68020/68881 or a 68030 or a 68040. The generated code does use the
10719 68881 instructions that are emulated on the 68060.
10721 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
10725 @opindex mhard-float
10727 Generate floating-point instructions. This is the default for 68020
10728 and above, and for ColdFire devices that have an FPU. It defines the
10729 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
10730 on ColdFire targets.
10733 @opindex msoft-float
10734 Do not generate floating-point instructions; use library calls instead.
10735 This is the default for 68000, 68010, and 68832 targets. It is also
10736 the default for ColdFire devices that have no FPU.
10742 Generate (do not generate) ColdFire hardware divide and remainder
10743 instructions. If @option{-march} is used without @option{-mcpu},
10744 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
10745 architectures. Otherwise, the default is taken from the target CPU
10746 (either the default CPU, or the one specified by @option{-mcpu}). For
10747 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
10748 @option{-mcpu=5206e}.
10750 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
10754 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10755 Additionally, parameters passed on the stack are also aligned to a
10756 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10759 @opindex -mno-short
10760 Do not consider type @code{int} to be 16 bits wide. This is the default.
10763 @itemx -mno-bitfield
10764 @opindex mnobitfield
10765 @opindex mno-bitfield
10766 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10767 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10771 Do use the bit-field instructions. The @option{-m68020} option implies
10772 @option{-mbitfield}. This is the default if you use a configuration
10773 designed for a 68020.
10777 Use a different function-calling convention, in which functions
10778 that take a fixed number of arguments return with the @code{rtd}
10779 instruction, which pops their arguments while returning. This
10780 saves one instruction in the caller since there is no need to pop
10781 the arguments there.
10783 This calling convention is incompatible with the one normally
10784 used on Unix, so you cannot use it if you need to call libraries
10785 compiled with the Unix compiler.
10787 Also, you must provide function prototypes for all functions that
10788 take variable numbers of arguments (including @code{printf});
10789 otherwise incorrect code will be generated for calls to those
10792 In addition, seriously incorrect code will result if you call a
10793 function with too many arguments. (Normally, extra arguments are
10794 harmlessly ignored.)
10796 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10797 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10801 Do not use the calling conventions selected by @option{-mrtd}.
10802 This is the default.
10805 @itemx -mno-align-int
10806 @opindex malign-int
10807 @opindex mno-align-int
10808 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10809 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10810 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10811 Aligning variables on 32-bit boundaries produces code that runs somewhat
10812 faster on processors with 32-bit busses at the expense of more memory.
10814 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10815 align structures containing the above types differently than
10816 most published application binary interface specifications for the m68k.
10820 Use the pc-relative addressing mode of the 68000 directly, instead of
10821 using a global offset table. At present, this option implies @option{-fpic},
10822 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10823 not presently supported with @option{-mpcrel}, though this could be supported for
10824 68020 and higher processors.
10826 @item -mno-strict-align
10827 @itemx -mstrict-align
10828 @opindex mno-strict-align
10829 @opindex mstrict-align
10830 Do not (do) assume that unaligned memory references will be handled by
10834 Generate code that allows the data segment to be located in a different
10835 area of memory from the text segment. This allows for execute in place in
10836 an environment without virtual memory management. This option implies
10839 @item -mno-sep-data
10840 Generate code that assumes that the data segment follows the text segment.
10841 This is the default.
10843 @item -mid-shared-library
10844 Generate code that supports shared libraries via the library ID method.
10845 This allows for execute in place and shared libraries in an environment
10846 without virtual memory management. This option implies @option{-fPIC}.
10848 @item -mno-id-shared-library
10849 Generate code that doesn't assume ID based shared libraries are being used.
10850 This is the default.
10852 @item -mshared-library-id=n
10853 Specified the identification number of the ID based shared library being
10854 compiled. Specifying a value of 0 will generate more compact code, specifying
10855 other values will force the allocation of that number to the current
10856 library but is no more space or time efficient than omitting this option.
10860 @node M68hc1x Options
10861 @subsection M68hc1x Options
10862 @cindex M68hc1x options
10864 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10865 microcontrollers. The default values for these options depends on
10866 which style of microcontroller was selected when the compiler was configured;
10867 the defaults for the most common choices are given below.
10874 Generate output for a 68HC11. This is the default
10875 when the compiler is configured for 68HC11-based systems.
10881 Generate output for a 68HC12. This is the default
10882 when the compiler is configured for 68HC12-based systems.
10888 Generate output for a 68HCS12.
10890 @item -mauto-incdec
10891 @opindex mauto-incdec
10892 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10899 Enable the use of 68HC12 min and max instructions.
10902 @itemx -mno-long-calls
10903 @opindex mlong-calls
10904 @opindex mno-long-calls
10905 Treat all calls as being far away (near). If calls are assumed to be
10906 far away, the compiler will use the @code{call} instruction to
10907 call a function and the @code{rtc} instruction for returning.
10911 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10913 @item -msoft-reg-count=@var{count}
10914 @opindex msoft-reg-count
10915 Specify the number of pseudo-soft registers which are used for the
10916 code generation. The maximum number is 32. Using more pseudo-soft
10917 register may or may not result in better code depending on the program.
10918 The default is 4 for 68HC11 and 2 for 68HC12.
10922 @node MCore Options
10923 @subsection MCore Options
10924 @cindex MCore options
10926 These are the @samp{-m} options defined for the Motorola M*Core
10932 @itemx -mno-hardlit
10934 @opindex mno-hardlit
10935 Inline constants into the code stream if it can be done in two
10936 instructions or less.
10942 Use the divide instruction. (Enabled by default).
10944 @item -mrelax-immediate
10945 @itemx -mno-relax-immediate
10946 @opindex mrelax-immediate
10947 @opindex mno-relax-immediate
10948 Allow arbitrary sized immediates in bit operations.
10950 @item -mwide-bitfields
10951 @itemx -mno-wide-bitfields
10952 @opindex mwide-bitfields
10953 @opindex mno-wide-bitfields
10954 Always treat bit-fields as int-sized.
10956 @item -m4byte-functions
10957 @itemx -mno-4byte-functions
10958 @opindex m4byte-functions
10959 @opindex mno-4byte-functions
10960 Force all functions to be aligned to a four byte boundary.
10962 @item -mcallgraph-data
10963 @itemx -mno-callgraph-data
10964 @opindex mcallgraph-data
10965 @opindex mno-callgraph-data
10966 Emit callgraph information.
10969 @itemx -mno-slow-bytes
10970 @opindex mslow-bytes
10971 @opindex mno-slow-bytes
10972 Prefer word access when reading byte quantities.
10974 @item -mlittle-endian
10975 @itemx -mbig-endian
10976 @opindex mlittle-endian
10977 @opindex mbig-endian
10978 Generate code for a little endian target.
10984 Generate code for the 210 processor.
10988 @subsection MIPS Options
10989 @cindex MIPS options
10995 Generate big-endian code.
10999 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11002 @item -march=@var{arch}
11004 Generate code that will run on @var{arch}, which can be the name of a
11005 generic MIPS ISA, or the name of a particular processor.
11007 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11008 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11009 The processor names are:
11010 @samp{4kc}, @samp{4km}, @samp{4kp},
11011 @samp{4kec}, @samp{4kem}, @samp{4kep},
11012 @samp{5kc}, @samp{5kf},
11014 @samp{24kc}, @samp{24kf}, @samp{24kx},
11015 @samp{24kec}, @samp{24kef}, @samp{24kex},
11016 @samp{34kc}, @samp{34kf}, @samp{34kx},
11019 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11020 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11021 @samp{rm7000}, @samp{rm9000},
11024 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11025 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11026 The special value @samp{from-abi} selects the
11027 most compatible architecture for the selected ABI (that is,
11028 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11030 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11031 (for example, @samp{-march=r2k}). Prefixes are optional, and
11032 @samp{vr} may be written @samp{r}.
11034 GCC defines two macros based on the value of this option. The first
11035 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11036 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11037 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11038 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11039 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11041 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11042 above. In other words, it will have the full prefix and will not
11043 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11044 the macro names the resolved architecture (either @samp{"mips1"} or
11045 @samp{"mips3"}). It names the default architecture when no
11046 @option{-march} option is given.
11048 @item -mtune=@var{arch}
11050 Optimize for @var{arch}. Among other things, this option controls
11051 the way instructions are scheduled, and the perceived cost of arithmetic
11052 operations. The list of @var{arch} values is the same as for
11055 When this option is not used, GCC will optimize for the processor
11056 specified by @option{-march}. By using @option{-march} and
11057 @option{-mtune} together, it is possible to generate code that will
11058 run on a family of processors, but optimize the code for one
11059 particular member of that family.
11061 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11062 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11063 @samp{-march} ones described above.
11067 Equivalent to @samp{-march=mips1}.
11071 Equivalent to @samp{-march=mips2}.
11075 Equivalent to @samp{-march=mips3}.
11079 Equivalent to @samp{-march=mips4}.
11083 Equivalent to @samp{-march=mips32}.
11087 Equivalent to @samp{-march=mips32r2}.
11091 Equivalent to @samp{-march=mips64}.
11096 @opindex mno-mips16
11097 Generate (do not generate) MIPS16 code. If GCC is targetting a
11098 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11110 Generate code for the given ABI@.
11112 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11113 generates 64-bit code when you select a 64-bit architecture, but you
11114 can use @option{-mgp32} to get 32-bit code instead.
11116 For information about the O64 ABI, see
11117 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11119 GCC supports a variant of the o32 ABI in which floating-point registers
11120 are 64 rather than 32 bits wide. You can select this combination with
11121 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11122 and @samp{mfhc1} instructions and is therefore only supported for
11123 MIPS32R2 processors.
11125 The register assignments for arguments and return values remain the
11126 same, but each scalar value is passed in a single 64-bit register
11127 rather than a pair of 32-bit registers. For example, scalar
11128 floating-point values are returned in @samp{$f0} only, not a
11129 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11130 remains the same, but all 64 bits are saved.
11133 @itemx -mno-abicalls
11135 @opindex mno-abicalls
11136 Generate (do not generate) code that is suitable for SVR4-style
11137 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11142 Generate (do not generate) code that is fully position-independent,
11143 and that can therefore be linked into shared libraries. This option
11144 only affects @option{-mabicalls}.
11146 All @option{-mabicalls} code has traditionally been position-independent,
11147 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11148 as an extension, the GNU toolchain allows executables to use absolute
11149 accesses for locally-binding symbols. It can also use shorter GP
11150 initialization sequences and generate direct calls to locally-defined
11151 functions. This mode is selected by @option{-mno-shared}.
11153 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11154 objects that can only be linked by the GNU linker. However, the option
11155 does not affect the ABI of the final executable; it only affects the ABI
11156 of relocatable objects. Using @option{-mno-shared} will generally make
11157 executables both smaller and quicker.
11159 @option{-mshared} is the default.
11165 Lift (do not lift) the usual restrictions on the size of the global
11168 GCC normally uses a single instruction to load values from the GOT@.
11169 While this is relatively efficient, it will only work if the GOT
11170 is smaller than about 64k. Anything larger will cause the linker
11171 to report an error such as:
11173 @cindex relocation truncated to fit (MIPS)
11175 relocation truncated to fit: R_MIPS_GOT16 foobar
11178 If this happens, you should recompile your code with @option{-mxgot}.
11179 It should then work with very large GOTs, although it will also be
11180 less efficient, since it will take three instructions to fetch the
11181 value of a global symbol.
11183 Note that some linkers can create multiple GOTs. If you have such a
11184 linker, you should only need to use @option{-mxgot} when a single object
11185 file accesses more than 64k's worth of GOT entries. Very few do.
11187 These options have no effect unless GCC is generating position
11192 Assume that general-purpose registers are 32 bits wide.
11196 Assume that general-purpose registers are 64 bits wide.
11200 Assume that floating-point registers are 32 bits wide.
11204 Assume that floating-point registers are 64 bits wide.
11207 @opindex mhard-float
11208 Use floating-point coprocessor instructions.
11211 @opindex msoft-float
11212 Do not use floating-point coprocessor instructions. Implement
11213 floating-point calculations using library calls instead.
11215 @item -msingle-float
11216 @opindex msingle-float
11217 Assume that the floating-point coprocessor only supports single-precision
11220 @itemx -mdouble-float
11221 @opindex mdouble-float
11222 Assume that the floating-point coprocessor supports double-precision
11223 operations. This is the default.
11229 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
11231 @itemx -mpaired-single
11232 @itemx -mno-paired-single
11233 @opindex mpaired-single
11234 @opindex mno-paired-single
11235 Use (do not use) paired-single floating-point instructions.
11236 @xref{MIPS Paired-Single Support}. This option can only be used
11237 when generating 64-bit code and requires hardware floating-point
11238 support to be enabled.
11243 @opindex mno-mips3d
11244 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
11245 The option @option{-mips3d} implies @option{-mpaired-single}.
11249 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
11250 an explanation of the default and the way that the pointer size is
11255 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
11257 The default size of @code{int}s, @code{long}s and pointers depends on
11258 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
11259 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
11260 32-bit @code{long}s. Pointers are the same size as @code{long}s,
11261 or the same size as integer registers, whichever is smaller.
11267 Assume (do not assume) that all symbols have 32-bit values, regardless
11268 of the selected ABI@. This option is useful in combination with
11269 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
11270 to generate shorter and faster references to symbolic addresses.
11274 @cindex smaller data references (MIPS)
11275 @cindex gp-relative references (MIPS)
11276 Put global and static items less than or equal to @var{num} bytes into
11277 the small data or bss section instead of the normal data or bss section.
11278 This allows the data to be accessed using a single instruction.
11280 All modules should be compiled with the same @option{-G @var{num}}
11283 @item -membedded-data
11284 @itemx -mno-embedded-data
11285 @opindex membedded-data
11286 @opindex mno-embedded-data
11287 Allocate variables to the read-only data section first if possible, then
11288 next in the small data section if possible, otherwise in data. This gives
11289 slightly slower code than the default, but reduces the amount of RAM required
11290 when executing, and thus may be preferred for some embedded systems.
11292 @item -muninit-const-in-rodata
11293 @itemx -mno-uninit-const-in-rodata
11294 @opindex muninit-const-in-rodata
11295 @opindex mno-uninit-const-in-rodata
11296 Put uninitialized @code{const} variables in the read-only data section.
11297 This option is only meaningful in conjunction with @option{-membedded-data}.
11299 @item -msplit-addresses
11300 @itemx -mno-split-addresses
11301 @opindex msplit-addresses
11302 @opindex mno-split-addresses
11303 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
11304 relocation operators. This option has been superseded by
11305 @option{-mexplicit-relocs} but is retained for backwards compatibility.
11307 @item -mexplicit-relocs
11308 @itemx -mno-explicit-relocs
11309 @opindex mexplicit-relocs
11310 @opindex mno-explicit-relocs
11311 Use (do not use) assembler relocation operators when dealing with symbolic
11312 addresses. The alternative, selected by @option{-mno-explicit-relocs},
11313 is to use assembler macros instead.
11315 @option{-mexplicit-relocs} is the default if GCC was configured
11316 to use an assembler that supports relocation operators.
