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
106 options of the same kind; for example, if you specify @option{-L} more
107 than once, the directories are searched in the order specified. Also,
108 the placement of the @option{-l} option is significant.
110 Many options have long names starting with @samp{-f} or with
111 @samp{-W}---for example,
112 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
113 these have both positive and negative forms; the negative form of
114 @option{-ffoo} would be @option{-fno-foo}. This manual documents
115 only one of these two forms, whichever one is not the default.
119 @xref{Option Index}, for an index to GCC's options.
122 * Option Summary:: Brief list of all options, without explanations.
123 * Overall Options:: Controlling the kind of output:
124 an executable, object files, assembler files,
125 or preprocessed source.
126 * Invoking G++:: Compiling C++ programs.
127 * C Dialect Options:: Controlling the variant of C language compiled.
128 * C++ Dialect Options:: Variations on C++.
129 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
131 * Language Independent Options:: Controlling how diagnostics should be
133 * Warning Options:: How picky should the compiler be?
134 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
135 * Optimize Options:: How much optimization?
136 * Preprocessor Options:: Controlling header files and macro definitions.
137 Also, getting dependency information for Make.
138 * Assembler Options:: Passing options to the assembler.
139 * Link Options:: Specifying libraries and so on.
140 * Directory Options:: Where to find header files and libraries.
141 Where to find the compiler executable files.
142 * Spec Files:: How to pass switches to sub-processes.
143 * Target Options:: Running a cross-compiler, or an old version of GCC.
144 * Submodel Options:: Specifying minor hardware or convention variations,
145 such as 68010 vs 68020.
146 * Code Gen Options:: Specifying conventions for function calls, data layout
148 * Environment Variables:: Env vars that affect GCC.
149 * Precompiled Headers:: Compiling a header once, and using it many times.
150 * Running Protoize:: Automatically adding or removing function prototypes.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
165 -x @var{language} -v -### --help@r{[}=@var{class}@r{]} --target-help @gol
166 --version @@@var{file}}
168 @item C Language Options
169 @xref{C Dialect Options,,Options Controlling C Dialect}.
170 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
171 -aux-info @var{filename} @gol
172 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
173 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
174 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
175 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
176 -fsigned-bitfields -fsigned-char @gol
177 -funsigned-bitfields -funsigned-char}
179 @item C++ Language Options
180 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
181 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
182 -fconserve-space -ffriend-injection @gol
183 -fno-elide-constructors @gol
184 -fno-enforce-eh-specs @gol
185 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
186 -fno-implicit-templates @gol
187 -fno-implicit-inline-templates @gol
188 -fno-implement-inlines -fms-extensions @gol
189 -fno-nonansi-builtins -fno-operator-names @gol
190 -fno-optional-diags -fpermissive @gol
191 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
192 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
193 -fno-default-inline -fvisibility-inlines-hidden @gol
194 -fvisibility-ms-compat @gol
195 -Wabi -Wctor-dtor-privacy @gol
196 -Wnon-virtual-dtor -Wreorder @gol
197 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
198 -Wno-non-template-friend -Wold-style-cast @gol
199 -Woverloaded-virtual -Wno-pmf-conversions @gol
202 @item Objective-C and Objective-C++ Language Options
203 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
204 Objective-C and Objective-C++ Dialects}.
205 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
206 -fgnu-runtime -fnext-runtime @gol
207 -fno-nil-receivers @gol
208 -fobjc-call-cxx-cdtors @gol
209 -fobjc-direct-dispatch @gol
210 -fobjc-exceptions @gol
212 -freplace-objc-classes @gol
215 -Wassign-intercept @gol
216 -Wno-protocol -Wselector @gol
217 -Wstrict-selector-match @gol
218 -Wundeclared-selector}
220 @item Language Independent Options
221 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
222 @gccoptlist{-fmessage-length=@var{n} @gol
223 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
224 -fdiagnostics-show-option}
226 @item Warning Options
227 @xref{Warning Options,,Options to Request or Suppress Warnings}.
228 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
229 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
230 -Wno-attributes -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
231 -Wchar-subscripts -Wclobbered -Wcomment @gol
232 -Wconversion -Wcoverage-mismatch -Wno-deprecated-declarations @gol
233 -Wdisabled-optimization -Wno-div-by-zero @gol
234 -Wempty-body -Wno-endif-labels @gol
235 -Werror -Werror=* @gol
236 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
237 -Wno-format-extra-args -Wformat-nonliteral @gol
238 -Wformat-security -Wformat-y2k @gol
239 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
240 -Wimport -Wno-import -Winit-self -Winline @gol
241 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
242 -Winvalid-pch -Wlarger-than-@var{len} -Wunsafe-loop-optimizations @gol
243 -Wlogical-op -Wlong-long @gol
244 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
245 -Wmissing-format-attribute -Wmissing-include-dirs @gol
246 -Wmissing-noreturn @gol
247 -Wno-multichar -Wnonnull -Wno-overflow @gol
248 -Woverlength-strings -Wpacked -Wpadded @gol
249 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
250 -Wredundant-decls @gol
251 -Wreturn-type -Wsequence-point -Wshadow @gol
252 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
253 -Wstrict-aliasing -Wstrict-aliasing=n @gol
254 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
255 -Wswitch -Wswitch-default -Wswitch-enum @gol
256 -Wsystem-headers -Wtrigraphs -Wtype-limits -Wundef -Wuninitialized @gol
257 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
258 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
259 -Wunused-value -Wunused-variable @gol
260 -Wvariadic-macros -Wvla @gol
261 -Wvolatile-register-var -Wwrite-strings}
263 @item C-only Warning Options
264 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
265 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
266 -Wold-style-declaration -Wold-style-definition @gol
267 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
268 -Wdeclaration-after-statement -Wpointer-sign}
270 @item Debugging Options
271 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
272 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
273 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
274 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
275 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
276 -fdump-ipa-all -fdump-ipa-cgraph @gol
278 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
283 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
287 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
289 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
291 -fdump-tree-nrv -fdump-tree-vect @gol
292 -fdump-tree-sink @gol
293 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
294 -fdump-tree-salias @gol
295 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
296 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
297 -ftree-vectorizer-verbose=@var{n} @gol
298 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
299 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
300 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
301 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
302 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
303 -ftest-coverage -ftime-report -fvar-tracking @gol
304 -g -g@var{level} -gcoff -gdwarf-2 @gol
305 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
306 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
307 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
308 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
309 -print-multi-directory -print-multi-lib @gol
310 -print-prog-name=@var{program} -print-search-dirs -Q @gol
311 -print-sysroot-headers-suffix @gol
314 @item Optimization Options
315 @xref{Optimize Options,,Options that Control Optimization}.
316 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
317 -falign-labels=@var{n} -falign-loops=@var{n} @gol
318 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
319 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
320 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
321 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
322 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
323 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
324 -fexpensive-optimizations -ffast-math -ffloat-store @gol
325 -fforce-addr -fforward-propagate -ffunction-sections @gol
326 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
327 -fcrossjumping -fif-conversion -fif-conversion2 @gol
328 -finline-functions -finline-functions-called-once @gol
329 -finline-limit=@var{n} -fkeep-inline-functions @gol
330 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
331 -fmodulo-sched -fno-branch-count-reg @gol
332 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
333 -fno-function-cse -fno-guess-branch-probability @gol
334 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
335 -funsafe-math-optimizations -funsafe-loop-optimizations @gol
336 -ffinite-math-only -fno-signed-zeros @gol
337 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
338 -fomit-frame-pointer -foptimize-register-move @gol
339 -foptimize-sibling-calls -fpredictive-commoning -fprefetch-loop-arrays @gol
340 -fprofile-generate -fprofile-use @gol
341 -fregmove -frename-registers @gol
342 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
343 -frerun-cse-after-loop @gol
344 -frounding-math -frtl-abstract-sequences @gol
345 -fschedule-insns -fschedule-insns2 @gol
346 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
347 -fsched-spec-load-dangerous @gol
348 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
349 -fsched2-use-superblocks @gol
350 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
351 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
352 -fno-split-wide-types -fstack-protector -fstack-protector-all @gol
353 -fstrict-aliasing -fstrict-overflow -ftracer -fthread-jumps @gol
354 -funroll-all-loops -funroll-loops -fpeel-loops @gol
355 -fsplit-ivs-in-unroller -funswitch-loops @gol
356 -fvariable-expansion-in-unroller -ftree-reassoc @gol
357 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
358 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
359 -fcheck-data-deps @gol
360 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
361 -ftree-ch -ftree-sra -ftree-ter -ftree-fre -ftree-vectorize @gol
362 -ftree-vect-loop-version -fvect-cost-model -ftree-salias -fipa-pta -fweb @gol
363 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
364 --param @var{name}=@var{value}
365 -O -O0 -O1 -O2 -O3 -Os}
367 @item Preprocessor Options
368 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
369 @gccoptlist{-A@var{question}=@var{answer} @gol
370 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
371 -C -dD -dI -dM -dN @gol
372 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
373 -idirafter @var{dir} @gol
374 -include @var{file} -imacros @var{file} @gol
375 -iprefix @var{file} -iwithprefix @var{dir} @gol
376 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
377 -imultilib @var{dir} -isysroot @var{dir} @gol
378 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
379 -P -fworking-directory -remap @gol
380 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
381 -Xpreprocessor @var{option}}
383 @item Assembler Option
384 @xref{Assembler Options,,Passing Options to the Assembler}.