11318 @item -mcheck-zero-division
11319 @itemx -mno-check-zero-division
11320 @opindex mcheck-zero-division
11321 @opindex mno-check-zero-division
11322 Trap (do not trap) on integer division by zero.
11324 The default is @option{-mcheck-zero-division}.
11326 @item -mdivide-traps
11327 @itemx -mdivide-breaks
11328 @opindex mdivide-traps
11329 @opindex mdivide-breaks
11330 MIPS systems check for division by zero by generating either a
11331 conditional trap or a break instruction. Using traps results in
11332 smaller code, but is only supported on MIPS II and later. Also, some
11333 versions of the Linux kernel have a bug that prevents trap from
11334 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
11335 allow conditional traps on architectures that support them and
11336 @option{-mdivide-breaks} to force the use of breaks.
11338 The default is usually @option{-mdivide-traps}, but this can be
11339 overridden at configure time using @option{--with-divide=breaks}.
11340 Divide-by-zero checks can be completely disabled using
11341 @option{-mno-check-zero-division}.
11346 @opindex mno-memcpy
11347 Force (do not force) the use of @code{memcpy()} for non-trivial block
11348 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
11349 most constant-sized copies.
11352 @itemx -mno-long-calls
11353 @opindex mlong-calls
11354 @opindex mno-long-calls
11355 Disable (do not disable) use of the @code{jal} instruction. Calling
11356 functions using @code{jal} is more efficient but requires the caller
11357 and callee to be in the same 256 megabyte segment.
11359 This option has no effect on abicalls code. The default is
11360 @option{-mno-long-calls}.
11366 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
11367 instructions, as provided by the R4650 ISA@.
11370 @itemx -mno-fused-madd
11371 @opindex mfused-madd
11372 @opindex mno-fused-madd
11373 Enable (disable) use of the floating point multiply-accumulate
11374 instructions, when they are available. The default is
11375 @option{-mfused-madd}.
11377 When multiply-accumulate instructions are used, the intermediate
11378 product is calculated to infinite precision and is not subject to
11379 the FCSR Flush to Zero bit. This may be undesirable in some
11384 Tell the MIPS assembler to not run its preprocessor over user
11385 assembler files (with a @samp{.s} suffix) when assembling them.
11388 @itemx -mno-fix-r4000
11389 @opindex mfix-r4000
11390 @opindex mno-fix-r4000
11391 Work around certain R4000 CPU errata:
11394 A double-word or a variable shift may give an incorrect result if executed
11395 immediately after starting an integer division.
11397 A double-word or a variable shift may give an incorrect result if executed
11398 while an integer multiplication is in progress.
11400 An integer division may give an incorrect result if started in a delay slot
11401 of a taken branch or a jump.
11405 @itemx -mno-fix-r4400
11406 @opindex mfix-r4400
11407 @opindex mno-fix-r4400
11408 Work around certain R4400 CPU errata:
11411 A double-word or a variable shift may give an incorrect result if executed
11412 immediately after starting an integer division.
11416 @itemx -mno-fix-vr4120
11417 @opindex mfix-vr4120
11418 Work around certain VR4120 errata:
11421 @code{dmultu} does not always produce the correct result.
11423 @code{div} and @code{ddiv} do not always produce the correct result if one
11424 of the operands is negative.
11426 The workarounds for the division errata rely on special functions in
11427 @file{libgcc.a}. At present, these functions are only provided by
11428 the @code{mips64vr*-elf} configurations.
11430 Other VR4120 errata require a nop to be inserted between certain pairs of
11431 instructions. These errata are handled by the assembler, not by GCC itself.
11434 @opindex mfix-vr4130
11435 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
11436 workarounds are implemented by the assembler rather than by GCC,
11437 although GCC will avoid using @code{mflo} and @code{mfhi} if the
11438 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
11439 instructions are available instead.
11442 @itemx -mno-fix-sb1
11444 Work around certain SB-1 CPU core errata.
11445 (This flag currently works around the SB-1 revision 2
11446 ``F1'' and ``F2'' floating point errata.)
11448 @item -mflush-func=@var{func}
11449 @itemx -mno-flush-func
11450 @opindex mflush-func
11451 Specifies the function to call to flush the I and D caches, or to not
11452 call any such function. If called, the function must take the same
11453 arguments as the common @code{_flush_func()}, that is, the address of the
11454 memory range for which the cache is being flushed, the size of the
11455 memory range, and the number 3 (to flush both caches). The default
11456 depends on the target GCC was configured for, but commonly is either
11457 @samp{_flush_func} or @samp{__cpu_flush}.
11459 @item -mbranch-likely
11460 @itemx -mno-branch-likely
11461 @opindex mbranch-likely
11462 @opindex mno-branch-likely
11463 Enable or disable use of Branch Likely instructions, regardless of the
11464 default for the selected architecture. By default, Branch Likely
11465 instructions may be generated if they are supported by the selected
11466 architecture. An exception is for the MIPS32 and MIPS64 architectures
11467 and processors which implement those architectures; for those, Branch
11468 Likely instructions will not be generated by default because the MIPS32
11469 and MIPS64 architectures specifically deprecate their use.
11471 @item -mfp-exceptions
11472 @itemx -mno-fp-exceptions
11473 @opindex mfp-exceptions
11474 Specifies whether FP exceptions are enabled. This affects how we schedule
11475 FP instructions for some processors. The default is that FP exceptions are
11478 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11479 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11482 @item -mvr4130-align
11483 @itemx -mno-vr4130-align
11484 @opindex mvr4130-align
11485 The VR4130 pipeline is two-way superscalar, but can only issue two
11486 instructions together if the first one is 8-byte aligned. When this
11487 option is enabled, GCC will align pairs of instructions that it
11488 thinks should execute in parallel.
11490 This option only has an effect when optimizing for the VR4130.
11491 It normally makes code faster, but at the expense of making it bigger.
11492 It is enabled by default at optimization level @option{-O3}.
11496 @subsection MMIX Options
11497 @cindex MMIX Options
11499 These options are defined for the MMIX:
11503 @itemx -mno-libfuncs
11505 @opindex mno-libfuncs
11506 Specify that intrinsic library functions are being compiled, passing all
11507 values in registers, no matter the size.
11510 @itemx -mno-epsilon
11512 @opindex mno-epsilon
11513 Generate floating-point comparison instructions that compare with respect
11514 to the @code{rE} epsilon register.
11516 @item -mabi=mmixware
11518 @opindex mabi-mmixware
11520 Generate code that passes function parameters and return values that (in
11521 the called function) are seen as registers @code{$0} and up, as opposed to
11522 the GNU ABI which uses global registers @code{$231} and up.
11524 @item -mzero-extend
11525 @itemx -mno-zero-extend
11526 @opindex mzero-extend
11527 @opindex mno-zero-extend
11528 When reading data from memory in sizes shorter than 64 bits, use (do not
11529 use) zero-extending load instructions by default, rather than
11530 sign-extending ones.
11533 @itemx -mno-knuthdiv
11535 @opindex mno-knuthdiv
11536 Make the result of a division yielding a remainder have the same sign as
11537 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
11538 remainder follows the sign of the dividend. Both methods are
11539 arithmetically valid, the latter being almost exclusively used.
11541 @item -mtoplevel-symbols
11542 @itemx -mno-toplevel-symbols
11543 @opindex mtoplevel-symbols
11544 @opindex mno-toplevel-symbols
11545 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
11546 code can be used with the @code{PREFIX} assembly directive.
11550 Generate an executable in the ELF format, rather than the default
11551 @samp{mmo} format used by the @command{mmix} simulator.
11553 @item -mbranch-predict
11554 @itemx -mno-branch-predict
11555 @opindex mbranch-predict
11556 @opindex mno-branch-predict
11557 Use (do not use) the probable-branch instructions, when static branch
11558 prediction indicates a probable branch.
11560 @item -mbase-addresses
11561 @itemx -mno-base-addresses
11562 @opindex mbase-addresses
11563 @opindex mno-base-addresses
11564 Generate (do not generate) code that uses @emph{base addresses}. Using a
11565 base address automatically generates a request (handled by the assembler
11566 and the linker) for a constant to be set up in a global register. The
11567 register is used for one or more base address requests within the range 0
11568 to 255 from the value held in the register. The generally leads to short
11569 and fast code, but the number of different data items that can be
11570 addressed is limited. This means that a program that uses lots of static
11571 data may require @option{-mno-base-addresses}.
11573 @item -msingle-exit
11574 @itemx -mno-single-exit
11575 @opindex msingle-exit
11576 @opindex mno-single-exit
11577 Force (do not force) generated code to have a single exit point in each
11581 @node MN10300 Options
11582 @subsection MN10300 Options
11583 @cindex MN10300 options
11585 These @option{-m} options are defined for Matsushita MN10300 architectures:
11590 Generate code to avoid bugs in the multiply instructions for the MN10300
11591 processors. This is the default.
11593 @item -mno-mult-bug
11594 @opindex mno-mult-bug
11595 Do not generate code to avoid bugs in the multiply instructions for the
11596 MN10300 processors.
11600 Generate code which uses features specific to the AM33 processor.
11604 Do not generate code which uses features specific to the AM33 processor. This
11607 @item -mreturn-pointer-on-d0
11608 @opindex mreturn-pointer-on-d0
11609 When generating a function which returns a pointer, return the pointer
11610 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11611 only in a0, and attempts to call such functions without a prototype
11612 would result in errors. Note that this option is on by default; use
11613 @option{-mno-return-pointer-on-d0} to disable it.
11617 Do not link in the C run-time initialization object file.
11621 Indicate to the linker that it should perform a relaxation optimization pass
11622 to shorten branches, calls and absolute memory addresses. This option only
11623 has an effect when used on the command line for the final link step.
11625 This option makes symbolic debugging impossible.
11629 @subsection MT Options
11632 These @option{-m} options are defined for Morpho MT architectures:
11636 @item -march=@var{cpu-type}
11638 Generate code that will run on @var{cpu-type}, which is the name of a system
11639 representing a certain processor type. Possible values for
11640 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11641 @samp{ms1-16-003} and @samp{ms2}.
11643 When this option is not used, the default is @option{-march=ms1-16-002}.
11647 Use byte loads and stores when generating code.
11651 Do not use byte loads and stores when generating code.
11655 Use simulator runtime
11659 Do not link in the C run-time initialization object file
11660 @file{crti.o}. Other run-time initialization and termination files
11661 such as @file{startup.o} and @file{exit.o} are still included on the
11662 linker command line.
11666 @node PDP-11 Options
11667 @subsection PDP-11 Options
11668 @cindex PDP-11 Options
11670 These options are defined for the PDP-11:
11675 Use hardware FPP floating point. This is the default. (FIS floating
11676 point on the PDP-11/40 is not supported.)
11679 @opindex msoft-float
11680 Do not use hardware floating point.
11684 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11688 Return floating-point results in memory. This is the default.
11692 Generate code for a PDP-11/40.
11696 Generate code for a PDP-11/45. This is the default.
11700 Generate code for a PDP-11/10.
11702 @item -mbcopy-builtin
11703 @opindex bcopy-builtin
11704 Use inline @code{movmemhi} patterns for copying memory. This is the
11709 Do not use inline @code{movmemhi} patterns for copying memory.
11715 Use 16-bit @code{int}. This is the default.
11721 Use 32-bit @code{int}.
11724 @itemx -mno-float32
11726 @opindex mno-float32
11727 Use 64-bit @code{float}. This is the default.
11730 @itemx -mno-float64
11732 @opindex mno-float64
11733 Use 32-bit @code{float}.
11737 Use @code{abshi2} pattern. This is the default.
11741 Do not use @code{abshi2} pattern.
11743 @item -mbranch-expensive
11744 @opindex mbranch-expensive
11745 Pretend that branches are expensive. This is for experimenting with
11746 code generation only.
11748 @item -mbranch-cheap
11749 @opindex mbranch-cheap
11750 Do not pretend that branches are expensive. This is the default.
11754 Generate code for a system with split I&D@.
11758 Generate code for a system without split I&D@. This is the default.
11762 Use Unix assembler syntax. This is the default when configured for
11763 @samp{pdp11-*-bsd}.
11767 Use DEC assembler syntax. This is the default when configured for any
11768 PDP-11 target other than @samp{pdp11-*-bsd}.
11771 @node PowerPC Options
11772 @subsection PowerPC Options
11773 @cindex PowerPC options
11775 These are listed under @xref{RS/6000 and PowerPC Options}.
11777 @node RS/6000 and PowerPC Options
11778 @subsection IBM RS/6000 and PowerPC Options
11779 @cindex RS/6000 and PowerPC Options
11780 @cindex IBM RS/6000 and PowerPC Options
11782 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11789 @itemx -mno-powerpc
11790 @itemx -mpowerpc-gpopt
11791 @itemx -mno-powerpc-gpopt
11792 @itemx -mpowerpc-gfxopt
11793 @itemx -mno-powerpc-gfxopt
11795 @itemx -mno-powerpc64
11799 @itemx -mno-popcntb
11807 @opindex mno-power2
11809 @opindex mno-powerpc
11810 @opindex mpowerpc-gpopt
11811 @opindex mno-powerpc-gpopt
11812 @opindex mpowerpc-gfxopt
11813 @opindex mno-powerpc-gfxopt
11814 @opindex mpowerpc64
11815 @opindex mno-powerpc64
11819 @opindex mno-popcntb
11823 @opindex mno-mfpgpr
11824 GCC supports two related instruction set architectures for the
11825 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11826 instructions supported by the @samp{rios} chip set used in the original
11827 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11828 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11829 the IBM 4xx, 6xx, and follow-on microprocessors.
11831 Neither architecture is a subset of the other. However there is a
11832 large common subset of instructions supported by both. An MQ
11833 register is included in processors supporting the POWER architecture.
11835 You use these options to specify which instructions are available on the
11836 processor you are using. The default value of these options is
11837 determined when configuring GCC@. Specifying the
11838 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11839 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11840 rather than the options listed above.
11842 The @option{-mpower} option allows GCC to generate instructions that
11843 are found only in the POWER architecture and to use the MQ register.
11844 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11845 to generate instructions that are present in the POWER2 architecture but
11846 not the original POWER architecture.
11848 The @option{-mpowerpc} option allows GCC to generate instructions that
11849 are found only in the 32-bit subset of the PowerPC architecture.
11850 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11851 GCC to use the optional PowerPC architecture instructions in the
11852 General Purpose group, including floating-point square root. Specifying
11853 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11854 use the optional PowerPC architecture instructions in the Graphics
11855 group, including floating-point select.
11857 The @option{-mmfcrf} option allows GCC to generate the move from
11858 condition register field instruction implemented on the POWER4
11859 processor and other processors that support the PowerPC V2.01
11861 The @option{-mpopcntb} option allows GCC to generate the popcount and
11862 double precision FP reciprocal estimate instruction implemented on the
11863 POWER5 processor and other processors that support the PowerPC V2.02
11865 The @option{-mfprnd} option allows GCC to generate the FP round to
11866 integer instructions implemented on the POWER5+ processor and other
11867 processors that support the PowerPC V2.03 architecture.
11868 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
11869 general purpose register instructions implemented on the POWER6X
11870 processor and other processors that support the extended PowerPC V2.05
11873 The @option{-mpowerpc64} option allows GCC to generate the additional
11874 64-bit instructions that are found in the full PowerPC64 architecture
11875 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11876 @option{-mno-powerpc64}.
11878 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11879 will use only the instructions in the common subset of both
11880 architectures plus some special AIX common-mode calls, and will not use
11881 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11882 permits GCC to use any instruction from either architecture and to
11883 allow use of the MQ register; specify this for the Motorola MPC601.
11885 @item -mnew-mnemonics
11886 @itemx -mold-mnemonics
11887 @opindex mnew-mnemonics
11888 @opindex mold-mnemonics
11889 Select which mnemonics to use in the generated assembler code. With
11890 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11891 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11892 assembler mnemonics defined for the POWER architecture. Instructions
11893 defined in only one architecture have only one mnemonic; GCC uses that
11894 mnemonic irrespective of which of these options is specified.
11896 GCC defaults to the mnemonics appropriate for the architecture in
11897 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11898 value of these option. Unless you are building a cross-compiler, you
11899 should normally not specify either @option{-mnew-mnemonics} or
11900 @option{-mold-mnemonics}, but should instead accept the default.
11902 @item -mcpu=@var{cpu_type}
11904 Set architecture type, register usage, choice of mnemonics, and
11905 instruction scheduling parameters for machine type @var{cpu_type}.
11906 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11907 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11908 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11909 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11910 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11911 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11912 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11913 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11914 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
11915 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11917 @option{-mcpu=common} selects a completely generic processor. Code
11918 generated under this option will run on any POWER or PowerPC processor.
11919 GCC will use only the instructions in the common subset of both
11920 architectures, and will not use the MQ register. GCC assumes a generic
11921 processor model for scheduling purposes.
11923 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11924 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11925 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11926 types, with an appropriate, generic processor model assumed for
11927 scheduling purposes.
11929 The other options specify a specific processor. Code generated under
11930 those options will run best on that processor, and may not run at all on
11933 The @option{-mcpu} options automatically enable or disable the
11936 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
11937 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
11938 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
11940 The particular options set for any particular CPU will vary between
11941 compiler versions, depending on what setting seems to produce optimal
11942 code for that CPU; it doesn't necessarily reflect the actual hardware's
11943 capabilities. If you wish to set an individual option to a particular
11944 value, you may specify it after the @option{-mcpu} option, like
11945 @samp{-mcpu=970 -mno-altivec}.
11947 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11948 not enabled or disabled by the @option{-mcpu} option at present because
11949 AIX does not have full support for these options. You may still
11950 enable or disable them individually if you're sure it'll work in your
11953 @item -mtune=@var{cpu_type}
11955 Set the instruction scheduling parameters for machine type
11956 @var{cpu_type}, but do not set the architecture type, register usage, or
11957 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11958 values for @var{cpu_type} are used for @option{-mtune} as for
11959 @option{-mcpu}. If both are specified, the code generated will use the
11960 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11961 scheduling parameters set by @option{-mtune}.
11967 Generate code to compute division as reciprocal estimate and iterative
11968 refinement, creating opportunities for increased throughput. This
11969 feature requires: optional PowerPC Graphics instruction set for single
11970 precision and FRE instruction for double precision, assuming divides
11971 cannot generate user-visible traps, and the domain values not include
11972 Infinities, denormals or zero denominator.
11975 @itemx -mno-altivec
11977 @opindex mno-altivec
11978 Generate code that uses (does not use) AltiVec instructions, and also
11979 enable the use of built-in functions that allow more direct access to
11980 the AltiVec instruction set. You may also need to set
11981 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11987 @opindex mno-vrsave
11988 Generate VRSAVE instructions when generating AltiVec code.
11991 @opindex msecure-plt
11992 Generate code that allows ld and ld.so to build executables and shared
11993 libraries with non-exec .plt and .got sections. This is a PowerPC
11994 32-bit SYSV ABI option.
11998 Generate code that uses a BSS .plt section that ld.so fills in, and
11999 requires .plt and .got sections that are both writable and executable.
12000 This is a PowerPC 32-bit SYSV ABI option.
12006 This switch enables or disables the generation of ISEL instructions.
12008 @item -misel=@var{yes/no}
12009 This switch has been deprecated. Use @option{-misel} and
12010 @option{-mno-isel} instead.
12016 This switch enables or disables the generation of SPE simd
12019 @item -mspe=@var{yes/no}
12020 This option has been deprecated. Use @option{-mspe} and
12021 @option{-mno-spe} instead.
12023 @item -mfloat-gprs=@var{yes/single/double/no}
12024 @itemx -mfloat-gprs
12025 @opindex mfloat-gprs
12026 This switch enables or disables the generation of floating point
12027 operations on the general purpose registers for architectures that
12030 The argument @var{yes} or @var{single} enables the use of
12031 single-precision floating point operations.
12033 The argument @var{double} enables the use of single and
12034 double-precision floating point operations.
12036 The argument @var{no} disables floating point operations on the
12037 general purpose registers.
12039 This option is currently only available on the MPC854x.
12045 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12046 targets (including GNU/Linux). The 32-bit environment sets int, long
12047 and pointer to 32 bits and generates code that runs on any PowerPC
12048 variant. The 64-bit environment sets int to 32 bits and long and
12049 pointer to 64 bits, and generates code for PowerPC64, as for
12050 @option{-mpowerpc64}.
12053 @itemx -mno-fp-in-toc
12054 @itemx -mno-sum-in-toc
12055 @itemx -mminimal-toc
12057 @opindex mno-fp-in-toc
12058 @opindex mno-sum-in-toc
12059 @opindex mminimal-toc
12060 Modify generation of the TOC (Table Of Contents), which is created for
12061 every executable file. The @option{-mfull-toc} option is selected by
12062 default. In that case, GCC will allocate at least one TOC entry for
12063 each unique non-automatic variable reference in your program. GCC
12064 will also place floating-point constants in the TOC@. However, only
12065 16,384 entries are available in the TOC@.
12067 If you receive a linker error message that saying you have overflowed
12068 the available TOC space, you can reduce the amount of TOC space used
12069 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
12070 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
12071 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
12072 generate code to calculate the sum of an address and a constant at
12073 run-time instead of putting that sum into the TOC@. You may specify one
12074 or both of these options. Each causes GCC to produce very slightly
12075 slower and larger code at the expense of conserving TOC space.
12077 If you still run out of space in the TOC even when you specify both of
12078 these options, specify @option{-mminimal-toc} instead. This option causes
12079 GCC to make only one TOC entry for every file. When you specify this
12080 option, GCC will produce code that is slower and larger but which
12081 uses extremely little TOC space. You may wish to use this option
12082 only on files that contain less frequently executed code.
12088 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
12089 @code{long} type, and the infrastructure needed to support them.
12090 Specifying @option{-maix64} implies @option{-mpowerpc64} and
12091 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
12092 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
12095 @itemx -mno-xl-compat
12096 @opindex mxl-compat
12097 @opindex mno-xl-compat
12098 Produce code that conforms more closely to IBM XL compiler semantics
12099 when using AIX-compatible ABI. Pass floating-point arguments to
12100 prototyped functions beyond the register save area (RSA) on the stack
12101 in addition to argument FPRs. Do not assume that most significant
12102 double in 128-bit long double value is properly rounded when comparing
12103 values and converting to double. Use XL symbol names for long double
12106 The AIX calling convention was extended but not initially documented to
12107 handle an obscure K&R C case of calling a function that takes the
12108 address of its arguments with fewer arguments than declared. IBM XL
12109 compilers access floating point arguments which do not fit in the
12110 RSA from the stack when a subroutine is compiled without
12111 optimization. Because always storing floating-point arguments on the
12112 stack is inefficient and rarely needed, this option is not enabled by
12113 default and only is necessary when calling subroutines compiled by IBM
12114 XL compilers without optimization.
12118 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
12119 application written to use message passing with special startup code to
12120 enable the application to run. The system must have PE installed in the
12121 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
12122 must be overridden with the @option{-specs=} option to specify the
12123 appropriate directory location. The Parallel Environment does not
12124 support threads, so the @option{-mpe} option and the @option{-pthread}
12125 option are incompatible.
12127 @item -malign-natural
12128 @itemx -malign-power
12129 @opindex malign-natural
12130 @opindex malign-power
12131 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12132 @option{-malign-natural} overrides the ABI-defined alignment of larger
12133 types, such as floating-point doubles, on their natural size-based boundary.
12134 The option @option{-malign-power} instructs GCC to follow the ABI-specified
12135 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
12137 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
12141 @itemx -mhard-float
12142 @opindex msoft-float
12143 @opindex mhard-float
12144 Generate code that does not use (uses) the floating-point register set.
12145 Software floating point emulation is provided if you use the
12146 @option{-msoft-float} option, and pass the option to GCC when linking.
12149 @itemx -mno-multiple
12151 @opindex mno-multiple
12152 Generate code that uses (does not use) the load multiple word
12153 instructions and the store multiple word instructions. These
12154 instructions are generated by default on POWER systems, and not
12155 generated on PowerPC systems. Do not use @option{-mmultiple} on little
12156 endian PowerPC systems, since those instructions do not work when the
12157 processor is in little endian mode. The exceptions are PPC740 and
12158 PPC750 which permit the instructions usage in little endian mode.
12163 @opindex mno-string
12164 Generate code that uses (does not use) the load string instructions
12165 and the store string word instructions to save multiple registers and
12166 do small block moves. These instructions are generated by default on
12167 POWER systems, and not generated on PowerPC systems. Do not use
12168 @option{-mstring} on little endian PowerPC systems, since those
12169 instructions do not work when the processor is in little endian mode.
12170 The exceptions are PPC740 and PPC750 which permit the instructions
12171 usage in little endian mode.
12176 @opindex mno-update
12177 Generate code that uses (does not use) the load or store instructions
12178 that update the base register to the address of the calculated memory
12179 location. These instructions are generated by default. If you use
12180 @option{-mno-update}, there is a small window between the time that the
12181 stack pointer is updated and the address of the previous frame is
12182 stored, which means code that walks the stack frame across interrupts or
12183 signals may get corrupted data.
12186 @itemx -mno-fused-madd
12187 @opindex mfused-madd
12188 @opindex mno-fused-madd
12189 Generate code that uses (does not use) the floating point multiply and
12190 accumulate instructions. These instructions are generated by default if
12191 hardware floating is used.
12197 Generate code that uses (does not use) the half-word multiply and
12198 multiply-accumulate instructions on the IBM 405 and 440 processors.
12199 These instructions are generated by default when targetting those
12206 Generate code that uses (does not use) the string-search @samp{dlmzb}
12207 instruction on the IBM 405 and 440 processors. This instruction is
12208 generated by default when targetting those processors.
12210 @item -mno-bit-align
12212 @opindex mno-bit-align
12213 @opindex mbit-align
12214 On System V.4 and embedded PowerPC systems do not (do) force structures
12215 and unions that contain bit-fields to be aligned to the base type of the
12218 For example, by default a structure containing nothing but 8
12219 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
12220 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
12221 the structure would be aligned to a 1 byte boundary and be one byte in
12224 @item -mno-strict-align
12225 @itemx -mstrict-align
12226 @opindex mno-strict-align
12227 @opindex mstrict-align
12228 On System V.4 and embedded PowerPC systems do not (do) assume that
12229 unaligned memory references will be handled by the system.
12231 @item -mrelocatable
12232 @itemx -mno-relocatable
12233 @opindex mrelocatable
12234 @opindex mno-relocatable
12235 On embedded PowerPC systems generate code that allows (does not allow)
12236 the program to be relocated to a different address at runtime. If you
12237 use @option{-mrelocatable} on any module, all objects linked together must
12238 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
12240 @item -mrelocatable-lib
12241 @itemx -mno-relocatable-lib
12242 @opindex mrelocatable-lib
12243 @opindex mno-relocatable-lib
12244 On embedded PowerPC systems generate code that allows (does not allow)
12245 the program to be relocated to a different address at runtime. Modules
12246 compiled with @option{-mrelocatable-lib} can be linked with either modules
12247 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
12248 with modules compiled with the @option{-mrelocatable} options.
12254 On System V.4 and embedded PowerPC systems do not (do) assume that
12255 register 2 contains a pointer to a global area pointing to the addresses
12256 used in the program.
12259 @itemx -mlittle-endian
12261 @opindex mlittle-endian
12262 On System V.4 and embedded PowerPC systems compile code for the
12263 processor in little endian mode. The @option{-mlittle-endian} option is
12264 the same as @option{-mlittle}.
12267 @itemx -mbig-endian
12269 @opindex mbig-endian
12270 On System V.4 and embedded PowerPC systems compile code for the
12271 processor in big endian mode. The @option{-mbig-endian} option is
12272 the same as @option{-mbig}.
12274 @item -mdynamic-no-pic
12275 @opindex mdynamic-no-pic
12276 On Darwin and Mac OS X systems, compile code so that it is not
12277 relocatable, but that its external references are relocatable. The
12278 resulting code is suitable for applications, but not shared
12281 @item -mprioritize-restricted-insns=@var{priority}
12282 @opindex mprioritize-restricted-insns
12283 This option controls the priority that is assigned to
12284 dispatch-slot restricted instructions during the second scheduling
12285 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
12286 @var{no/highest/second-highest} priority to dispatch slot restricted
12289 @item -msched-costly-dep=@var{dependence_type}
12290 @opindex msched-costly-dep
12291 This option controls which dependences are considered costly
12292 by the target during instruction scheduling. The argument
12293 @var{dependence_type} takes one of the following values:
12294 @var{no}: no dependence is costly,
12295 @var{all}: all dependences are costly,
12296 @var{true_store_to_load}: a true dependence from store to load is costly,
12297 @var{store_to_load}: any dependence from store to load is costly,
12298 @var{number}: any dependence which latency >= @var{number} is costly.
12300 @item -minsert-sched-nops=@var{scheme}
12301 @opindex minsert-sched-nops
12302 This option controls which nop insertion scheme will be used during
12303 the second scheduling pass. The argument @var{scheme} takes one of the
12305 @var{no}: Don't insert nops.
12306 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
12307 according to the scheduler's grouping.
12308 @var{regroup_exact}: Insert nops to force costly dependent insns into
12309 separate groups. Insert exactly as many nops as needed to force an insn
12310 to a new group, according to the estimated processor grouping.
12311 @var{number}: Insert nops to force costly dependent insns into
12312 separate groups. Insert @var{number} nops to force an insn to a new group.