385 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
388 @xref{Link Options,,Options for Linking}.
389 @gccoptlist{@var{object-file-name} -l@var{library} @gol
390 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
391 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
392 -Wl,@var{option} -Xlinker @var{option} @gol
395 @item Directory Options
396 @xref{Directory Options,,Options for Directory Search}.
397 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
398 -specs=@var{file} -I- --sysroot=@var{dir}}
401 @c I wrote this xref this way to avoid overfull hbox. -- rms
402 @xref{Target Options}.
403 @gccoptlist{-V @var{version} -b @var{machine}}
405 @item Machine Dependent Options
406 @xref{Submodel Options,,Hardware Models and Configurations}.
407 @c This list is ordered alphanumerically by subsection name.
408 @c Try and put the significant identifier (CPU or system) first,
409 @c so users have a clue at guessing where the ones they want will be.
412 @gccoptlist{-EB -EL @gol
413 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
414 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
417 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
418 -mabi=@var{name} @gol
419 -mapcs-stack-check -mno-apcs-stack-check @gol
420 -mapcs-float -mno-apcs-float @gol
421 -mapcs-reentrant -mno-apcs-reentrant @gol
422 -msched-prolog -mno-sched-prolog @gol
423 -mlittle-endian -mbig-endian -mwords-little-endian @gol
424 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
425 -mthumb-interwork -mno-thumb-interwork @gol
426 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
427 -mstructure-size-boundary=@var{n} @gol
428 -mabort-on-noreturn @gol
429 -mlong-calls -mno-long-calls @gol
430 -msingle-pic-base -mno-single-pic-base @gol
431 -mpic-register=@var{reg} @gol
432 -mnop-fun-dllimport @gol
433 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
434 -mpoke-function-name @gol
436 -mtpcs-frame -mtpcs-leaf-frame @gol
437 -mcaller-super-interworking -mcallee-super-interworking @gol
441 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
442 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
444 @emph{Blackfin Options}
445 @gccoptlist{-mcpu=@var{cpu} -msim -momit-leaf-frame-pointer @gol
446 -mno-omit-leaf-frame-pointer -mspecld-anomaly -mno-specld-anomaly @gol
447 -mcsync-anomaly -mno-csync-anomaly -mlow-64k -mno-low64k @gol
448 -mstack-check-l1 -mid-shared-library -mno-id-shared-library @gol
449 -mshared-library-id=@var{n} -mleaf-id-shared-library @gol
450 -mno-leaf-id-shared-library -msep-data -mno-sep-data -mlong-calls @gol
454 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
455 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
456 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
457 -mstack-align -mdata-align -mconst-align @gol
458 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
459 -melf -maout -melinux -mlinux -sim -sim2 @gol
460 -mmul-bug-workaround -mno-mul-bug-workaround}
463 @gccoptlist{-mmac -mpush-args}
465 @emph{Darwin Options}
466 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
467 -arch_only -bind_at_load -bundle -bundle_loader @gol
468 -client_name -compatibility_version -current_version @gol
470 -dependency-file -dylib_file -dylinker_install_name @gol
471 -dynamic -dynamiclib -exported_symbols_list @gol
472 -filelist -flat_namespace -force_cpusubtype_ALL @gol
473 -force_flat_namespace -headerpad_max_install_names @gol
475 -image_base -init -install_name -keep_private_externs @gol
476 -multi_module -multiply_defined -multiply_defined_unused @gol
477 -noall_load -no_dead_strip_inits_and_terms @gol
478 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
479 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
480 -private_bundle -read_only_relocs -sectalign @gol
481 -sectobjectsymbols -whyload -seg1addr @gol
482 -sectcreate -sectobjectsymbols -sectorder @gol
483 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
484 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
485 -segprot -segs_read_only_addr -segs_read_write_addr @gol
486 -single_module -static -sub_library -sub_umbrella @gol
487 -twolevel_namespace -umbrella -undefined @gol
488 -unexported_symbols_list -weak_reference_mismatches @gol
489 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
490 -mkernel -mone-byte-bool}
492 @emph{DEC Alpha Options}
493 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
494 -mieee -mieee-with-inexact -mieee-conformant @gol
495 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
496 -mtrap-precision=@var{mode} -mbuild-constants @gol
497 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
498 -mbwx -mmax -mfix -mcix @gol
499 -mfloat-vax -mfloat-ieee @gol
500 -mexplicit-relocs -msmall-data -mlarge-data @gol
501 -msmall-text -mlarge-text @gol
502 -mmemory-latency=@var{time}}
504 @emph{DEC Alpha/VMS Options}
505 @gccoptlist{-mvms-return-codes}
508 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
509 -mhard-float -msoft-float @gol
510 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
511 -mdouble -mno-double @gol
512 -mmedia -mno-media -mmuladd -mno-muladd @gol
513 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
514 -mlinked-fp -mlong-calls -malign-labels @gol
515 -mlibrary-pic -macc-4 -macc-8 @gol
516 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
517 -moptimize-membar -mno-optimize-membar @gol
518 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
519 -mvliw-branch -mno-vliw-branch @gol
520 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
521 -mno-nested-cond-exec -mtomcat-stats @gol
525 @emph{GNU/Linux Options}
526 @gccoptlist{-muclibc}
528 @emph{H8/300 Options}
529 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
532 @gccoptlist{-march=@var{architecture-type} @gol
533 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
534 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
535 -mfixed-range=@var{register-range} @gol
536 -mjump-in-delay -mlinker-opt -mlong-calls @gol
537 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
538 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
539 -mno-jump-in-delay -mno-long-load-store @gol
540 -mno-portable-runtime -mno-soft-float @gol
541 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
542 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
543 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
544 -munix=@var{unix-std} -nolibdld -static -threads}
546 @emph{i386 and x86-64 Options}
547 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
548 -mfpmath=@var{unit} @gol
549 -masm=@var{dialect} -mno-fancy-math-387 @gol
550 -mno-fp-ret-in-387 -msoft-float @gol
551 -mno-wide-multiply -mrtd -malign-double @gol
552 -mpreferred-stack-boundary=@var{num} -mcx16 -msahf -mrecip @gol
553 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 @gol
554 -msse4a -m3dnow -mpopcnt -mabm @gol
555 -mthreads -mno-align-stringops -minline-all-stringops @gol
556 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
557 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
558 -mpc32 -mpc64 -mpc80 mstackrealign @gol
559 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
560 -mcmodel=@var{code-model} @gol
561 -m32 -m64 -mlarge-data-threshold=@var{num}}
564 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
565 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
566 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
567 -minline-float-divide-max-throughput @gol
568 -minline-int-divide-min-latency @gol
569 -minline-int-divide-max-throughput @gol
570 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
571 -mno-dwarf2-asm -mearly-stop-bits @gol
572 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
573 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
574 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
575 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
576 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
577 -mno-sched-prefer-non-data-spec-insns @gol
578 -mno-sched-prefer-non-control-spec-insns @gol
579 -mno-sched-count-spec-in-critical-path}
581 @emph{M32R/D Options}
582 @gccoptlist{-m32r2 -m32rx -m32r @gol
584 -malign-loops -mno-align-loops @gol
585 -missue-rate=@var{number} @gol
586 -mbranch-cost=@var{number} @gol
587 -mmodel=@var{code-size-model-type} @gol
588 -msdata=@var{sdata-type} @gol
589 -mno-flush-func -mflush-func=@var{name} @gol
590 -mno-flush-trap -mflush-trap=@var{number} @gol
594 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
596 @emph{M680x0 Options}
597 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
598 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
599 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
600 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