12315 @opindex mcall-sysv
12316 On System V.4 and embedded PowerPC systems compile code using calling
12317 conventions that adheres to the March 1995 draft of the System V
12318 Application Binary Interface, PowerPC processor supplement. This is the
12319 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
12321 @item -mcall-sysv-eabi
12322 @opindex mcall-sysv-eabi
12323 Specify both @option{-mcall-sysv} and @option{-meabi} options.
12325 @item -mcall-sysv-noeabi
12326 @opindex mcall-sysv-noeabi
12327 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
12329 @item -mcall-solaris
12330 @opindex mcall-solaris
12331 On System V.4 and embedded PowerPC systems compile code for the Solaris
12335 @opindex mcall-linux
12336 On System V.4 and embedded PowerPC systems compile code for the
12337 Linux-based GNU system.
12341 On System V.4 and embedded PowerPC systems compile code for the
12342 Hurd-based GNU system.
12344 @item -mcall-netbsd
12345 @opindex mcall-netbsd
12346 On System V.4 and embedded PowerPC systems compile code for the
12347 NetBSD operating system.
12349 @item -maix-struct-return
12350 @opindex maix-struct-return
12351 Return all structures in memory (as specified by the AIX ABI)@.
12353 @item -msvr4-struct-return
12354 @opindex msvr4-struct-return
12355 Return structures smaller than 8 bytes in registers (as specified by the
12358 @item -mabi=@var{abi-type}
12360 Extend the current ABI with a particular extension, or remove such extension.
12361 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
12362 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
12366 Extend the current ABI with SPE ABI extensions. This does not change
12367 the default ABI, instead it adds the SPE ABI extensions to the current
12371 @opindex mabi=no-spe
12372 Disable Booke SPE ABI extensions for the current ABI@.
12374 @item -mabi=ibmlongdouble
12375 @opindex mabi=ibmlongdouble
12376 Change the current ABI to use IBM extended precision long double.
12377 This is a PowerPC 32-bit SYSV ABI option.
12379 @item -mabi=ieeelongdouble
12380 @opindex mabi=ieeelongdouble
12381 Change the current ABI to use IEEE extended precision long double.
12382 This is a PowerPC 32-bit Linux ABI option.
12385 @itemx -mno-prototype
12386 @opindex mprototype
12387 @opindex mno-prototype
12388 On System V.4 and embedded PowerPC systems assume that all calls to
12389 variable argument functions are properly prototyped. Otherwise, the
12390 compiler must insert an instruction before every non prototyped call to
12391 set or clear bit 6 of the condition code register (@var{CR}) to
12392 indicate whether floating point values were passed in the floating point
12393 registers in case the function takes a variable arguments. With
12394 @option{-mprototype}, only calls to prototyped variable argument functions
12395 will set or clear the bit.
12399 On embedded PowerPC systems, assume that the startup module is called
12400 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
12401 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
12406 On embedded PowerPC systems, assume that the startup module is called
12407 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
12412 On embedded PowerPC systems, assume that the startup module is called
12413 @file{crt0.o} and the standard C libraries are @file{libads.a} and
12416 @item -myellowknife
12417 @opindex myellowknife
12418 On embedded PowerPC systems, assume that the startup module is called
12419 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
12424 On System V.4 and embedded PowerPC systems, specify that you are
12425 compiling for a VxWorks system.
12429 Specify that you are compiling for the WindISS simulation environment.
12433 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
12434 header to indicate that @samp{eabi} extended relocations are used.
12440 On System V.4 and embedded PowerPC systems do (do not) adhere to the
12441 Embedded Applications Binary Interface (eabi) which is a set of
12442 modifications to the System V.4 specifications. Selecting @option{-meabi}
12443 means that the stack is aligned to an 8 byte boundary, a function
12444 @code{__eabi} is called to from @code{main} to set up the eabi
12445 environment, and the @option{-msdata} option can use both @code{r2} and
12446 @code{r13} to point to two separate small data areas. Selecting
12447 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
12448 do not call an initialization function from @code{main}, and the
12449 @option{-msdata} option will only use @code{r13} to point to a single
12450 small data area. The @option{-meabi} option is on by default if you
12451 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
12454 @opindex msdata=eabi
12455 On System V.4 and embedded PowerPC systems, put small initialized
12456 @code{const} global and static data in the @samp{.sdata2} section, which
12457 is pointed to by register @code{r2}. Put small initialized
12458 non-@code{const} global and static data in the @samp{.sdata} section,
12459 which is pointed to by register @code{r13}. Put small uninitialized
12460 global and static data in the @samp{.sbss} section, which is adjacent to
12461 the @samp{.sdata} section. The @option{-msdata=eabi} option is
12462 incompatible with the @option{-mrelocatable} option. The
12463 @option{-msdata=eabi} option also sets the @option{-memb} option.
12466 @opindex msdata=sysv
12467 On System V.4 and embedded PowerPC systems, put small global and static
12468 data in the @samp{.sdata} section, which is pointed to by register
12469 @code{r13}. Put small uninitialized global and static data in the
12470 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
12471 The @option{-msdata=sysv} option is incompatible with the
12472 @option{-mrelocatable} option.
12474 @item -msdata=default
12476 @opindex msdata=default
12478 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12479 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12480 same as @option{-msdata=sysv}.
12483 @opindex msdata-data
12484 On System V.4 and embedded PowerPC systems, put small global
12485 data in the @samp{.sdata} section. Put small uninitialized global
12486 data in the @samp{.sbss} section. Do not use register @code{r13}
12487 to address small data however. This is the default behavior unless
12488 other @option{-msdata} options are used.
12492 @opindex msdata=none
12494 On embedded PowerPC systems, put all initialized global and static data
12495 in the @samp{.data} section, and all uninitialized data in the
12496 @samp{.bss} section.
12500 @cindex smaller data references (PowerPC)
12501 @cindex .sdata/.sdata2 references (PowerPC)
12502 On embedded PowerPC systems, put global and static items less than or
12503 equal to @var{num} bytes into the small data or bss sections instead of
12504 the normal data or bss section. By default, @var{num} is 8. The
12505 @option{-G @var{num}} switch is also passed to the linker.
12506 All modules should be compiled with the same @option{-G @var{num}} value.
12509 @itemx -mno-regnames
12511 @opindex mno-regnames
12512 On System V.4 and embedded PowerPC systems do (do not) emit register
12513 names in the assembly language output using symbolic forms.
12516 @itemx -mno-longcall
12518 @opindex mno-longcall
12519 By default assume that all calls are far away so that a longer more
12520 expensive calling sequence is required. This is required for calls
12521 further than 32 megabytes (33,554,432 bytes) from the current location.
12522 A short call will be generated if the compiler knows
12523 the call cannot be that far away. This setting can be overridden by
12524 the @code{shortcall} function attribute, or by @code{#pragma
12527 Some linkers are capable of detecting out-of-range calls and generating
12528 glue code on the fly. On these systems, long calls are unnecessary and
12529 generate slower code. As of this writing, the AIX linker can do this,
12530 as can the GNU linker for PowerPC/64. It is planned to add this feature
12531 to the GNU linker for 32-bit PowerPC systems as well.
12533 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
12534 callee, L42'', plus a ``branch island'' (glue code). The two target
12535 addresses represent the callee and the ``branch island''. The
12536 Darwin/PPC linker will prefer the first address and generate a ``bl
12537 callee'' if the PPC ``bl'' instruction will reach the callee directly;
12538 otherwise, the linker will generate ``bl L42'' to call the ``branch
12539 island''. The ``branch island'' is appended to the body of the
12540 calling function; it computes the full 32-bit address of the callee
12543 On Mach-O (Darwin) systems, this option directs the compiler emit to
12544 the glue for every direct call, and the Darwin linker decides whether
12545 to use or discard it.
12547 In the future, we may cause GCC to ignore all longcall specifications
12548 when the linker is known to generate glue.
12552 Adds support for multithreading with the @dfn{pthreads} library.
12553 This option sets flags for both the preprocessor and linker.
12557 @node S/390 and zSeries Options
12558 @subsection S/390 and zSeries Options
12559 @cindex S/390 and zSeries Options
12561 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
12565 @itemx -msoft-float
12566 @opindex mhard-float
12567 @opindex msoft-float
12568 Use (do not use) the hardware floating-point instructions and registers
12569 for floating-point operations. When @option{-msoft-float} is specified,
12570 functions in @file{libgcc.a} will be used to perform floating-point
12571 operations. When @option{-mhard-float} is specified, the compiler
12572 generates IEEE floating-point instructions. This is the default.
12574 @item -mlong-double-64
12575 @itemx -mlong-double-128
12576 @opindex mlong-double-64
12577 @opindex mlong-double-128
12578 These switches control the size of @code{long double} type. A size
12579 of 64bit makes the @code{long double} type equivalent to the @code{double}
12580 type. This is the default.
12583 @itemx -mno-backchain
12584 @opindex mbackchain
12585 @opindex mno-backchain
12586 Store (do not store) the address of the caller's frame as backchain pointer
12587 into the callee's stack frame.
12588 A backchain may be needed to allow debugging using tools that do not understand
12589 DWARF-2 call frame information.
12590 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12591 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12592 the backchain is placed into the topmost word of the 96/160 byte register
12595 In general, code compiled with @option{-mbackchain} is call-compatible with
12596 code compiled with @option{-mmo-backchain}; however, use of the backchain
12597 for debugging purposes usually requires that the whole binary is built with
12598 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12599 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12600 to build a linux kernel use @option{-msoft-float}.
12602 The default is to not maintain the backchain.
12604 @item -mpacked-stack
12605 @item -mno-packed-stack
12606 @opindex mpacked-stack
12607 @opindex mno-packed-stack
12608 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12609 specified, the compiler uses the all fields of the 96/160 byte register save
12610 area only for their default purpose; unused fields still take up stack space.
12611 When @option{-mpacked-stack} is specified, register save slots are densely
12612 packed at the top of the register save area; unused space is reused for other
12613 purposes, allowing for more efficient use of the available stack space.
12614 However, when @option{-mbackchain} is also in effect, the topmost word of
12615 the save area is always used to store the backchain, and the return address
12616 register is always saved two words below the backchain.
12618 As long as the stack frame backchain is not used, code generated with
12619 @option{-mpacked-stack} is call-compatible with code generated with
12620 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12621 S/390 or zSeries generated code that uses the stack frame backchain at run
12622 time, not just for debugging purposes. Such code is not call-compatible
12623 with code compiled with @option{-mpacked-stack}. Also, note that the
12624 combination of @option{-mbackchain},
12625 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12626 to build a linux kernel use @option{-msoft-float}.
12628 The default is to not use the packed stack layout.
12631 @itemx -mno-small-exec
12632 @opindex msmall-exec
12633 @opindex mno-small-exec
12634 Generate (or do not generate) code using the @code{bras} instruction
12635 to do subroutine calls.
12636 This only works reliably if the total executable size does not
12637 exceed 64k. The default is to use the @code{basr} instruction instead,
12638 which does not have this limitation.
12644 When @option{-m31} is specified, generate code compliant to the
12645 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12646 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12647 particular to generate 64-bit instructions. For the @samp{s390}
12648 targets, the default is @option{-m31}, while the @samp{s390x}
12649 targets default to @option{-m64}.
12655 When @option{-mzarch} is specified, generate code using the
12656 instructions available on z/Architecture.
12657 When @option{-mesa} is specified, generate code using the
12658 instructions available on ESA/390. Note that @option{-mesa} is
12659 not possible with @option{-m64}.
12660 When generating code compliant to the GNU/Linux for S/390 ABI,
12661 the default is @option{-mesa}. When generating code compliant
12662 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12668 Generate (or do not generate) code using the @code{mvcle} instruction
12669 to perform block moves. When @option{-mno-mvcle} is specified,
12670 use a @code{mvc} loop instead. This is the default unless optimizing for
12677 Print (or do not print) additional debug information when compiling.
12678 The default is to not print debug information.
12680 @item -march=@var{cpu-type}
12682 Generate code that will run on @var{cpu-type}, which is the name of a system
12683 representing a certain processor type. Possible values for
12684 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12685 When generating code using the instructions available on z/Architecture,
12686 the default is @option{-march=z900}. Otherwise, the default is
12687 @option{-march=g5}.
12689 @item -mtune=@var{cpu-type}
12691 Tune to @var{cpu-type} everything applicable about the generated code,
12692 except for the ABI and the set of available instructions.
12693 The list of @var{cpu-type} values is the same as for @option{-march}.
12694 The default is the value used for @option{-march}.
12697 @itemx -mno-tpf-trace
12698 @opindex mtpf-trace
12699 @opindex mno-tpf-trace
12700 Generate code that adds (does not add) in TPF OS specific branches to trace
12701 routines in the operating system. This option is off by default, even
12702 when compiling for the TPF OS@.
12705 @itemx -mno-fused-madd
12706 @opindex mfused-madd
12707 @opindex mno-fused-madd
12708 Generate code that uses (does not use) the floating point multiply and
12709 accumulate instructions. These instructions are generated by default if
12710 hardware floating point is used.
12712 @item -mwarn-framesize=@var{framesize}
12713 @opindex mwarn-framesize
12714 Emit a warning if the current function exceeds the given frame size. Because
12715 this is a compile time check it doesn't need to be a real problem when the program
12716 runs. It is intended to identify functions which most probably cause
12717 a stack overflow. It is useful to be used in an environment with limited stack
12718 size e.g.@: the linux kernel.
12720 @item -mwarn-dynamicstack
12721 @opindex mwarn-dynamicstack
12722 Emit a warning if the function calls alloca or uses dynamically
12723 sized arrays. This is generally a bad idea with a limited stack size.
12725 @item -mstack-guard=@var{stack-guard}
12726 @item -mstack-size=@var{stack-size}
12727 @opindex mstack-guard
12728 @opindex mstack-size
12729 These arguments always have to be used in conjunction. If they are present the s390
12730 back end emits additional instructions in the function prologue which trigger a trap
12731 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12732 (remember that the stack on s390 grows downward). These options are intended to
12733 be used to help debugging stack overflow problems. The additionally emitted code
12734 causes only little overhead and hence can also be used in production like systems
12735 without greater performance degradation. The given values have to be exact
12736 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12738 In order to be efficient the extra code makes the assumption that the stack starts
12739 at an address aligned to the value given by @var{stack-size}.
12742 @node Score Options
12743 @subsection Score Options
12744 @cindex Score Options
12746 These options are defined for Score implementations:
12751 Compile code for little endian mode.
12755 Compile code for big endian mode. This is the default.
12759 Enable the use of multiply-accumulate instructions. Disabled by default.
12763 Specify the SCORE5U of the target architecture.
12767 Specify the SCORE7 of the target architecture. This is the default.
12771 @subsection SH Options
12773 These @samp{-m} options are defined for the SH implementations:
12778 Generate code for the SH1.
12782 Generate code for the SH2.
12785 Generate code for the SH2e.
12789 Generate code for the SH3.
12793 Generate code for the SH3e.
12797 Generate code for the SH4 without a floating-point unit.
12799 @item -m4-single-only
12800 @opindex m4-single-only
12801 Generate code for the SH4 with a floating-point unit that only
12802 supports single-precision arithmetic.
12806 Generate code for the SH4 assuming the floating-point unit is in
12807 single-precision mode by default.
12811 Generate code for the SH4.
12815 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12816 floating-point unit is not used.
12818 @item -m4a-single-only
12819 @opindex m4a-single-only
12820 Generate code for the SH4a, in such a way that no double-precision
12821 floating point operations are used.
12824 @opindex m4a-single
12825 Generate code for the SH4a assuming the floating-point unit is in
12826 single-precision mode by default.
12830 Generate code for the SH4a.
12834 Same as @option{-m4a-nofpu}, except that it implicitly passes
12835 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12836 instructions at the moment.
12840 Compile code for the processor in big endian mode.
12844 Compile code for the processor in little endian mode.
12848 Align doubles at 64-bit boundaries. Note that this changes the calling
12849 conventions, and thus some functions from the standard C library will
12850 not work unless you recompile it first with @option{-mdalign}.
12854 Shorten some address references at link time, when possible; uses the
12855 linker option @option{-relax}.
12859 Use 32-bit offsets in @code{switch} tables. The default is to use
12864 Enable the use of the instruction @code{fmovd}.
12868 Comply with the calling conventions defined by Renesas.
12872 Comply with the calling conventions defined by Renesas.
12876 Comply with the calling conventions defined for GCC before the Renesas
12877 conventions were available. This option is the default for all
12878 targets of the SH toolchain except for @samp{sh-symbianelf}.
12881 @opindex mnomacsave
12882 Mark the @code{MAC} register as call-clobbered, even if
12883 @option{-mhitachi} is given.
12887 Increase IEEE-compliance of floating-point code.
12888 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12889 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12890 comparisons of NANs / infinities incurs extra overhead in every
12891 floating point comparison, therefore the default is set to
12892 @option{-ffinite-math-only}.
12894 @item -minline-ic_invalidate
12895 @opindex minline-ic_invalidate
12896 Inline code to invalidate instruction cache entries after setting up
12897 nested function trampolines.
12898 This option has no effect if -musermode is in effect and the selected
12899 code generation option (e.g. -m4) does not allow the use of the icbi
12901 If the selected code generation option does not allow the use of the icbi
12902 instruction, and -musermode is not in effect, the inlined code will
12903 manipulate the instruction cache address array directly with an associative
12904 write. This not only requires privileged mode, but it will also
12905 fail if the cache line had been mapped via the TLB and has become unmapped.
12909 Dump instruction size and location in the assembly code.
12912 @opindex mpadstruct
12913 This option is deprecated. It pads structures to multiple of 4 bytes,
12914 which is incompatible with the SH ABI@.
12918 Optimize for space instead of speed. Implied by @option{-Os}.
12921 @opindex mprefergot
12922 When generating position-independent code, emit function calls using
12923 the Global Offset Table instead of the Procedure Linkage Table.
12927 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
12928 if the inlined code would not work in user mode.
12929 This is the default when the target is @code{sh-*-linux*}.
12931 @item -multcost=@var{number}
12932 @opindex multcost=@var{number}
12933 Set the cost to assume for a multiply insn.
12935 @item -mdiv=@var{strategy}
12936 @opindex mdiv=@var{strategy}
12937 Set the division strategy to use for SHmedia code. @var{strategy} must be
12938 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12939 inv:call2, inv:fp .
12940 "fp" performs the operation in floating point. This has a very high latency,
12941 but needs only a few instructions, so it might be a good choice if
12942 your code has enough easily exploitable ILP to allow the compiler to
12943 schedule the floating point instructions together with other instructions.
12944 Division by zero causes a floating point exception.
12945 "inv" uses integer operations to calculate the inverse of the divisor,
12946 and then multiplies the dividend with the inverse. This strategy allows
12947 cse and hoisting of the inverse calculation. Division by zero calculates
12948 an unspecified result, but does not trap.
12949 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12950 have been found, or if the entire operation has been hoisted to the same
12951 place, the last stages of the inverse calculation are intertwined with the
12952 final multiply to reduce the overall latency, at the expense of using a few
12953 more instructions, and thus offering fewer scheduling opportunities with
12955 "call" calls a library function that usually implements the inv:minlat
12957 This gives high code density for m5-*media-nofpu compilations.
12958 "call2" uses a different entry point of the same library function, where it
12959 assumes that a pointer to a lookup table has already been set up, which
12960 exposes the pointer load to cse / code hoisting optimizations.
12961 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12962 code generation, but if the code stays unoptimized, revert to the "call",
12963 "call2", or "fp" strategies, respectively. Note that the
12964 potentially-trapping side effect of division by zero is carried by a
12965 separate instruction, so it is possible that all the integer instructions
12966 are hoisted out, but the marker for the side effect stays where it is.
12967 A recombination to fp operations or a call is not possible in that case.
12968 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12969 that the inverse calculation was nor separated from the multiply, they speed
12970 up division where the dividend fits into 20 bits (plus sign where applicable),
12971 by inserting a test to skip a number of operations in this case; this test
12972 slows down the case of larger dividends. inv20u assumes the case of a such
12973 a small dividend to be unlikely, and inv20l assumes it to be likely.
12975 @item -mdivsi3_libfunc=@var{name}
12976 @opindex mdivsi3_libfunc=@var{name}
12977 Set the name of the library function used for 32 bit signed division to
12978 @var{name}. This only affect the name used in the call and inv:call
12979 division strategies, and the compiler will still expect the same
12980 sets of input/output/clobbered registers as if this option was not present.
12982 @item -madjust-unroll
12983 @opindex madjust-unroll
12984 Throttle unrolling to avoid thrashing target registers.
12985 This option only has an effect if the gcc code base supports the
12986 TARGET_ADJUST_UNROLL_MAX target hook.
12988 @item -mindexed-addressing
12989 @opindex mindexed-addressing
12990 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12991 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12992 semantics for the indexed addressing mode. The architecture allows the
12993 implementation of processors with 64 bit MMU, which the OS could use to
12994 get 32 bit addressing, but since no current hardware implementation supports
12995 this or any other way to make the indexed addressing mode safe to use in
12996 the 32 bit ABI, the default is -mno-indexed-addressing.
12998 @item -mgettrcost=@var{number}
12999 @opindex mgettrcost=@var{number}
13000 Set the cost assumed for the gettr instruction to @var{number}.
13001 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
13005 Assume pt* instructions won't trap. This will generally generate better
13006 scheduled code, but is unsafe on current hardware. The current architecture
13007 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
13008 This has the unintentional effect of making it unsafe to schedule ptabs /
13009 ptrel before a branch, or hoist it out of a loop. For example,
13010 __do_global_ctors, a part of libgcc that runs constructors at program
13011 startup, calls functions in a list which is delimited by @minus{}1. With the
13012 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
13013 That means that all the constructors will be run a bit quicker, but when
13014 the loop comes to the end of the list, the program crashes because ptabs
13015 loads @minus{}1 into a target register. Since this option is unsafe for any
13016 hardware implementing the current architecture specification, the default
13017 is -mno-pt-fixed. Unless the user specifies a specific cost with
13018 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
13019 this deters register allocation using target registers for storing
13022 @item -minvalid-symbols
13023 @opindex minvalid-symbols
13024 Assume symbols might be invalid. Ordinary function symbols generated by
13025 the compiler will always be valid to load with movi/shori/ptabs or
13026 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
13027 to generate symbols that will cause ptabs / ptrel to trap.
13028 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
13029 It will then prevent cross-basic-block cse, hoisting and most scheduling
13030 of symbol loads. The default is @option{-mno-invalid-symbols}.
13033 @node SPARC Options
13034 @subsection SPARC Options
13035 @cindex SPARC options
13037 These @samp{-m} options are supported on the SPARC:
13040 @item -mno-app-regs
13042 @opindex mno-app-regs
13044 Specify @option{-mapp-regs} to generate output using the global registers
13045 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
13048 To be fully SVR4 ABI compliant at the cost of some performance loss,
13049 specify @option{-mno-app-regs}. You should compile libraries and system
13050 software with this option.
13053 @itemx -mhard-float
13055 @opindex mhard-float
13056 Generate output containing floating point instructions. This is the
13060 @itemx -msoft-float
13062 @opindex msoft-float
13063 Generate output containing library calls for floating point.
13064 @strong{Warning:} the requisite libraries are not available for all SPARC
13065 targets. Normally the facilities of the machine's usual C compiler are
13066 used, but this cannot be done directly in cross-compilation. You must make
13067 your own arrangements to provide suitable library functions for
13068 cross-compilation. The embedded targets @samp{sparc-*-aout} and
13069 @samp{sparclite-*-*} do provide software floating point support.
13071 @option{-msoft-float} changes the calling convention in the output file;
13072 therefore, it is only useful if you compile @emph{all} of a program with
13073 this option. In particular, you need to compile @file{libgcc.a}, the
13074 library that comes with GCC, with @option{-msoft-float} in order for
13077 @item -mhard-quad-float
13078 @opindex mhard-quad-float
13079 Generate output containing quad-word (long double) floating point
13082 @item -msoft-quad-float
13083 @opindex msoft-quad-float
13084 Generate output containing library calls for quad-word (long double)
13085 floating point instructions. The functions called are those specified
13086 in the SPARC ABI@. This is the default.
13088 As of this writing, there are no SPARC implementations that have hardware
13089 support for the quad-word floating point instructions. They all invoke
13090 a trap handler for one of these instructions, and then the trap handler
13091 emulates the effect of the instruction. Because of the trap handler overhead,
13092 this is much slower than calling the ABI library routines. Thus the
13093 @option{-msoft-quad-float} option is the default.
13095 @item -mno-unaligned-doubles
13096 @itemx -munaligned-doubles
13097 @opindex mno-unaligned-doubles
13098 @opindex munaligned-doubles
13099 Assume that doubles have 8 byte alignment. This is the default.
13101 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
13102 alignment only if they are contained in another type, or if they have an
13103 absolute address. Otherwise, it assumes they have 4 byte alignment.
13104 Specifying this option avoids some rare compatibility problems with code
13105 generated by other compilers. It is not the default because it results
13106 in a performance loss, especially for floating point code.
13108 @item -mno-faster-structs
13109 @itemx -mfaster-structs
13110 @opindex mno-faster-structs
13111 @opindex mfaster-structs
13112 With @option{-mfaster-structs}, the compiler assumes that structures
13113 should have 8 byte alignment. This enables the use of pairs of
13114 @code{ldd} and @code{std} instructions for copies in structure
13115 assignment, in place of twice as many @code{ld} and @code{st} pairs.
13116 However, the use of this changed alignment directly violates the SPARC
13117 ABI@. Thus, it's intended only for use on targets where the developer
13118 acknowledges that their resulting code will not be directly in line with
13119 the rules of the ABI@.
13121 @item -mimpure-text
13122 @opindex mimpure-text
13123 @option{-mimpure-text}, used in addition to @option{-shared}, tells
13124 the compiler to not pass @option{-z text} to the linker when linking a
13125 shared object. Using this option, you can link position-dependent
13126 code into a shared object.
13128 @option{-mimpure-text} suppresses the ``relocations remain against
13129 allocatable but non-writable sections'' linker error message.
13130 However, the necessary relocations will trigger copy-on-write, and the
13131 shared object is not actually shared across processes. Instead of
13132 using @option{-mimpure-text}, you should compile all source code with
13133 @option{-fpic} or @option{-fPIC}.
13135 This option is only available on SunOS and Solaris.
13137 @item -mcpu=@var{cpu_type}
13139 Set the instruction set, register set, and instruction scheduling parameters
13140 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
13141 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
13142 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
13143 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
13144 @samp{ultrasparc3}, and @samp{niagara}.
13146 Default instruction scheduling parameters are used for values that select
13147 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
13148 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
13150 Here is a list of each supported architecture and their supported
13155 v8: supersparc, hypersparc
13156 sparclite: f930, f934, sparclite86x
13158 v9: ultrasparc, ultrasparc3, niagara
13161 By default (unless configured otherwise), GCC generates code for the V7
13162 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
13163 additionally optimizes it for the Cypress CY7C602 chip, as used in the
13164 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
13165 SPARCStation 1, 2, IPX etc.
13167 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
13168 architecture. The only difference from V7 code is that the compiler emits
13169 the integer multiply and integer divide instructions which exist in SPARC-V8
13170 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
13171 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
13174 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
13175 the SPARC architecture. This adds the integer multiply, integer divide step
13176 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
13177 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
13178 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
13179 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
13180 MB86934 chip, which is the more recent SPARClite with FPU@.
13182 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
13183 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
13184 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
13185 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
13186 optimizes it for the TEMIC SPARClet chip.
13188 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
13189 architecture. This adds 64-bit integer and floating-point move instructions,
13190 3 additional floating-point condition code registers and conditional move
13191 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
13192 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
13193 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
13194 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
13195 @option{-mcpu=niagara}, the compiler additionally optimizes it for
13196 Sun UltraSPARC T1 chips.
13198 @item -mtune=@var{cpu_type}
13200 Set the instruction scheduling parameters for machine type
13201 @var{cpu_type}, but do not set the instruction set or register set that the
13202 option @option{-mcpu=@var{cpu_type}} would.
13204 The same values for @option{-mcpu=@var{cpu_type}} can be used for
13205 @option{-mtune=@var{cpu_type}}, but the only useful values are those
13206 that select a particular cpu implementation. Those are @samp{cypress},
13207 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
13208 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
13209 @samp{ultrasparc3}, and @samp{niagara}.
13214 @opindex mno-v8plus
13215 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
13216 difference from the V8 ABI is that the global and out registers are
13217 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
13218 mode for all SPARC-V9 processors.
13224 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
13225 Visual Instruction Set extensions. The default is @option{-mno-vis}.
13228 These @samp{-m} options are supported in addition to the above
13229 on SPARC-V9 processors in 64-bit environments:
13232 @item -mlittle-endian
13233 @opindex mlittle-endian
13234 Generate code for a processor running in little-endian mode. It is only
13235 available for a few configurations and most notably not on Solaris and Linux.
13241 Generate code for a 32-bit or 64-bit environment.
13242 The 32-bit environment sets int, long and pointer to 32 bits.
13243 The 64-bit environment sets int to 32 bits and long and pointer
13246 @item -mcmodel=medlow
13247 @opindex mcmodel=medlow
13248 Generate code for the Medium/Low code model: 64-bit addresses, programs
13249 must be linked in the low 32 bits of memory. Programs can be statically
13250 or dynamically linked.
13252 @item -mcmodel=medmid
13253 @opindex mcmodel=medmid
13254 Generate code for the Medium/Middle code model: 64-bit addresses, programs
13255 must be linked in the low 44 bits of memory, the text and data segments must
13256 be less than 2GB in size and the data segment must be located within 2GB of
13259 @item -mcmodel=medany
13260 @opindex mcmodel=medany
13261 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
13262 may be linked anywhere in memory, the text and data segments must be less
13263 than 2GB in size and the data segment must be located within 2GB of the
13266 @item -mcmodel=embmedany
13267 @opindex mcmodel=embmedany
13268 Generate code for the Medium/Anywhere code model for embedded systems:
13269 64-bit addresses, the text and data segments must be less than 2GB in
13270 size, both starting anywhere in memory (determined at link time). The
13271 global register %g4 points to the base of the data segment. Programs
13272 are statically linked and PIC is not supported.
13275 @itemx -mno-stack-bias
13276 @opindex mstack-bias
13277 @opindex mno-stack-bias
13278 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
13279 frame pointer if present, are offset by @minus{}2047 which must be added back
13280 when making stack frame references. This is the default in 64-bit mode.
13281 Otherwise, assume no such offset is present.
13284 These switches are supported in addition to the above on Solaris:
13289 Add support for multithreading using the Solaris threads library. This
13290 option sets flags for both the preprocessor and linker. This option does
13291 not affect the thread safety of object code produced by the compiler or
13292 that of libraries supplied with it.
13296 Add support for multithreading using the POSIX threads library. This
13297 option sets flags for both the preprocessor and linker. This option does
13298 not affect the thread safety of object code produced by the compiler or
13299 that of libraries supplied with it.
13303 This is a synonym for @option{-pthreads}.
13307 @subsection SPU Options
13308 @cindex SPU options
13310 These @samp{-m} options are supported on the SPU:
13314 @itemx -merror-reloc
13315 @opindex mwarn-reloc
13316 @opindex merror-reloc
13318 The loader for SPU does not handle dynamic relocations. By default, GCC
13319 will give an error when it generates code that requires a dynamic
13320 relocation. @option{-mno-error-reloc} disables the error,
13321 @option{-mwarn-reloc} will generate a warning instead.