601 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
602 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
603 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
604 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
606 @emph{M68hc1x Options}
607 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
608 -mauto-incdec -minmax -mlong-calls -mshort @gol
609 -msoft-reg-count=@var{count}}
612 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
613 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
614 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
615 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
616 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
619 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
620 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
621 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
622 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
623 -mfp32 -mfp64 -mhard-float -msoft-float @gol
624 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
625 -msmartmips -mno-smartmips @gol
626 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
627 -mips3d -mno-mips3d -mmt -mno-mt @gol
628 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
629 -G@var{num} -membedded-data -mno-embedded-data @gol
630 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
631 -msplit-addresses -mno-split-addresses @gol
632 -mexplicit-relocs -mno-explicit-relocs @gol
633 -mcheck-zero-division -mno-check-zero-division @gol
634 -mdivide-traps -mdivide-breaks @gol
635 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
636 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
637 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
638 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 -mno-fix-vr4130 @gol
639 -mfix-sb1 -mno-fix-sb1 @gol
640 -mflush-func=@var{func} -mno-flush-func @gol
641 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
642 -mfp-exceptions -mno-fp-exceptions @gol
643 -mvr4130-align -mno-vr4130-align}
646 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
647 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
648 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
649 -mno-base-addresses -msingle-exit -mno-single-exit}
651 @emph{MN10300 Options}
652 @gccoptlist{-mmult-bug -mno-mult-bug @gol
653 -mam33 -mno-am33 @gol
654 -mam33-2 -mno-am33-2 @gol
655 -mreturn-pointer-on-d0 @gol
659 @gccoptlist{-mno-crt0 -mbacc -msim @gol
660 -march=@var{cpu-type} }
662 @emph{PDP-11 Options}
663 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
664 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
665 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
666 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
667 -mbranch-expensive -mbranch-cheap @gol
668 -msplit -mno-split -munix-asm -mdec-asm}
670 @emph{PowerPC Options}
671 See RS/6000 and PowerPC Options.
673 @emph{RS/6000 and PowerPC Options}
674 @gccoptlist{-mcpu=@var{cpu-type} @gol
675 -mtune=@var{cpu-type} @gol
676 -mpower -mno-power -mpower2 -mno-power2 @gol
677 -mpowerpc -mpowerpc64 -mno-powerpc @gol
678 -maltivec -mno-altivec @gol
679 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
680 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
681 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
682 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mdfp -mno-dfp @gol
683 -mnew-mnemonics -mold-mnemonics @gol
684 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
685 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
686 -malign-power -malign-natural @gol
687 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
688 -mstring -mno-string -mupdate -mno-update @gol
689 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
690 -mstrict-align -mno-strict-align -mrelocatable @gol
691 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
692 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
693 -mdynamic-no-pic -maltivec -mswdiv @gol
694 -mprioritize-restricted-insns=@var{priority} @gol
695 -msched-costly-dep=@var{dependence_type} @gol
696 -minsert-sched-nops=@var{scheme} @gol
697 -mcall-sysv -mcall-netbsd @gol
698 -maix-struct-return -msvr4-struct-return @gol
699 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
700 -misel -mno-isel @gol
701 -misel=yes -misel=no @gol
703 -mspe=yes -mspe=no @gol
704 -mvrsave -mno-vrsave @gol
705 -mmulhw -mno-mulhw @gol
706 -mdlmzb -mno-dlmzb @gol
707 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
708 -mprototype -mno-prototype @gol
709 -msim -mmvme -mads -myellowknife -memb -msdata @gol
710 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
712 @emph{S/390 and zSeries Options}
713 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
714 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
715 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
716 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
717 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
718 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
719 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
722 @gccoptlist{-meb -mel @gol
726 -mscore5 -mscore5u -mscore7 -mscore7d}
729 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
730 -m4-nofpu -m4-single-only -m4-single -m4 @gol
731 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
732 -m5-64media -m5-64media-nofpu @gol
733 -m5-32media -m5-32media-nofpu @gol
734 -m5-compact -m5-compact-nofpu @gol
735 -mb -ml -mdalign -mrelax @gol
736 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
737 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
738 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
739 -mdivsi3_libfunc=@var{name} @gol
740 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
744 @gccoptlist{-mcpu=@var{cpu-type} @gol
745 -mtune=@var{cpu-type} @gol
746 -mcmodel=@var{code-model} @gol
747 -m32 -m64 -mapp-regs -mno-app-regs @gol
748 -mfaster-structs -mno-faster-structs @gol
749 -mfpu -mno-fpu -mhard-float -msoft-float @gol
750 -mhard-quad-float -msoft-quad-float @gol
751 -mimpure-text -mno-impure-text -mlittle-endian @gol
752 -mstack-bias -mno-stack-bias @gol
753 -munaligned-doubles -mno-unaligned-doubles @gol
754 -mv8plus -mno-v8plus -mvis -mno-vis
755 -threads -pthreads -pthread}
758 @gccoptlist{-mwarn-reloc -merror-reloc @gol
759 -msafe-dma -munsafe-dma @gol
761 -msmall-mem -mlarge-mem -mstdmain @gol
762 -mfixed-range=@var{register-range}}
764 @emph{System V Options}
765 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
767 @emph{TMS320C3x/C4x Options}
768 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
769 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
770 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
771 -mparallel-insns -mparallel-mpy -mpreserve-float}
774 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
775 -mprolog-function -mno-prolog-function -mspace @gol
776 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
777 -mapp-regs -mno-app-regs @gol
778 -mdisable-callt -mno-disable-callt @gol
784 @gccoptlist{-mg -mgnu -munix}
786 @emph{VxWorks Options}
787 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
788 -Xbind-lazy -Xbind-now}
790 @emph{x86-64 Options}
791 See i386 and x86-64 Options.
793 @emph{Xstormy16 Options}
796 @emph{Xtensa Options}
797 @gccoptlist{-mconst16 -mno-const16 @gol
798 -mfused-madd -mno-fused-madd @gol
799 -mtext-section-literals -mno-text-section-literals @gol
800 -mtarget-align -mno-target-align @gol
801 -mlongcalls -mno-longcalls}
803 @emph{zSeries Options}
804 See S/390 and zSeries Options.
806 @item Code Generation Options
807 @xref{Code Gen Options,,Options for Code Generation Conventions}.
808 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
809 -ffixed-@var{reg} -fexceptions @gol
810 -fnon-call-exceptions -funwind-tables @gol
811 -fasynchronous-unwind-tables @gol
812 -finhibit-size-directive -finstrument-functions @gol
813 -fno-common -fno-ident @gol
814 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
815 -fno-jump-tables @gol
816 -frecord-gcc-switches @gol
817 -freg-struct-return -fshort-enums @gol
818 -fshort-double -fshort-wchar @gol
819 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
820 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
821 -fargument-alias -fargument-noalias @gol
822 -fargument-noalias-global -fargument-noalias-anything
823 -fleading-underscore -ftls-model=@var{model} @gol
824 -ftrapv -fwrapv -fbounds-check @gol
829 * Overall Options:: Controlling the kind of output:
830 an executable, object files, assembler files,
831 or preprocessed source.
832 * C Dialect Options:: Controlling the variant of C language compiled.
833 * C++ Dialect Options:: Variations on C++.
834 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
836 * Language Independent Options:: Controlling how diagnostics should be
838 * Warning Options:: How picky should the compiler be?
839 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
840 * Optimize Options:: How much optimization?
841 * Preprocessor Options:: Controlling header files and macro definitions.
842 Also, getting dependency information for Make.
843 * Assembler Options:: Passing options to the assembler.
844 * Link Options:: Specifying libraries and so on.
845 * Directory Options:: Where to find header files and libraries.
846 Where to find the compiler executable files.
847 * Spec Files:: How to pass switches to sub-processes.
848 * Target Options:: Running a cross-compiler, or an old version of GCC.
851 @node Overall Options
852 @section Options Controlling the Kind of Output
854 Compilation can involve up to four stages: preprocessing, compilation
855 proper, assembly and linking, always in that order. GCC is capable of
856 preprocessing and compiling several files either into several
857 assembler input files, or into one assembler input file; then each
858 assembler input file produces an object file, and linking combines all
859 the object files (those newly compiled, and those specified as input)
860 into an executable file.
862 @cindex file name suffix
863 For any given input file, the file name suffix determines what kind of
868 C source code which must be preprocessed.
871 C source code which should not be preprocessed.
874 C++ source code which should not be preprocessed.
877 Objective-C source code. Note that you must link with the @file{libobjc}
878 library to make an Objective-C program work.
881 Objective-C source code which should not be preprocessed.
885 Objective-C++ source code. Note that you must link with the @file{libobjc}
886 library to make an Objective-C++ program work. Note that @samp{.M} refers
887 to a literal capital M@.
890 Objective-C++ source code which should not be preprocessed.
893 C, C++, Objective-C or Objective-C++ header file to be turned into a
898 @itemx @var{file}.cxx
899 @itemx @var{file}.cpp
900 @itemx @var{file}.CPP
901 @itemx @var{file}.c++
903 C++ source code which must be preprocessed. Note that in @samp{.cxx},
904 the last two letters must both be literally @samp{x}. Likewise,
905 @samp{.C} refers to a literal capital C@.
909 Objective-C++ source code which must be preprocessed.
912 Objective-C++ source code which should not be preprocessed.
916 C++ header file to be turned into a precompiled header.
919 @itemx @var{file}.for
920 @itemx @var{file}.FOR
921 Fixed form Fortran source code which should not be preprocessed.
924 @itemx @var{file}.fpp
925 @itemx @var{file}.FPP
926 Fixed form Fortran source code which must be preprocessed (with the traditional
930 @itemx @var{file}.f95
931 Free form Fortran source code which should not be preprocessed.
934 @itemx @var{file}.F95
935 Free form Fortran source code which must be preprocessed (with the
936 traditional preprocessor).
938 @c FIXME: Descriptions of Java file types.
945 Ada source code file which contains a library unit declaration (a
946 declaration of a package, subprogram, or generic, or a generic
947 instantiation), or a library unit renaming declaration (a package,
948 generic, or subprogram renaming declaration). Such files are also
951 @itemx @var{file}.adb
952 Ada source code file containing a library unit body (a subprogram or
953 package body). Such files are also called @dfn{bodies}.
955 @c GCC also knows about some suffixes for languages not yet included:
966 Assembler code which must be preprocessed.
969 An object file to be fed straight into linking.
970 Any file name with no recognized suffix is treated this way.
974 You can specify the input language explicitly with the @option{-x} option:
977 @item -x @var{language}
978 Specify explicitly the @var{language} for the following input files
979 (rather than letting the compiler choose a default based on the file
980 name suffix). This option applies to all following input files until
981 the next @option{-x} option. Possible values for @var{language} are:
983 c c-header c-cpp-output
984 c++ c++-header c++-cpp-output
985 objective-c objective-c-header objective-c-cpp-output
986 objective-c++ objective-c++-header objective-c++-cpp-output
987 assembler assembler-with-cpp
995 Turn off any specification of a language, so that subsequent files are
996 handled according to their file name suffixes (as they are if @option{-x}
997 has not been used at all).
999 @item -pass-exit-codes
1000 @opindex pass-exit-codes
1001 Normally the @command{gcc} program will exit with the code of 1 if any
1002 phase of the compiler returns a non-success return code. If you specify
1003 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1004 numerically highest error produced by any phase that returned an error
1005 indication. The C, C++, and Fortran frontends return 4, if an internal
1006 compiler error is encountered.
1009 If you only want some of the stages of compilation, you can use
1010 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1011 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1012 @command{gcc} is to stop. Note that some combinations (for example,
1013 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1018 Compile or assemble the source files, but do not link. The linking
1019 stage simply is not done. The ultimate output is in the form of an
1020 object file for each source file.
1022 By default, the object file name for a source file is made by replacing
1023 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1025 Unrecognized input files, not requiring compilation or assembly, are
1030 Stop after the stage of compilation proper; do not assemble. The output
1031 is in the form of an assembler code file for each non-assembler input
1034 By default, the assembler file name for a source file is made by
1035 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1037 Input files that don't require compilation are ignored.
1041 Stop after the preprocessing stage; do not run the compiler proper. The
1042 output is in the form of preprocessed source code, which is sent to the
1045 Input files which don't require preprocessing are ignored.
1047 @cindex output file option
1050 Place output in file @var{file}. This applies regardless to whatever
1051 sort of output is being produced, whether it be an executable file,
1052 an object file, an assembler file or preprocessed C code.
1054 If @option{-o} is not specified, the default is to put an executable
1055 file in @file{a.out}, the object file for
1056 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1057 assembler file in @file{@var{source}.s}, a precompiled header file in
1058 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1063 Print (on standard error output) the commands executed to run the stages
1064 of compilation. Also print the version number of the compiler driver
1065 program and of the preprocessor and the compiler proper.
1069 Like @option{-v} except the commands are not executed and all command
1070 arguments are quoted. This is useful for shell scripts to capture the
1071 driver-generated command lines.
1075 Use pipes rather than temporary files for communication between the
1076 various stages of compilation. This fails to work on some systems where
1077 the assembler is unable to read from a pipe; but the GNU assembler has
1082 If you are compiling multiple source files, this option tells the driver
1083 to pass all the source files to the compiler at once (for those
1084 languages for which the compiler can handle this). This will allow
1085 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1086 language for which this is supported is C@. If you pass source files for
1087 multiple languages to the driver, using this option, the driver will invoke
1088 the compiler(s) that support IMA once each, passing each compiler all the
1089 source files appropriate for it. For those languages that do not support
1090 IMA this option will be ignored, and the compiler will be invoked once for
1091 each source file in that language. If you use this option in conjunction
1092 with @option{-save-temps}, the compiler will generate multiple
1094 (one for each source file), but only one (combined) @file{.o} or
1099 Print (on the standard output) a description of the command line options
1100 understood by @command{gcc}. If the @option{-v} option is also specified
1101 then @option{--help} will also be passed on to the various processes
1102 invoked by @command{gcc}, so that they can display the command line options
1103 they accept. If the @option{-Wextra} option has also been specified
1104 (prior to the @option{--help} option), then command line options which
1105 have no documentation associated with them will also be displayed.
1108 @opindex target-help
1109 Print (on the standard output) a description of target-specific command
1110 line options for each tool. For some targets extra target-specific
1111 information may also be printed.
1113 @item --help=@var{class}@r{[},@var{qualifier}@r{]}
1114 Print (on the standard output) a description of the command line
1115 options understood by the compiler that fit into a specific class.
1116 The class can be one of @samp{optimizers}, @samp{warnings}, @samp{target},
1117 @samp{params}, or @var{language}:
1120 @item @samp{optimizers}
1121 This will display all of the optimization options supported by the
1124 @item @samp{warnings}
1125 This will display all of the options controlling warning messages
1126 produced by the compiler.
1129 This will display target-specific options. Unlike the
1130 @option{--target-help} option however, target-specific options of the
1131 linker and assembler will not be displayed. This is because those
1132 tools do not currently support the extended @option{--help=} syntax.