13324 @itemx -munsafe-dma
13326 @opindex munsafe-dma
13328 Instructions which initiate or test completion of DMA must not be
13329 reordered with respect to loads and stores of the memory which is being
13330 accessed. Users typically address this problem using the volatile
13331 keyword, but that can lead to inefficient code in places where the
13332 memory is known to not change. Rather than mark the memory as volatile
13333 we treat the DMA instructions as potentially effecting all memory. With
13334 @option{-munsafe-dma} users must use the volatile keyword to protect
13337 @item -mbranch-hints
13338 @opindex mbranch-hints
13340 By default, GCC will generate a branch hint instruction to avoid
13341 pipeline stalls for always taken or probably taken branches. A hint
13342 will not be generated closer than 8 instructions away from its branch.
13343 There is little reason to disable them, except for debugging purposes,
13344 or to make an object a little bit smaller.
13348 @opindex msmall-mem
13349 @opindex mlarge-mem
13351 By default, GCC generates code assuming that addresses are never larger
13352 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
13353 a full 32 bit address.
13358 By default, GCC links against startup code that assumes the SPU-style
13359 main function interface (which has an unconventional parameter list).
13360 With @option{-mstdmain}, GCC will link your program against startup
13361 code that assumes a C99-style interface to @code{main}, including a
13362 local copy of @code{argv} strings.
13364 @item -mfixed-range=@var{register-range}
13365 @opindex mfixed-range
13366 Generate code treating the given register range as fixed registers.
13367 A fixed register is one that the register allocator can not use. This is
13368 useful when compiling kernel code. A register range is specified as
13369 two registers separated by a dash. Multiple register ranges can be
13370 specified separated by a comma.
13374 @node System V Options
13375 @subsection Options for System V
13377 These additional options are available on System V Release 4 for
13378 compatibility with other compilers on those systems:
13383 Create a shared object.
13384 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
13388 Identify the versions of each tool used by the compiler, in a
13389 @code{.ident} assembler directive in the output.
13393 Refrain from adding @code{.ident} directives to the output file (this is
13396 @item -YP,@var{dirs}
13398 Search the directories @var{dirs}, and no others, for libraries
13399 specified with @option{-l}.
13401 @item -Ym,@var{dir}
13403 Look in the directory @var{dir} to find the M4 preprocessor.
13404 The assembler uses this option.
13405 @c This is supposed to go with a -Yd for predefined M4 macro files, but
13406 @c the generic assembler that comes with Solaris takes just -Ym.
13409 @node TMS320C3x/C4x Options
13410 @subsection TMS320C3x/C4x Options
13411 @cindex TMS320C3x/C4x Options
13413 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
13417 @item -mcpu=@var{cpu_type}
13419 Set the instruction set, register set, and instruction scheduling
13420 parameters for machine type @var{cpu_type}. Supported values for
13421 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
13422 @samp{c44}. The default is @samp{c40} to generate code for the
13427 @itemx -msmall-memory
13429 @opindex mbig-memory
13431 @opindex msmall-memory
13433 Generates code for the big or small memory model. The small memory
13434 model assumed that all data fits into one 64K word page. At run-time
13435 the data page (DP) register must be set to point to the 64K page
13436 containing the .bss and .data program sections. The big memory model is
13437 the default and requires reloading of the DP register for every direct
13444 Allow (disallow) allocation of general integer operands into the block
13445 count register BK@.
13451 Enable (disable) generation of code using decrement and branch,
13452 DBcond(D), instructions. This is enabled by default for the C4x. To be
13453 on the safe side, this is disabled for the C3x, since the maximum
13454 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
13455 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
13456 that it can utilize the decrement and branch instruction, but will give
13457 up if there is more than one memory reference in the loop. Thus a loop
13458 where the loop counter is decremented can generate slightly more
13459 efficient code, in cases where the RPTB instruction cannot be utilized.
13461 @item -mdp-isr-reload
13463 @opindex mdp-isr-reload
13465 Force the DP register to be saved on entry to an interrupt service
13466 routine (ISR), reloaded to point to the data section, and restored on
13467 exit from the ISR@. This should not be required unless someone has
13468 violated the small memory model by modifying the DP register, say within
13475 For the C3x use the 24-bit MPYI instruction for integer multiplies
13476 instead of a library call to guarantee 32-bit results. Note that if one
13477 of the operands is a constant, then the multiplication will be performed
13478 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
13479 then squaring operations are performed inline instead of a library call.
13482 @itemx -mno-fast-fix
13484 @opindex mno-fast-fix
13485 The C3x/C4x FIX instruction to convert a floating point value to an
13486 integer value chooses the nearest integer less than or equal to the
13487 floating point value rather than to the nearest integer. Thus if the
13488 floating point number is negative, the result will be incorrectly
13489 truncated an additional code is necessary to detect and correct this
13490 case. This option can be used to disable generation of the additional
13491 code required to correct the result.
13497 Enable (disable) generation of repeat block sequences using the RPTB
13498 instruction for zero overhead looping. The RPTB construct is only used
13499 for innermost loops that do not call functions or jump across the loop
13500 boundaries. There is no advantage having nested RPTB loops due to the
13501 overhead required to save and restore the RC, RS, and RE registers.
13502 This is enabled by default with @option{-O2}.
13504 @item -mrpts=@var{count}
13508 Enable (disable) the use of the single instruction repeat instruction
13509 RPTS@. If a repeat block contains a single instruction, and the loop
13510 count can be guaranteed to be less than the value @var{count}, GCC will
13511 emit a RPTS instruction instead of a RPTB@. If no value is specified,
13512 then a RPTS will be emitted even if the loop count cannot be determined
13513 at compile time. Note that the repeated instruction following RPTS does
13514 not have to be reloaded from memory each iteration, thus freeing up the
13515 CPU buses for operands. However, since interrupts are blocked by this
13516 instruction, it is disabled by default.
13518 @item -mloop-unsigned
13519 @itemx -mno-loop-unsigned
13520 @opindex mloop-unsigned
13521 @opindex mno-loop-unsigned
13522 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
13523 is @math{2^{31} + 1} since these instructions test if the iteration count is
13524 negative to terminate the loop. If the iteration count is unsigned
13525 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
13526 exceeded. This switch allows an unsigned iteration count.
13530 Try to emit an assembler syntax that the TI assembler (asm30) is happy
13531 with. This also enforces compatibility with the API employed by the TI
13532 C3x C compiler. For example, long doubles are passed as structures
13533 rather than in floating point registers.
13539 Generate code that uses registers (stack) for passing arguments to functions.
13540 By default, arguments are passed in registers where possible rather
13541 than by pushing arguments on to the stack.
13543 @item -mparallel-insns
13544 @itemx -mno-parallel-insns
13545 @opindex mparallel-insns
13546 @opindex mno-parallel-insns
13547 Allow the generation of parallel instructions. This is enabled by
13548 default with @option{-O2}.
13550 @item -mparallel-mpy
13551 @itemx -mno-parallel-mpy
13552 @opindex mparallel-mpy
13553 @opindex mno-parallel-mpy
13554 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
13555 provided @option{-mparallel-insns} is also specified. These instructions have
13556 tight register constraints which can pessimize the code generation
13557 of large functions.
13562 @subsection V850 Options
13563 @cindex V850 Options
13565 These @samp{-m} options are defined for V850 implementations:
13569 @itemx -mno-long-calls
13570 @opindex mlong-calls
13571 @opindex mno-long-calls
13572 Treat all calls as being far away (near). If calls are assumed to be
13573 far away, the compiler will always load the functions address up into a
13574 register, and call indirect through the pointer.
13580 Do not optimize (do optimize) basic blocks that use the same index
13581 pointer 4 or more times to copy pointer into the @code{ep} register, and
13582 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
13583 option is on by default if you optimize.
13585 @item -mno-prolog-function
13586 @itemx -mprolog-function
13587 @opindex mno-prolog-function
13588 @opindex mprolog-function
13589 Do not use (do use) external functions to save and restore registers
13590 at the prologue and epilogue of a function. The external functions
13591 are slower, but use less code space if more than one function saves
13592 the same number of registers. The @option{-mprolog-function} option
13593 is on by default if you optimize.
13597 Try to make the code as small as possible. At present, this just turns
13598 on the @option{-mep} and @option{-mprolog-function} options.
13600 @item -mtda=@var{n}
13602 Put static or global variables whose size is @var{n} bytes or less into
13603 the tiny data area that register @code{ep} points to. The tiny data
13604 area can hold up to 256 bytes in total (128 bytes for byte references).
13606 @item -msda=@var{n}
13608 Put static or global variables whose size is @var{n} bytes or less into
13609 the small data area that register @code{gp} points to. The small data
13610 area can hold up to 64 kilobytes.
13612 @item -mzda=@var{n}
13614 Put static or global variables whose size is @var{n} bytes or less into
13615 the first 32 kilobytes of memory.
13619 Specify that the target processor is the V850.
13622 @opindex mbig-switch
13623 Generate code suitable for big switch tables. Use this option only if
13624 the assembler/linker complain about out of range branches within a switch
13629 This option will cause r2 and r5 to be used in the code generated by
13630 the compiler. This setting is the default.
13632 @item -mno-app-regs
13633 @opindex mno-app-regs
13634 This option will cause r2 and r5 to be treated as fixed registers.
13638 Specify that the target processor is the V850E1. The preprocessor
13639 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13640 this option is used.
13644 Specify that the target processor is the V850E@. The preprocessor
13645 constant @samp{__v850e__} will be defined if this option is used.
13647 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13648 are defined then a default target processor will be chosen and the
13649 relevant @samp{__v850*__} preprocessor constant will be defined.
13651 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13652 defined, regardless of which processor variant is the target.
13654 @item -mdisable-callt
13655 @opindex mdisable-callt
13656 This option will suppress generation of the CALLT instruction for the
13657 v850e and v850e1 flavors of the v850 architecture. The default is
13658 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13663 @subsection VAX Options
13664 @cindex VAX options
13666 These @samp{-m} options are defined for the VAX:
13671 Do not output certain jump instructions (@code{aobleq} and so on)
13672 that the Unix assembler for the VAX cannot handle across long
13677 Do output those jump instructions, on the assumption that you
13678 will assemble with the GNU assembler.
13682 Output code for g-format floating point numbers instead of d-format.
13685 @node x86-64 Options
13686 @subsection x86-64 Options
13687 @cindex x86-64 options
13689 These are listed under @xref{i386 and x86-64 Options}.
13691 @node Xstormy16 Options
13692 @subsection Xstormy16 Options
13693 @cindex Xstormy16 Options
13695 These options are defined for Xstormy16:
13700 Choose startup files and linker script suitable for the simulator.
13703 @node Xtensa Options
13704 @subsection Xtensa Options
13705 @cindex Xtensa Options
13707 These options are supported for Xtensa targets:
13711 @itemx -mno-const16
13713 @opindex mno-const16
13714 Enable or disable use of @code{CONST16} instructions for loading
13715 constant values. The @code{CONST16} instruction is currently not a
13716 standard option from Tensilica. When enabled, @code{CONST16}
13717 instructions are always used in place of the standard @code{L32R}
13718 instructions. The use of @code{CONST16} is enabled by default only if
13719 the @code{L32R} instruction is not available.
13722 @itemx -mno-fused-madd
13723 @opindex mfused-madd
13724 @opindex mno-fused-madd
13725 Enable or disable use of fused multiply/add and multiply/subtract
13726 instructions in the floating-point option. This has no effect if the
13727 floating-point option is not also enabled. Disabling fused multiply/add
13728 and multiply/subtract instructions forces the compiler to use separate
13729 instructions for the multiply and add/subtract operations. This may be
13730 desirable in some cases where strict IEEE 754-compliant results are
13731 required: the fused multiply add/subtract instructions do not round the
13732 intermediate result, thereby producing results with @emph{more} bits of
13733 precision than specified by the IEEE standard. Disabling fused multiply
13734 add/subtract instructions also ensures that the program output is not
13735 sensitive to the compiler's ability to combine multiply and add/subtract
13738 @item -mtext-section-literals
13739 @itemx -mno-text-section-literals
13740 @opindex mtext-section-literals
13741 @opindex mno-text-section-literals
13742 Control the treatment of literal pools. The default is
13743 @option{-mno-text-section-literals}, which places literals in a separate
13744 section in the output file. This allows the literal pool to be placed
13745 in a data RAM/ROM, and it also allows the linker to combine literal
13746 pools from separate object files to remove redundant literals and
13747 improve code size. With @option{-mtext-section-literals}, the literals
13748 are interspersed in the text section in order to keep them as close as
13749 possible to their references. This may be necessary for large assembly
13752 @item -mtarget-align
13753 @itemx -mno-target-align
13754 @opindex mtarget-align
13755 @opindex mno-target-align
13756 When this option is enabled, GCC instructs the assembler to
13757 automatically align instructions to reduce branch penalties at the
13758 expense of some code density. The assembler attempts to widen density
13759 instructions to align branch targets and the instructions following call
13760 instructions. If there are not enough preceding safe density
13761 instructions to align a target, no widening will be performed. The
13762 default is @option{-mtarget-align}. These options do not affect the
13763 treatment of auto-aligned instructions like @code{LOOP}, which the
13764 assembler will always align, either by widening density instructions or
13765 by inserting no-op instructions.
13768 @itemx -mno-longcalls
13769 @opindex mlongcalls
13770 @opindex mno-longcalls
13771 When this option is enabled, GCC instructs the assembler to translate
13772 direct calls to indirect calls unless it can determine that the target
13773 of a direct call is in the range allowed by the call instruction. This
13774 translation typically occurs for calls to functions in other source
13775 files. Specifically, the assembler translates a direct @code{CALL}
13776 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13777 The default is @option{-mno-longcalls}. This option should be used in
13778 programs where the call target can potentially be out of range. This
13779 option is implemented in the assembler, not the compiler, so the
13780 assembly code generated by GCC will still show direct call
13781 instructions---look at the disassembled object code to see the actual
13782 instructions. Note that the assembler will use an indirect call for
13783 every cross-file call, not just those that really will be out of range.
13786 @node zSeries Options
13787 @subsection zSeries Options
13788 @cindex zSeries options
13790 These are listed under @xref{S/390 and zSeries Options}.
13792 @node Code Gen Options
13793 @section Options for Code Generation Conventions
13794 @cindex code generation conventions
13795 @cindex options, code generation
13796 @cindex run-time options
13798 These machine-independent options control the interface conventions
13799 used in code generation.
13801 Most of them have both positive and negative forms; the negative form
13802 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13803 one of the forms is listed---the one which is not the default. You
13804 can figure out the other form by either removing @samp{no-} or adding
13808 @item -fbounds-check
13809 @opindex fbounds-check
13810 For front-ends that support it, generate additional code to check that
13811 indices used to access arrays are within the declared range. This is
13812 currently only supported by the Java and Fortran front-ends, where
13813 this option defaults to true and false respectively.
13817 This option generates traps for signed overflow on addition, subtraction,
13818 multiplication operations.
13822 This option instructs the compiler to assume that signed arithmetic
13823 overflow of addition, subtraction and multiplication wraps around
13824 using twos-complement representation. This flag enables some optimizations
13825 and disables others. This option is enabled by default for the Java
13826 front-end, as required by the Java language specification.
13829 @opindex fexceptions
13830 Enable exception handling. Generates extra code needed to propagate
13831 exceptions. For some targets, this implies GCC will generate frame
13832 unwind information for all functions, which can produce significant data
13833 size overhead, although it does not affect execution. If you do not
13834 specify this option, GCC will enable it by default for languages like
13835 C++ which normally require exception handling, and disable it for
13836 languages like C that do not normally require it. However, you may need
13837 to enable this option when compiling C code that needs to interoperate
13838 properly with exception handlers written in C++. You may also wish to
13839 disable this option if you are compiling older C++ programs that don't
13840 use exception handling.