1135 This will display the values recognized by the @option{--param}
1138 @item @var{language}
1139 This will display the options supported for @var{language}, where
1140 @var{language} is the name of one of the languages supported in this
1144 This will display the options that are common to all languages.
1147 It is possible to further refine the output of the @option{--help=}
1148 option by adding a comma separated list of qualifiers after the
1149 class. These can be any from the following list:
1152 @item @samp{undocumented}
1153 Display only those options which are undocumented.
1156 Display options which take an argument that appears after an equal
1157 sign in the same continuous piece of text, such as:
1158 @samp{--help=target}.
1160 @item @samp{separate}
1161 Display options which take an argument that appears as a separate word
1162 following the original option, such as: @samp{-o output-file}.
1165 Thus for example to display all the undocumented target-specific
1166 switches supported by the compiler the following can be used:
1169 --help=target,undocumented
1172 The sense of a qualifier can be inverted by prefixing it with the
1173 @var{^} character, so for example to display all binary warning
1174 options (i.e. ones that are either on or off and that do not take an
1175 argument), which have a description the following can be used:
1178 --help=warnings,^joined,^undocumented
1181 A class can also be used as a qualifier, although this usually
1182 restricts the output by so much that there is nothing to display. One
1183 case where it does work however is when one of the classes is
1184 @var{target}. So for example to display all the target-specific
1185 optimization options the following can be used:
1188 --help=target,optimizers
1191 The @option{--help=} option can be repeated on the command line. Each
1192 successive use will display its requested class of options, skipping
1193 those that have already been displayed.
1195 If the @option{-Q} option appears on the command line before the
1196 @option{--help=} option, then the descriptive text displayed by
1197 @option{--help=} is changed. Instead of describing the displayed
1198 options, an indication is given as to whether the option is enabled,
1199 disabled or set to a specific value (assuming that the compiler
1200 knows this at the point where the @option{--help=} option is used).
1202 Here is a truncated example from the ARM port of @command{gcc}:
1205 % gcc -Q -mabi=2 --help=target -c
1206 The following options are target specific:
1208 -mabort-on-noreturn [disabled]
1212 The output is sensitive to the effects of previous command line
1213 options, so for example it is possible to find out which optimizations
1214 are enabled at @option{-O2} by using:
1217 -O2 --help=optimizers
1220 Alternatively you can discover which binary optimizations are enabled
1221 by @option{-O3} by using:
1224 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1225 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1226 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1231 Display the version number and copyrights of the invoked GCC@.
1233 @include @value{srcdir}/../libiberty/at-file.texi
1237 @section Compiling C++ Programs
1239 @cindex suffixes for C++ source
1240 @cindex C++ source file suffixes
1241 C++ source files conventionally use one of the suffixes @samp{.C},
1242 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1243 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1244 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1245 files with these names and compiles them as C++ programs even if you
1246 call the compiler the same way as for compiling C programs (usually
1247 with the name @command{gcc}).
1251 However, the use of @command{gcc} does not add the C++ library.
1252 @command{g++} is a program that calls GCC and treats @samp{.c},
1253 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1254 files unless @option{-x} is used, and automatically specifies linking
1255 against the C++ library. This program is also useful when
1256 precompiling a C header file with a @samp{.h} extension for use in C++
1257 compilations. On many systems, @command{g++} is also installed with
1258 the name @command{c++}.
1260 @cindex invoking @command{g++}
1261 When you compile C++ programs, you may specify many of the same
1262 command-line options that you use for compiling programs in any
1263 language; or command-line options meaningful for C and related
1264 languages; or options that are meaningful only for C++ programs.
1265 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1266 explanations of options for languages related to C@.
1267 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1268 explanations of options that are meaningful only for C++ programs.
1270 @node C Dialect Options
1271 @section Options Controlling C Dialect
1272 @cindex dialect options
1273 @cindex language dialect options
1274 @cindex options, dialect
1276 The following options control the dialect of C (or languages derived
1277 from C, such as C++, Objective-C and Objective-C++) that the compiler
1281 @cindex ANSI support
1285 In C mode, support all ISO C90 programs. In C++ mode,
1286 remove GNU extensions that conflict with ISO C++.
1288 This turns off certain features of GCC that are incompatible with ISO
1289 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1290 such as the @code{asm} and @code{typeof} keywords, and
1291 predefined macros such as @code{unix} and @code{vax} that identify the
1292 type of system you are using. It also enables the undesirable and
1293 rarely used ISO trigraph feature. For the C compiler,
1294 it disables recognition of C++ style @samp{//} comments as well as
1295 the @code{inline} keyword.
1297 The alternate keywords @code{__asm__}, @code{__extension__},
1298 @code{__inline__} and @code{__typeof__} continue to work despite
1299 @option{-ansi}. You would not want to use them in an ISO C program, of
1300 course, but it is useful to put them in header files that might be included
1301 in compilations done with @option{-ansi}. Alternate predefined macros
1302 such as @code{__unix__} and @code{__vax__} are also available, with or
1303 without @option{-ansi}.
1305 The @option{-ansi} option does not cause non-ISO programs to be
1306 rejected gratuitously. For that, @option{-pedantic} is required in
1307 addition to @option{-ansi}. @xref{Warning Options}.
1309 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1310 option is used. Some header files may notice this macro and refrain
1311 from declaring certain functions or defining certain macros that the
1312 ISO standard doesn't call for; this is to avoid interfering with any
1313 programs that might use these names for other things.
1315 Functions which would normally be built in but do not have semantics
1316 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1317 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1318 built-in functions provided by GCC}, for details of the functions
1323 Determine the language standard. This option is currently only
1324 supported when compiling C or C++. A value for this option must be
1325 provided; possible values are
1330 ISO C90 (same as @option{-ansi}).
1332 @item iso9899:199409
1333 ISO C90 as modified in amendment 1.
1339 ISO C99. Note that this standard is not yet fully supported; see
1340 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1341 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1344 Default, ISO C90 plus GNU extensions (including some C99 features).
1348 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1349 this will become the default. The name @samp{gnu9x} is deprecated.
1352 The 1998 ISO C++ standard plus amendments.
1355 The same as @option{-std=c++98} plus GNU extensions. This is the
1356 default for C++ code.
1359 The working draft of the upcoming ISO C++0x standard. This option
1360 enables experimental features that are likely to be included in
1361 C++0x. The working draft is constantly changing, and any feature that is
1362 enabled by this flag may be removed from future versions of GCC if it is
1363 not part of the C++0x standard.
1366 The same as @option{-std=c++0x} plus GNU extensions. As with
1367 @option{-std=c++0x}, this option enables experimental features that may
1368 be removed in future versions of GCC.
1371 Even when this option is not specified, you can still use some of the
1372 features of newer standards in so far as they do not conflict with
1373 previous C standards. For example, you may use @code{__restrict__} even
1374 when @option{-std=c99} is not specified.
1376 The @option{-std} options specifying some version of ISO C have the same
1377 effects as @option{-ansi}, except that features that were not in ISO C90
1378 but are in the specified version (for example, @samp{//} comments and
1379 the @code{inline} keyword in ISO C99) are not disabled.
1381 @xref{Standards,,Language Standards Supported by GCC}, for details of
1382 these standard versions.
1384 @item -fgnu89-inline
1385 @opindex fgnu89-inline
1386 The option @option{-fgnu89-inline} tells GCC to use the traditional
1387 GNU semantics for @code{inline} functions when in C99 mode.
1388 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1389 is accepted and ignored by GCC versions 4.1.3 up to but not including
1390 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1391 C99 mode. Using this option is roughly equivalent to adding the
1392 @code{gnu_inline} function attribute to all inline functions
1393 (@pxref{Function Attributes}).
1395 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1396 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1397 specifies the default behavior). This option was first supported in
1398 GCC 4.3. This option is not supported in C89 or gnu89 mode.
1400 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1401 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1402 in effect for @code{inline} functions. @xref{Common Predefined
1403 Macros,,,cpp,The C Preprocessor}.
1405 @item -aux-info @var{filename}
1407 Output to the given filename prototyped declarations for all functions
1408 declared and/or defined in a translation unit, including those in header
1409 files. This option is silently ignored in any language other than C@.
1411 Besides declarations, the file indicates, in comments, the origin of
1412 each declaration (source file and line), whether the declaration was
1413 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1414 @samp{O} for old, respectively, in the first character after the line
1415 number and the colon), and whether it came from a declaration or a
1416 definition (@samp{C} or @samp{F}, respectively, in the following
1417 character). In the case of function definitions, a K&R-style list of
1418 arguments followed by their declarations is also provided, inside
1419 comments, after the declaration.
1423 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1424 keyword, so that code can use these words as identifiers. You can use
1425 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1426 instead. @option{-ansi} implies @option{-fno-asm}.
1428 In C++, this switch only affects the @code{typeof} keyword, since
1429 @code{asm} and @code{inline} are standard keywords. You may want to
1430 use the @option{-fno-gnu-keywords} flag instead, which has the same
1431 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1432 switch only affects the @code{asm} and @code{typeof} keywords, since
1433 @code{inline} is a standard keyword in ISO C99.
1436 @itemx -fno-builtin-@var{function}
1437 @opindex fno-builtin
1438 @cindex built-in functions
1439 Don't recognize built-in functions that do not begin with
1440 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1441 functions provided by GCC}, for details of the functions affected,
1442 including those which are not built-in functions when @option{-ansi} or
1443 @option{-std} options for strict ISO C conformance are used because they
1444 do not have an ISO standard meaning.
1446 GCC normally generates special code to handle certain built-in functions
1447 more efficiently; for instance, calls to @code{alloca} may become single
1448 instructions that adjust the stack directly, and calls to @code{memcpy}
1449 may become inline copy loops. The resulting code is often both smaller
1450 and faster, but since the function calls no longer appear as such, you
1451 cannot set a breakpoint on those calls, nor can you change the behavior
1452 of the functions by linking with a different library. In addition,
1453 when a function is recognized as a built-in function, GCC may use
1454 information about that function to warn about problems with calls to
1455 that function, or to generate more efficient code, even if the
1456 resulting code still contains calls to that function. For example,
1457 warnings are given with @option{-Wformat} for bad calls to
1458 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1459 known not to modify global memory.
1461 With the @option{-fno-builtin-@var{function}} option
1462 only the built-in function @var{function} is
1463 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1464 function is named this is not built-in in this version of GCC, this
1465 option is ignored. There is no corresponding
1466 @option{-fbuiltin-@var{function}} option; if you wish to enable
1467 built-in functions selectively when using @option{-fno-builtin} or
1468 @option{-ffreestanding}, you may define macros such as:
1471 #define abs(n) __builtin_abs ((n))
1472 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1477 @cindex hosted environment
1479 Assert that compilation takes place in a hosted environment. This implies
1480 @option{-fbuiltin}. A hosted environment is one in which the
1481 entire standard library is available, and in which @code{main} has a return
1482 type of @code{int}. Examples are nearly everything except a kernel.
1483 This is equivalent to @option{-fno-freestanding}.
1485 @item -ffreestanding
1486 @opindex ffreestanding
1487 @cindex hosted environment
1489 Assert that compilation takes place in a freestanding environment. This
1490 implies @option{-fno-builtin}. A freestanding environment
1491 is one in which the standard library may not exist, and program startup may
1492 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1493 This is equivalent to @option{-fno-hosted}.
1495 @xref{Standards,,Language Standards Supported by GCC}, for details of
1496 freestanding and hosted environments.
1500 @cindex openmp parallel
1501 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1502 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1503 compiler generates parallel code according to the OpenMP Application
1504 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1506 @item -fms-extensions
1507 @opindex fms-extensions
1508 Accept some non-standard constructs used in Microsoft header files.
1510 Some cases of unnamed fields in structures and unions are only
1511 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1512 fields within structs/unions}, for details.
1516 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1517 options for strict ISO C conformance) implies @option{-trigraphs}.
1519 @item -no-integrated-cpp
1520 @opindex no-integrated-cpp
1521 Performs a compilation in two passes: preprocessing and compiling. This
1522 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1523 @option{-B} option. The user supplied compilation step can then add in
1524 an additional preprocessing step after normal preprocessing but before
1525 compiling. The default is to use the integrated cpp (internal cpp)
1527 The semantics of this option will change if "cc1", "cc1plus", and
1528 "cc1obj" are merged.
1530 @cindex traditional C language
1531 @cindex C language, traditional
1533 @itemx -traditional-cpp
1534 @opindex traditional-cpp
1535 @opindex traditional
1536 Formerly, these options caused GCC to attempt to emulate a pre-standard
1537 C compiler. They are now only supported with the @option{-E} switch.
1538 The preprocessor continues to support a pre-standard mode. See the GNU
1539 CPP manual for details.
1541 @item -fcond-mismatch
1542 @opindex fcond-mismatch
1543 Allow conditional expressions with mismatched types in the second and
1544 third arguments. The value of such an expression is void. This option
1545 is not supported for C++.
1547 @item -flax-vector-conversions
1548 @opindex flax-vector-conversions
1549 Allow implicit conversions between vectors with differing numbers of
1550 elements and/or incompatible element types. This option should not be
1553 @item -funsigned-char
1554 @opindex funsigned-char
1555 Let the type @code{char} be unsigned, like @code{unsigned char}.
1557 Each kind of machine has a default for what @code{char} should
1558 be. It is either like @code{unsigned char} by default or like
1559 @code{signed char} by default.
1561 Ideally, a portable program should always use @code{signed char} or
1562 @code{unsigned char} when it depends on the signedness of an object.
1563 But many programs have been written to use plain @code{char} and
1564 expect it to be signed, or expect it to be unsigned, depending on the
1565 machines they were written for. This option, and its inverse, let you
1566 make such a program work with the opposite default.
1568 The type @code{char} is always a distinct type from each of
1569 @code{signed char} or @code{unsigned char}, even though its behavior
1570 is always just like one of those two.
1573 @opindex fsigned-char
1574 Let the type @code{char} be signed, like @code{signed char}.
1576 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1577 the negative form of @option{-funsigned-char}. Likewise, the option
1578 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1580 @item -fsigned-bitfields
1581 @itemx -funsigned-bitfields
1582 @itemx -fno-signed-bitfields
1583 @itemx -fno-unsigned-bitfields
1584 @opindex fsigned-bitfields
1585 @opindex funsigned-bitfields
1586 @opindex fno-signed-bitfields
1587 @opindex fno-unsigned-bitfields
1588 These options control whether a bit-field is signed or unsigned, when the
1589 declaration does not use either @code{signed} or @code{unsigned}. By
1590 default, such a bit-field is signed, because this is consistent: the
1591 basic integer types such as @code{int} are signed types.
1594 @node C++ Dialect Options
1595 @section Options Controlling C++ Dialect
1597 @cindex compiler options, C++
1598 @cindex C++ options, command line
1599 @cindex options, C++
1600 This section describes the command-line options that are only meaningful
1601 for C++ programs; but you can also use most of the GNU compiler options
1602 regardless of what language your program is in. For example, you
1603 might compile a file @code{firstClass.C} like this:
1606 g++ -g -frepo -O -c firstClass.C
1610 In this example, only @option{-frepo} is an option meant
1611 only for C++ programs; you can use the other options with any
1612 language supported by GCC@.
1614 Here is a list of options that are @emph{only} for compiling C++ programs:
1618 @item -fabi-version=@var{n}
1619 @opindex fabi-version
1620 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1621 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1622 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1623 the version that conforms most closely to the C++ ABI specification.
1624 Therefore, the ABI obtained using version 0 will change as ABI bugs
1627 The default is version 2.
1629 @item -fno-access-control
1630 @opindex fno-access-control
1631 Turn off all access checking. This switch is mainly useful for working
1632 around bugs in the access control code.
1636 Check that the pointer returned by @code{operator new} is non-null
1637 before attempting to modify the storage allocated. This check is
1638 normally unnecessary because the C++ standard specifies that
1639 @code{operator new} will only return @code{0} if it is declared
1640 @samp{throw()}, in which case the compiler will always check the
1641 return value even without this option. In all other cases, when
1642 @code{operator new} has a non-empty exception specification, memory
1643 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1644 @samp{new (nothrow)}.