13842 @item -fnon-call-exceptions
13843 @opindex fnon-call-exceptions
13844 Generate code that allows trapping instructions to throw exceptions.
13845 Note that this requires platform-specific runtime support that does
13846 not exist everywhere. Moreover, it only allows @emph{trapping}
13847 instructions to throw exceptions, i.e.@: memory references or floating
13848 point instructions. It does not allow exceptions to be thrown from
13849 arbitrary signal handlers such as @code{SIGALRM}.
13851 @item -funwind-tables
13852 @opindex funwind-tables
13853 Similar to @option{-fexceptions}, except that it will just generate any needed
13854 static data, but will not affect the generated code in any other way.
13855 You will normally not enable this option; instead, a language processor
13856 that needs this handling would enable it on your behalf.
13858 @item -fasynchronous-unwind-tables
13859 @opindex fasynchronous-unwind-tables
13860 Generate unwind table in dwarf2 format, if supported by target machine. The
13861 table is exact at each instruction boundary, so it can be used for stack
13862 unwinding from asynchronous events (such as debugger or garbage collector).
13864 @item -fpcc-struct-return
13865 @opindex fpcc-struct-return
13866 Return ``short'' @code{struct} and @code{union} values in memory like
13867 longer ones, rather than in registers. This convention is less
13868 efficient, but it has the advantage of allowing intercallability between
13869 GCC-compiled files and files compiled with other compilers, particularly
13870 the Portable C Compiler (pcc).
13872 The precise convention for returning structures in memory depends
13873 on the target configuration macros.
13875 Short structures and unions are those whose size and alignment match
13876 that of some integer type.
13878 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13879 switch is not binary compatible with code compiled with the
13880 @option{-freg-struct-return} switch.
13881 Use it to conform to a non-default application binary interface.
13883 @item -freg-struct-return
13884 @opindex freg-struct-return
13885 Return @code{struct} and @code{union} values in registers when possible.
13886 This is more efficient for small structures than
13887 @option{-fpcc-struct-return}.
13889 If you specify neither @option{-fpcc-struct-return} nor
13890 @option{-freg-struct-return}, GCC defaults to whichever convention is
13891 standard for the target. If there is no standard convention, GCC
13892 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13893 the principal compiler. In those cases, we can choose the standard, and
13894 we chose the more efficient register return alternative.
13896 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13897 switch is not binary compatible with code compiled with the
13898 @option{-fpcc-struct-return} switch.
13899 Use it to conform to a non-default application binary interface.
13901 @item -fshort-enums
13902 @opindex fshort-enums
13903 Allocate to an @code{enum} type only as many bytes as it needs for the
13904 declared range of possible values. Specifically, the @code{enum} type
13905 will be equivalent to the smallest integer type which has enough room.
13907 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13908 code that is not binary compatible with code generated without that switch.
13909 Use it to conform to a non-default application binary interface.
13911 @item -fshort-double
13912 @opindex fshort-double
13913 Use the same size for @code{double} as for @code{float}.
13915 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13916 code that is not binary compatible with code generated without that switch.
13917 Use it to conform to a non-default application binary interface.
13919 @item -fshort-wchar
13920 @opindex fshort-wchar
13921 Override the underlying type for @samp{wchar_t} to be @samp{short
13922 unsigned int} instead of the default for the target. This option is
13923 useful for building programs to run under WINE@.
13925 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13926 code that is not binary compatible with code generated without that switch.
13927 Use it to conform to a non-default application binary interface.
13930 @opindex fno-common
13931 In C, allocate even uninitialized global variables in the data section of the
13932 object file, rather than generating them as common blocks. This has the
13933 effect that if the same variable is declared (without @code{extern}) in
13934 two different compilations, you will get an error when you link them.
13935 The only reason this might be useful is if you wish to verify that the
13936 program will work on other systems which always work this way.
13940 Ignore the @samp{#ident} directive.
13942 @item -finhibit-size-directive
13943 @opindex finhibit-size-directive
13944 Don't output a @code{.size} assembler directive, or anything else that
13945 would cause trouble if the function is split in the middle, and the
13946 two halves are placed at locations far apart in memory. This option is
13947 used when compiling @file{crtstuff.c}; you should not need to use it
13950 @item -fverbose-asm
13951 @opindex fverbose-asm
13952 Put extra commentary information in the generated assembly code to
13953 make it more readable. This option is generally only of use to those
13954 who actually need to read the generated assembly code (perhaps while
13955 debugging the compiler itself).
13957 @option{-fno-verbose-asm}, the default, causes the
13958 extra information to be omitted and is useful when comparing two assembler
13961 @item -frecord-gcc-switches
13962 @opindex frecord-gcc-switches
13963 This switch causes the command line that was used to invoke the
13964 compiler to be recorded into the object file that is being created.
13965 This switch is only implemented on some targets and the exact format
13966 of the recording is target and binary file format dependent, but it
13967 usually takes the form of a section containing ASCII text. This
13968 switch is related to the @option{-fverbose-asm} switch, but that
13969 switch only records information in the assembler output file as
13970 comments, so it never reaches the object file.
13974 @cindex global offset table
13976 Generate position-independent code (PIC) suitable for use in a shared
13977 library, if supported for the target machine. Such code accesses all
13978 constant addresses through a global offset table (GOT)@. The dynamic
13979 loader resolves the GOT entries when the program starts (the dynamic
13980 loader is not part of GCC; it is part of the operating system). If
13981 the GOT size for the linked executable exceeds a machine-specific
13982 maximum size, you get an error message from the linker indicating that
13983 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13984 instead. (These maximums are 8k on the SPARC and 32k
13985 on the m68k and RS/6000. The 386 has no such limit.)
13987 Position-independent code requires special support, and therefore works
13988 only on certain machines. For the 386, GCC supports PIC for System V
13989 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13990 position-independent.
13992 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13997 If supported for the target machine, emit position-independent code,
13998 suitable for dynamic linking and avoiding any limit on the size of the
13999 global offset table. This option makes a difference on the m68k,
14000 PowerPC and SPARC@.
14002 Position-independent code requires special support, and therefore works
14003 only on certain machines.
14005 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14012 These options are similar to @option{-fpic} and @option{-fPIC}, but
14013 generated position independent code can be only linked into executables.
14014 Usually these options are used when @option{-pie} GCC option will be
14015 used during linking.
14017 @item -fno-jump-tables
14018 @opindex fno-jump-tables
14019 Do not use jump tables for switch statements even where it would be
14020 more efficient than other code generation strategies. This option is
14021 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14022 building code which forms part of a dynamic linker and cannot
14023 reference the address of a jump table. On some targets, jump tables
14024 do not require a GOT and this option is not needed.
14026 @item -ffixed-@var{reg}
14028 Treat the register named @var{reg} as a fixed register; generated code
14029 should never refer to it (except perhaps as a stack pointer, frame
14030 pointer or in some other fixed role).
14032 @var{reg} must be the name of a register. The register names accepted
14033 are machine-specific and are defined in the @code{REGISTER_NAMES}
14034 macro in the machine description macro file.
14036 This flag does not have a negative form, because it specifies a
14039 @item -fcall-used-@var{reg}
14040 @opindex fcall-used
14041 Treat the register named @var{reg} as an allocable register that is
14042 clobbered by function calls. It may be allocated for temporaries or
14043 variables that do not live across a call. Functions compiled this way
14044 will not save and restore the register @var{reg}.
14046 It is an error to used this flag with the frame pointer or stack pointer.
14047 Use of this flag for other registers that have fixed pervasive roles in
14048 the machine's execution model will produce disastrous results.
14050 This flag does not have a negative form, because it specifies a
14053 @item -fcall-saved-@var{reg}
14054 @opindex fcall-saved
14055 Treat the register named @var{reg} as an allocable register saved by
14056 functions. It may be allocated even for temporaries or variables that
14057 live across a call. Functions compiled this way will save and restore
14058 the register @var{reg} if they use it.
14060 It is an error to used this flag with the frame pointer or stack pointer.
14061 Use of this flag for other registers that have fixed pervasive roles in
14062 the machine's execution model will produce disastrous results.
14064 A different sort of disaster will result from the use of this flag for
14065 a register in which function values may be returned.
14067 This flag does not have a negative form, because it specifies a
14070 @item -fpack-struct[=@var{n}]
14071 @opindex fpack-struct
14072 Without a value specified, pack all structure members together without
14073 holes. When a value is specified (which must be a small power of two), pack
14074 structure members according to this value, representing the maximum
14075 alignment (that is, objects with default alignment requirements larger than
14076 this will be output potentially unaligned at the next fitting location.
14078 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14079 code that is not binary compatible with code generated without that switch.
14080 Additionally, it makes the code suboptimal.
14081 Use it to conform to a non-default application binary interface.
14083 @item -finstrument-functions
14084 @opindex finstrument-functions
14085 Generate instrumentation calls for entry and exit to functions. Just
14086 after function entry and just before function exit, the following
14087 profiling functions will be called with the address of the current
14088 function and its call site. (On some platforms,
14089 @code{__builtin_return_address} does not work beyond the current
14090 function, so the call site information may not be available to the
14091 profiling functions otherwise.)
14094 void __cyg_profile_func_enter (void *this_fn,
14096 void __cyg_profile_func_exit (void *this_fn,
14100 The first argument is the address of the start of the current function,
14101 which may be looked up exactly in the symbol table.
14103 This instrumentation is also done for functions expanded inline in other
14104 functions. The profiling calls will indicate where, conceptually, the
14105 inline function is entered and exited. This means that addressable
14106 versions of such functions must be available. If all your uses of a
14107 function are expanded inline, this may mean an additional expansion of
14108 code size. If you use @samp{extern inline} in your C code, an
14109 addressable version of such functions must be provided. (This is
14110 normally the case anyways, but if you get lucky and the optimizer always
14111 expands the functions inline, you might have gotten away without
14112 providing static copies.)
14114 A function may be given the attribute @code{no_instrument_function}, in
14115 which case this instrumentation will not be done. This can be used, for
14116 example, for the profiling functions listed above, high-priority
14117 interrupt routines, and any functions from which the profiling functions
14118 cannot safely be called (perhaps signal handlers, if the profiling
14119 routines generate output or allocate memory).
14121 @item -fstack-check
14122 @opindex fstack-check
14123 Generate code to verify that you do not go beyond the boundary of the
14124 stack. You should specify this flag if you are running in an
14125 environment with multiple threads, but only rarely need to specify it in
14126 a single-threaded environment since stack overflow is automatically
14127 detected on nearly all systems if there is only one stack.
14129 Note that this switch does not actually cause checking to be done; the
14130 operating system must do that. The switch causes generation of code
14131 to ensure that the operating system sees the stack being extended.
14133 @item -fstack-limit-register=@var{reg}
14134 @itemx -fstack-limit-symbol=@var{sym}
14135 @itemx -fno-stack-limit
14136 @opindex fstack-limit-register
14137 @opindex fstack-limit-symbol
14138 @opindex fno-stack-limit
14139 Generate code to ensure that the stack does not grow beyond a certain value,
14140 either the value of a register or the address of a symbol. If the stack
14141 would grow beyond the value, a signal is raised. For most targets,
14142 the signal is raised before the stack overruns the boundary, so
14143 it is possible to catch the signal without taking special precautions.
14145 For instance, if the stack starts at absolute address @samp{0x80000000}
14146 and grows downwards, you can use the flags
14147 @option{-fstack-limit-symbol=__stack_limit} and
14148 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
14149 of 128KB@. Note that this may only work with the GNU linker.
14151 @cindex aliasing of parameters
14152 @cindex parameters, aliased
14153 @item -fargument-alias
14154 @itemx -fargument-noalias
14155 @itemx -fargument-noalias-global
14156 @itemx -fargument-noalias-anything
14157 @opindex fargument-alias
14158 @opindex fargument-noalias
14159 @opindex fargument-noalias-global
14160 @opindex fargument-noalias-anything
14161 Specify the possible relationships among parameters and between
14162 parameters and global data.
14164 @option{-fargument-alias} specifies that arguments (parameters) may
14165 alias each other and may alias global storage.@*
14166 @option{-fargument-noalias} specifies that arguments do not alias
14167 each other, but may alias global storage.@*
14168 @option{-fargument-noalias-global} specifies that arguments do not
14169 alias each other and do not alias global storage.
14170 @option{-fargument-noalias-anything} specifies that arguments do not
14171 alias any other storage.
14173 Each language will automatically use whatever option is required by
14174 the language standard. You should not need to use these options yourself.
14176 @item -fleading-underscore
14177 @opindex fleading-underscore
14178 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14179 change the way C symbols are represented in the object file. One use
14180 is to help link with legacy assembly code.
14182 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14183 generate code that is not binary compatible with code generated without that
14184 switch. Use it to conform to a non-default application binary interface.
14185 Not all targets provide complete support for this switch.
14187 @item -ftls-model=@var{model}
14188 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14189 The @var{model} argument should be one of @code{global-dynamic},
14190 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
14192 The default without @option{-fpic} is @code{initial-exec}; with
14193 @option{-fpic} the default is @code{global-dynamic}.
14195 @item -fvisibility=@var{default|internal|hidden|protected}
14196 @opindex fvisibility
14197 Set the default ELF image symbol visibility to the specified option---all
14198 symbols will be marked with this unless overridden within the code.
14199 Using this feature can very substantially improve linking and
14200 load times of shared object libraries, produce more optimized
14201 code, provide near-perfect API export and prevent symbol clashes.
14202 It is @strong{strongly} recommended that you use this in any shared objects
14205 Despite the nomenclature, @code{default} always means public ie;
14206 available to be linked against from outside the shared object.
14207 @code{protected} and @code{internal} are pretty useless in real-world
14208 usage so the only other commonly used option will be @code{hidden}.
14209 The default if @option{-fvisibility} isn't specified is
14210 @code{default}, i.e., make every
14211 symbol public---this causes the same behavior as previous versions of
14214 A good explanation of the benefits offered by ensuring ELF
14215 symbols have the correct visibility is given by ``How To Write
14216 Shared Libraries'' by Ulrich Drepper (which can be found at
14217 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
14218 solution made possible by this option to marking things hidden when
14219 the default is public is to make the default hidden and mark things
14220 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
14221 and @code{__attribute__ ((visibility("default")))} instead of
14222 @code{__declspec(dllexport)} you get almost identical semantics with
14223 identical syntax. This is a great boon to those working with
14224 cross-platform projects.
14226 For those adding visibility support to existing code, you may find
14227 @samp{#pragma GCC visibility} of use. This works by you enclosing
14228 the declarations you wish to set visibility for with (for example)
14229 @samp{#pragma GCC visibility push(hidden)} and
14230 @samp{#pragma GCC visibility pop}.
14231 Bear in mind that symbol visibility should be viewed @strong{as
14232 part of the API interface contract} and thus all new code should
14233 always specify visibility when it is not the default ie; declarations
14234 only for use within the local DSO should @strong{always} be marked explicitly
14235 as hidden as so to avoid PLT indirection overheads---making this
14236 abundantly clear also aids readability and self-documentation of the code.
14237 Note that due to ISO C++ specification requirements, operator new and
14238 operator delete must always be of default visibility.