1646 @item -fconserve-space
1647 @opindex fconserve-space
1648 Put uninitialized or runtime-initialized global variables into the
1649 common segment, as C does. This saves space in the executable at the
1650 cost of not diagnosing duplicate definitions. If you compile with this
1651 flag and your program mysteriously crashes after @code{main()} has
1652 completed, you may have an object that is being destroyed twice because
1653 two definitions were merged.
1655 This option is no longer useful on most targets, now that support has
1656 been added for putting variables into BSS without making them common.
1658 @item -ffriend-injection
1659 @opindex ffriend-injection
1660 Inject friend functions into the enclosing namespace, so that they are
1661 visible outside the scope of the class in which they are declared.
1662 Friend functions were documented to work this way in the old Annotated
1663 C++ Reference Manual, and versions of G++ before 4.1 always worked
1664 that way. However, in ISO C++ a friend function which is not declared
1665 in an enclosing scope can only be found using argument dependent
1666 lookup. This option causes friends to be injected as they were in
1669 This option is for compatibility, and may be removed in a future
1672 @item -fno-elide-constructors
1673 @opindex fno-elide-constructors
1674 The C++ standard allows an implementation to omit creating a temporary
1675 which is only used to initialize another object of the same type.
1676 Specifying this option disables that optimization, and forces G++ to
1677 call the copy constructor in all cases.
1679 @item -fno-enforce-eh-specs
1680 @opindex fno-enforce-eh-specs
1681 Don't generate code to check for violation of exception specifications
1682 at runtime. This option violates the C++ standard, but may be useful
1683 for reducing code size in production builds, much like defining
1684 @samp{NDEBUG}. This does not give user code permission to throw
1685 exceptions in violation of the exception specifications; the compiler
1686 will still optimize based on the specifications, so throwing an
1687 unexpected exception will result in undefined behavior.
1690 @itemx -fno-for-scope
1692 @opindex fno-for-scope
1693 If @option{-ffor-scope} is specified, the scope of variables declared in
1694 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1695 as specified by the C++ standard.
1696 If @option{-fno-for-scope} is specified, the scope of variables declared in
1697 a @i{for-init-statement} extends to the end of the enclosing scope,
1698 as was the case in old versions of G++, and other (traditional)
1699 implementations of C++.
1701 The default if neither flag is given to follow the standard,
1702 but to allow and give a warning for old-style code that would
1703 otherwise be invalid, or have different behavior.
1705 @item -fno-gnu-keywords
1706 @opindex fno-gnu-keywords
1707 Do not recognize @code{typeof} as a keyword, so that code can use this
1708 word as an identifier. You can use the keyword @code{__typeof__} instead.
1709 @option{-ansi} implies @option{-fno-gnu-keywords}.
1711 @item -fno-implicit-templates
1712 @opindex fno-implicit-templates
1713 Never emit code for non-inline templates which are instantiated
1714 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1715 @xref{Template Instantiation}, for more information.
1717 @item -fno-implicit-inline-templates
1718 @opindex fno-implicit-inline-templates
1719 Don't emit code for implicit instantiations of inline templates, either.
1720 The default is to handle inlines differently so that compiles with and
1721 without optimization will need the same set of explicit instantiations.
1723 @item -fno-implement-inlines
1724 @opindex fno-implement-inlines
1725 To save space, do not emit out-of-line copies of inline functions
1726 controlled by @samp{#pragma implementation}. This will cause linker
1727 errors if these functions are not inlined everywhere they are called.
1729 @item -fms-extensions
1730 @opindex fms-extensions
1731 Disable pedantic warnings about constructs used in MFC, such as implicit
1732 int and getting a pointer to member function via non-standard syntax.
1734 @item -fno-nonansi-builtins
1735 @opindex fno-nonansi-builtins
1736 Disable built-in declarations of functions that are not mandated by
1737 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1738 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1740 @item -fno-operator-names
1741 @opindex fno-operator-names
1742 Do not treat the operator name keywords @code{and}, @code{bitand},
1743 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1744 synonyms as keywords.
1746 @item -fno-optional-diags
1747 @opindex fno-optional-diags
1748 Disable diagnostics that the standard says a compiler does not need to
1749 issue. Currently, the only such diagnostic issued by G++ is the one for
1750 a name having multiple meanings within a class.
1753 @opindex fpermissive
1754 Downgrade some diagnostics about nonconformant code from errors to
1755 warnings. Thus, using @option{-fpermissive} will allow some
1756 nonconforming code to compile.
1760 Enable automatic template instantiation at link time. This option also
1761 implies @option{-fno-implicit-templates}. @xref{Template
1762 Instantiation}, for more information.
1766 Disable generation of information about every class with virtual
1767 functions for use by the C++ runtime type identification features
1768 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1769 of the language, you can save some space by using this flag. Note that
1770 exception handling uses the same information, but it will generate it as
1771 needed. The @samp{dynamic_cast} operator can still be used for casts that
1772 do not require runtime type information, i.e. casts to @code{void *} or to
1773 unambiguous base classes.
1777 Emit statistics about front-end processing at the end of the compilation.
1778 This information is generally only useful to the G++ development team.
1780 @item -ftemplate-depth-@var{n}
1781 @opindex ftemplate-depth
1782 Set the maximum instantiation depth for template classes to @var{n}.
1783 A limit on the template instantiation depth is needed to detect
1784 endless recursions during template class instantiation. ANSI/ISO C++
1785 conforming programs must not rely on a maximum depth greater than 17.
1787 @item -fno-threadsafe-statics
1788 @opindex fno-threadsafe-statics
1789 Do not emit the extra code to use the routines specified in the C++
1790 ABI for thread-safe initialization of local statics. You can use this
1791 option to reduce code size slightly in code that doesn't need to be
1794 @item -fuse-cxa-atexit
1795 @opindex fuse-cxa-atexit
1796 Register destructors for objects with static storage duration with the
1797 @code{__cxa_atexit} function rather than the @code{atexit} function.
1798 This option is required for fully standards-compliant handling of static
1799 destructors, but will only work if your C library supports
1800 @code{__cxa_atexit}.
1802 @item -fno-use-cxa-get-exception-ptr
1803 @opindex fno-use-cxa-get-exception-ptr
1804 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1805 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1806 if the runtime routine is not available.
1808 @item -fvisibility-inlines-hidden
1809 @opindex fvisibility-inlines-hidden
1810 This switch declares that the user does not attempt to compare
1811 pointers to inline methods where the addresses of the two functions
1812 were taken in different shared objects.
1814 The effect of this is that GCC may, effectively, mark inline methods with
1815 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1816 appear in the export table of a DSO and do not require a PLT indirection
1817 when used within the DSO@. Enabling this option can have a dramatic effect
1818 on load and link times of a DSO as it massively reduces the size of the
1819 dynamic export table when the library makes heavy use of templates.
1821 The behavior of this switch is not quite the same as marking the
1822 methods as hidden directly, because it does not affect static variables
1823 local to the function or cause the compiler to deduce that
1824 the function is defined in only one shared object.
1826 You may mark a method as having a visibility explicitly to negate the
1827 effect of the switch for that method. For example, if you do want to
1828 compare pointers to a particular inline method, you might mark it as
1829 having default visibility. Marking the enclosing class with explicit
1830 visibility will have no effect.
1832 Explicitly instantiated inline methods are unaffected by this option
1833 as their linkage might otherwise cross a shared library boundary.
1834 @xref{Template Instantiation}.
1836 @item -fvisibility-ms-compat
1837 @opindex fvisibility-ms-compat
1838 This flag attempts to use visibility settings to make GCC's C++
1839 linkage model compatible with that of Microsoft Visual Studio.
1841 The flag makes these changes to GCC's linkage model:
1845 It sets the default visibility to @code{hidden}, like
1846 @option{-fvisibility=hidden}.
1849 Types, but not their members, are not hidden by default.
1852 The One Definition Rule is relaxed for types without explicit
1853 visibility specifications which are defined in more than one different
1854 shared object: those declarations are permitted if they would have
1855 been permitted when this option was not used.