14240 Be aware that headers from outside your project, in particular system
14241 headers and headers from any other library you use, may not be
14242 expecting to be compiled with visibility other than the default. You
14243 may need to explicitly say @samp{#pragma GCC visibility push(default)}
14244 before including any such headers.
14246 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
14247 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
14248 no modifications. However, this means that calls to @samp{extern}
14249 functions with no explicit visibility will use the PLT, so it is more
14250 effective to use @samp{__attribute ((visibility))} and/or
14251 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
14252 declarations should be treated as hidden.
14254 Note that @samp{-fvisibility} does affect C++ vague linkage
14255 entities. This means that, for instance, an exception class that will
14256 be thrown between DSOs must be explicitly marked with default
14257 visibility so that the @samp{type_info} nodes will be unified between
14260 An overview of these techniques, their benefits and how to use them
14261 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
14267 @node Environment Variables
14268 @section Environment Variables Affecting GCC
14269 @cindex environment variables
14271 @c man begin ENVIRONMENT
14272 This section describes several environment variables that affect how GCC
14273 operates. Some of them work by specifying directories or prefixes to use
14274 when searching for various kinds of files. Some are used to specify other
14275 aspects of the compilation environment.
14277 Note that you can also specify places to search using options such as
14278 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
14279 take precedence over places specified using environment variables, which
14280 in turn take precedence over those specified by the configuration of GCC@.
14281 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
14282 GNU Compiler Collection (GCC) Internals}.
14287 @c @itemx LC_COLLATE
14289 @c @itemx LC_MONETARY
14290 @c @itemx LC_NUMERIC
14295 @c @findex LC_COLLATE
14296 @findex LC_MESSAGES
14297 @c @findex LC_MONETARY
14298 @c @findex LC_NUMERIC
14302 These environment variables control the way that GCC uses
14303 localization information that allow GCC to work with different
14304 national conventions. GCC inspects the locale categories
14305 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
14306 so. These locale categories can be set to any value supported by your
14307 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
14308 Kingdom encoded in UTF-8.
14310 The @env{LC_CTYPE} environment variable specifies character
14311 classification. GCC uses it to determine the character boundaries in
14312 a string; this is needed for some multibyte encodings that contain quote
14313 and escape characters that would otherwise be interpreted as a string
14316 The @env{LC_MESSAGES} environment variable specifies the language to
14317 use in diagnostic messages.
14319 If the @env{LC_ALL} environment variable is set, it overrides the value
14320 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
14321 and @env{LC_MESSAGES} default to the value of the @env{LANG}
14322 environment variable. If none of these variables are set, GCC
14323 defaults to traditional C English behavior.
14327 If @env{TMPDIR} is set, it specifies the directory to use for temporary
14328 files. GCC uses temporary files to hold the output of one stage of
14329 compilation which is to be used as input to the next stage: for example,
14330 the output of the preprocessor, which is the input to the compiler
14333 @item GCC_EXEC_PREFIX
14334 @findex GCC_EXEC_PREFIX
14335 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
14336 names of the subprograms executed by the compiler. No slash is added
14337 when this prefix is combined with the name of a subprogram, but you can
14338 specify a prefix that ends with a slash if you wish.
14340 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
14341 an appropriate prefix to use based on the pathname it was invoked with.
14343 If GCC cannot find the subprogram using the specified prefix, it
14344 tries looking in the usual places for the subprogram.
14346 The default value of @env{GCC_EXEC_PREFIX} is
14347 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
14348 the installed compiler. In many cases @var{prefix} is the value
14349 of @code{prefix} when you ran the @file{configure} script.
14351 Other prefixes specified with @option{-B} take precedence over this prefix.
14353 This prefix is also used for finding files such as @file{crt0.o} that are
14356 In addition, the prefix is used in an unusual way in finding the
14357 directories to search for header files. For each of the standard
14358 directories whose name normally begins with @samp{/usr/local/lib/gcc}
14359 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
14360 replacing that beginning with the specified prefix to produce an
14361 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
14362 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
14363 These alternate directories are searched first; the standard directories
14364 come next. If a standard directory begins with the configured
14365 @var{prefix} then the value of @var{prefix} is replaced by
14366 @env{GCC_EXEC_PREFIX} when looking for header files.
14368 @item COMPILER_PATH
14369 @findex COMPILER_PATH
14370 The value of @env{COMPILER_PATH} is a colon-separated list of
14371 directories, much like @env{PATH}. GCC tries the directories thus
14372 specified when searching for subprograms, if it can't find the
14373 subprograms using @env{GCC_EXEC_PREFIX}.
14376 @findex LIBRARY_PATH
14377 The value of @env{LIBRARY_PATH} is a colon-separated list of
14378 directories, much like @env{PATH}. When configured as a native compiler,
14379 GCC tries the directories thus specified when searching for special
14380 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
14381 using GCC also uses these directories when searching for ordinary
14382 libraries for the @option{-l} option (but directories specified with
14383 @option{-L} come first).
14387 @cindex locale definition
14388 This variable is used to pass locale information to the compiler. One way in
14389 which this information is used is to determine the character set to be used
14390 when character literals, string literals and comments are parsed in C and C++.
14391 When the compiler is configured to allow multibyte characters,
14392 the following values for @env{LANG} are recognized:
14396 Recognize JIS characters.
14398 Recognize SJIS characters.
14400 Recognize EUCJP characters.
14403 If @env{LANG} is not defined, or if it has some other value, then the
14404 compiler will use mblen and mbtowc as defined by the default locale to
14405 recognize and translate multibyte characters.
14409 Some additional environments variables affect the behavior of the
14412 @include cppenv.texi
14416 @node Precompiled Headers
14417 @section Using Precompiled Headers
14418 @cindex precompiled headers
14419 @cindex speed of compilation
14421 Often large projects have many header files that are included in every
14422 source file. The time the compiler takes to process these header files
14423 over and over again can account for nearly all of the time required to
14424 build the project. To make builds faster, GCC allows users to
14425 `precompile' a header file; then, if builds can use the precompiled
14426 header file they will be much faster.
14428 To create a precompiled header file, simply compile it as you would any
14429 other file, if necessary using the @option{-x} option to make the driver
14430 treat it as a C or C++ header file. You will probably want to use a
14431 tool like @command{make} to keep the precompiled header up-to-date when
14432 the headers it contains change.
14434 A precompiled header file will be searched for when @code{#include} is
14435 seen in the compilation. As it searches for the included file
14436 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
14437 compiler looks for a precompiled header in each directory just before it
14438 looks for the include file in that directory. The name searched for is
14439 the name specified in the @code{#include} with @samp{.gch} appended. If
14440 the precompiled header file can't be used, it is ignored.
14442 For instance, if you have @code{#include "all.h"}, and you have
14443 @file{all.h.gch} in the same directory as @file{all.h}, then the
14444 precompiled header file will be used if possible, and the original
14445 header will be used otherwise.
14447 Alternatively, you might decide to put the precompiled header file in a
14448 directory and use @option{-I} to ensure that directory is searched
14449 before (or instead of) the directory containing the original header.
14450 Then, if you want to check that the precompiled header file is always
14451 used, you can put a file of the same name as the original header in this
14452 directory containing an @code{#error} command.
14454 This also works with @option{-include}. So yet another way to use
14455 precompiled headers, good for projects not designed with precompiled
14456 header files in mind, is to simply take most of the header files used by
14457 a project, include them from another header file, precompile that header
14458 file, and @option{-include} the precompiled header. If the header files
14459 have guards against multiple inclusion, they will be skipped because
14460 they've already been included (in the precompiled header).
14462 If you need to precompile the same header file for different
14463 languages, targets, or compiler options, you can instead make a
14464 @emph{directory} named like @file{all.h.gch}, and put each precompiled
14465 header in the directory, perhaps using @option{-o}. It doesn't matter
14466 what you call the files in the directory, every precompiled header in
14467 the directory will be considered. The first precompiled header
14468 encountered in the directory that is valid for this compilation will
14469 be used; they're searched in no particular order.
14471 There are many other possibilities, limited only by your imagination,
14472 good sense, and the constraints of your build system.
14474 A precompiled header file can be used only when these conditions apply:
14478 Only one precompiled header can be used in a particular compilation.
14481 A precompiled header can't be used once the first C token is seen. You
14482 can have preprocessor directives before a precompiled header; you can
14483 even include a precompiled header from inside another header, so long as
14484 there are no C tokens before the @code{#include}.
14487 The precompiled header file must be produced for the same language as
14488 the current compilation. You can't use a C precompiled header for a C++
14492 The precompiled header file must have been produced by the same compiler
14493 binary as the current compilation is using.
14496 Any macros defined before the precompiled header is included must
14497 either be defined in the same way as when the precompiled header was
14498 generated, or must not affect the precompiled header, which usually
14499 means that they don't appear in the precompiled header at all.
14501 The @option{-D} option is one way to define a macro before a
14502 precompiled header is included; using a @code{#define} can also do it.
14503 There are also some options that define macros implicitly, like
14504 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
14507 @item If debugging information is output when using the precompiled
14508 header, using @option{-g} or similar, the same kind of debugging information
14509 must have been output when building the precompiled header. However,
14510 a precompiled header built using @option{-g} can be used in a compilation
14511 when no debugging information is being output.
14513 @item The same @option{-m} options must generally be used when building
14514 and using the precompiled header. @xref{Submodel Options},
14515 for any cases where this rule is relaxed.
14517 @item Each of the following options must be the same when building and using
14518 the precompiled header:
14520 @gccoptlist{-fexceptions -funit-at-a-time}
14523 Some other command-line options starting with @option{-f},
14524 @option{-p}, or @option{-O} must be defined in the same way as when
14525 the precompiled header was generated. At present, it's not clear
14526 which options are safe to change and which are not; the safest choice
14527 is to use exactly the same options when generating and using the
14528 precompiled header. The following are known to be safe:
14530 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
14531 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
14532 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
14537 For all of these except the last, the compiler will automatically
14538 ignore the precompiled header if the conditions aren't met. If you
14539 find an option combination that doesn't work and doesn't cause the
14540 precompiled header to be ignored, please consider filing a bug report,
14543 If you do use differing options when generating and using the
14544 precompiled header, the actual behavior will be a mixture of the
14545 behavior for the options. For instance, if you use @option{-g} to
14546 generate the precompiled header but not when using it, you may or may
14547 not get debugging information for routines in the precompiled header.
14549 @node Running Protoize
14550 @section Running Protoize
14552 The program @code{protoize} is an optional part of GCC@. You can use
14553 it to add prototypes to a program, thus converting the program to ISO
14554 C in one respect. The companion program @code{unprotoize} does the
14555 reverse: it removes argument types from any prototypes that are found.
14557 When you run these programs, you must specify a set of source files as
14558 command line arguments. The conversion programs start out by compiling
14559 these files to see what functions they define. The information gathered
14560 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
14562 After scanning comes actual conversion. The specified files are all
14563 eligible to be converted; any files they include (whether sources or
14564 just headers) are eligible as well.
14566 But not all the eligible files are converted. By default,
14567 @code{protoize} and @code{unprotoize} convert only source and header
14568 files in the current directory. You can specify additional directories
14569 whose files should be converted with the @option{-d @var{directory}}
14570 option. You can also specify particular files to exclude with the
14571 @option{-x @var{file}} option. A file is converted if it is eligible, its
14572 directory name matches one of the specified directory names, and its
14573 name within the directory has not been excluded.
14575 Basic conversion with @code{protoize} consists of rewriting most
14576 function definitions and function declarations to specify the types of
14577 the arguments. The only ones not rewritten are those for varargs
14580 @code{protoize} optionally inserts prototype declarations at the
14581 beginning of the source file, to make them available for any calls that
14582 precede the function's definition. Or it can insert prototype
14583 declarations with block scope in the blocks where undeclared functions
14586 Basic conversion with @code{unprotoize} consists of rewriting most
14587 function declarations to remove any argument types, and rewriting
14588 function definitions to the old-style pre-ISO form.
14590 Both conversion programs print a warning for any function declaration or
14591 definition that they can't convert. You can suppress these warnings
14594 The output from @code{protoize} or @code{unprotoize} replaces the
14595 original source file. The original file is renamed to a name ending
14596 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
14597 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
14598 for DOS) file already exists, then the source file is simply discarded.
14600 @code{protoize} and @code{unprotoize} both depend on GCC itself to
14601 scan the program and collect information about the functions it uses.
14602 So neither of these programs will work until GCC is installed.
14604 Here is a table of the options you can use with @code{protoize} and
14605 @code{unprotoize}. Each option works with both programs unless
14609 @item -B @var{directory}
14610 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
14611 usual directory (normally @file{/usr/local/lib}). This file contains
14612 prototype information about standard system functions. This option
14613 applies only to @code{protoize}.
14615 @item -c @var{compilation-options}
14616 Use @var{compilation-options} as the options when running @command{gcc} to
14617 produce the @samp{.X} files. The special option @option{-aux-info} is
14618 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
14620 Note that the compilation options must be given as a single argument to
14621 @code{protoize} or @code{unprotoize}. If you want to specify several
14622 @command{gcc} options, you must quote the entire set of compilation options
14623 to make them a single word in the shell.
14625 There are certain @command{gcc} arguments that you cannot use, because they
14626 would produce the wrong kind of output. These include @option{-g},
14627 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14628 the @var{compilation-options}, they are ignored.
14631 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14632 systems) instead of @samp{.c}. This is convenient if you are converting
14633 a C program to C++. This option applies only to @code{protoize}.
14636 Add explicit global declarations. This means inserting explicit
14637 declarations at the beginning of each source file for each function
14638 that is called in the file and was not declared. These declarations
14639 precede the first function definition that contains a call to an
14640 undeclared function. This option applies only to @code{protoize}.
14642 @item -i @var{string}
14643 Indent old-style parameter declarations with the string @var{string}.
14644 This option applies only to @code{protoize}.
14646 @code{unprotoize} converts prototyped function definitions to old-style
14647 function definitions, where the arguments are declared between the
14648 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14649 uses five spaces as the indentation. If you want to indent with just
14650 one space instead, use @option{-i " "}.
14653 Keep the @samp{.X} files. Normally, they are deleted after conversion
14657 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14658 a prototype declaration for each function in each block which calls the
14659 function without any declaration. This option applies only to
14663 Make no real changes. This mode just prints information about the conversions
14664 that would have been done without @option{-n}.
14667 Make no @samp{.save} files. The original files are simply deleted.
14668 Use this option with caution.
14670 @item -p @var{program}
14671 Use the program @var{program} as the compiler. Normally, the name
14672 @file{gcc} is used.
14675 Work quietly. Most warnings are suppressed.
14678 Print the version number, just like @option{-v} for @command{gcc}.
14681 If you need special compiler options to compile one of your program's
14682 source files, then you should generate that file's @samp{.X} file
14683 specially, by running @command{gcc} on that source file with the
14684 appropriate options and the option @option{-aux-info}. Then run
14685 @code{protoize} on the entire set of files. @code{protoize} will use
14686 the existing @samp{.X} file because it is newer than the source file.
14690 gcc -Dfoo=bar file1.c -aux-info file1.X
14695 You need to include the special files along with the rest in the
14696 @code{protoize} command, even though their @samp{.X} files already
14697 exist, because otherwise they won't get converted.
14699 @xref{Protoize Caveats}, for more information on how to use
14700 @code{protoize} successfully.