1858 In new code it is better to use @option{-fvisibility=hidden} and
1859 export those classes which are intended to be externally visible.
1860 Unfortunately it is possible for code to rely, perhaps accidentally,
1861 on the Visual Studio behavior.
1863 Among the consequences of these changes are that static data members
1864 of the same type with the same name but defined in different shared
1865 objects will be different, so changing one will not change the other;
1866 and that pointers to function members defined in different shared
1867 objects may not compare equal. When this flag is given, it is a
1868 violation of the ODR to define types with the same name differently.
1872 Do not use weak symbol support, even if it is provided by the linker.
1873 By default, G++ will use weak symbols if they are available. This
1874 option exists only for testing, and should not be used by end-users;
1875 it will result in inferior code and has no benefits. This option may
1876 be removed in a future release of G++.
1880 Do not search for header files in the standard directories specific to
1881 C++, but do still search the other standard directories. (This option
1882 is used when building the C++ library.)
1885 In addition, these optimization, warning, and code generation options
1886 have meanings only for C++ programs:
1889 @item -fno-default-inline
1890 @opindex fno-default-inline
1891 Do not assume @samp{inline} for functions defined inside a class scope.
1892 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1893 functions will have linkage like inline functions; they just won't be
1896 @item -Wabi @r{(C++ only)}
1898 Warn when G++ generates code that is probably not compatible with the
1899 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1900 all such cases, there are probably some cases that are not warned about,
1901 even though G++ is generating incompatible code. There may also be
1902 cases where warnings are emitted even though the code that is generated
1905 You should rewrite your code to avoid these warnings if you are
1906 concerned about the fact that code generated by G++ may not be binary
1907 compatible with code generated by other compilers.
1909 The known incompatibilities at this point include:
1914 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1915 pack data into the same byte as a base class. For example:
1918 struct A @{ virtual void f(); int f1 : 1; @};
1919 struct B : public A @{ int f2 : 1; @};
1923 In this case, G++ will place @code{B::f2} into the same byte
1924 as@code{A::f1}; other compilers will not. You can avoid this problem
1925 by explicitly padding @code{A} so that its size is a multiple of the
1926 byte size on your platform; that will cause G++ and other compilers to
1927 layout @code{B} identically.
1930 Incorrect handling of tail-padding for virtual bases. G++ does not use
1931 tail padding when laying out virtual bases. For example:
1934 struct A @{ virtual void f(); char c1; @};
1935 struct B @{ B(); char c2; @};
1936 struct C : public A, public virtual B @{@};
1940 In this case, G++ will not place @code{B} into the tail-padding for
1941 @code{A}; other compilers will. You can avoid this problem by
1942 explicitly padding @code{A} so that its size is a multiple of its
1943 alignment (ignoring virtual base classes); that will cause G++ and other
1944 compilers to layout @code{C} identically.
1947 Incorrect handling of bit-fields with declared widths greater than that
1948 of their underlying types, when the bit-fields appear in a union. For
1952 union U @{ int i : 4096; @};
1956 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1957 union too small by the number of bits in an @code{int}.
1960 Empty classes can be placed at incorrect offsets. For example:
1970 struct C : public B, public A @{@};
1974 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1975 it should be placed at offset zero. G++ mistakenly believes that the
1976 @code{A} data member of @code{B} is already at offset zero.
1979 Names of template functions whose types involve @code{typename} or
1980 template template parameters can be mangled incorrectly.
1983 template <typename Q>
1984 void f(typename Q::X) @{@}
1986 template <template <typename> class Q>
1987 void f(typename Q<int>::X) @{@}
1991 Instantiations of these templates may be mangled incorrectly.
1995 @item -Wctor-dtor-privacy @r{(C++ only)}
1996 @opindex Wctor-dtor-privacy
1997 Warn when a class seems unusable because all the constructors or
1998 destructors in that class are private, and it has neither friends nor
1999 public static member functions.
2001 @item -Wnon-virtual-dtor @r{(C++ only)}
2002 @opindex Wnon-virtual-dtor
2003 Warn when a class appears to be polymorphic, thereby requiring a virtual
2004 destructor, yet it declares a non-virtual one. This warning is also
2005 enabled if -Weffc++ is specified.
2007 @item -Wreorder @r{(C++ only)}
2009 @cindex reordering, warning
2010 @cindex warning for reordering of member initializers
2011 Warn when the order of member initializers given in the code does not
2012 match the order in which they must be executed. For instance:
2018 A(): j (0), i (1) @{ @}
2022 The compiler will rearrange the member initializers for @samp{i}
2023 and @samp{j} to match the declaration order of the members, emitting
2024 a warning to that effect. This warning is enabled by @option{-Wall}.
2027 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2030 @item -Weffc++ @r{(C++ only)}
2032 Warn about violations of the following style guidelines from Scott Meyers'
2033 @cite{Effective C++} book:
2037 Item 11: Define a copy constructor and an assignment operator for classes
2038 with dynamically allocated memory.
2041 Item 12: Prefer initialization to assignment in constructors.
2044 Item 14: Make destructors virtual in base classes.
2047 Item 15: Have @code{operator=} return a reference to @code{*this}.
2050 Item 23: Don't try to return a reference when you must return an object.
2054 Also warn about violations of the following style guidelines from
2055 Scott Meyers' @cite{More Effective C++} book:
2059 Item 6: Distinguish between prefix and postfix forms of increment and
2060 decrement operators.
2063 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2067 When selecting this option, be aware that the standard library
2068 headers do not obey all of these guidelines; use @samp{grep -v}
2069 to filter out those warnings.
2071 @item -Wno-deprecated @r{(C++ only)}
2072 @opindex Wno-deprecated
2073 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
2075 @item -Wstrict-null-sentinel @r{(C++ only)}
2076 @opindex Wstrict-null-sentinel
2077 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2078 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2079 to @code{__null}. Although it is a null pointer constant not a null pointer,
2080 it is guaranteed to of the same size as a pointer. But this use is
2081 not portable across different compilers.
2083 @item -Wno-non-template-friend @r{(C++ only)}
2084 @opindex Wno-non-template-friend
2085 Disable warnings when non-templatized friend functions are declared
2086 within a template. Since the advent of explicit template specification
2087 support in G++, if the name of the friend is an unqualified-id (i.e.,
2088 @samp{friend foo(int)}), the C++ language specification demands that the
2089 friend declare or define an ordinary, nontemplate function. (Section
2090 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2091 could be interpreted as a particular specialization of a templatized
2092 function. Because this non-conforming behavior is no longer the default
2093 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2094 check existing code for potential trouble spots and is on by default.
2095 This new compiler behavior can be turned off with
2096 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2097 but disables the helpful warning.
2099 @item -Wold-style-cast @r{(C++ only)}
2100 @opindex Wold-style-cast
2101 Warn if an old-style (C-style) cast to a non-void type is used within
2102 a C++ program. The new-style casts (@samp{dynamic_cast},
2103 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2104 less vulnerable to unintended effects and much easier to search for.
2106 @item -Woverloaded-virtual @r{(C++ only)}
2107 @opindex Woverloaded-virtual
2108 @cindex overloaded virtual fn, warning
2109 @cindex warning for overloaded virtual fn
2110 Warn when a function declaration hides virtual functions from a
2111 base class. For example, in:
2118 struct B: public A @{
2123 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2131 will fail to compile.
2133 @item -Wno-pmf-conversions @r{(C++ only)}
2134 @opindex Wno-pmf-conversions
2135 Disable the diagnostic for converting a bound pointer to member function
2138 @item -Wsign-promo @r{(C++ only)}
2139 @opindex Wsign-promo
2140 Warn when overload resolution chooses a promotion from unsigned or
2141 enumerated type to a signed type, over a conversion to an unsigned type of
2142 the same size. Previous versions of G++ would try to preserve
2143 unsignedness, but the standard mandates the current behavior.
2148 A& operator = (int);
2158 In this example, G++ will synthesize a default @samp{A& operator =
2159 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2162 @node Objective-C and Objective-C++ Dialect Options
2163 @section Options Controlling Objective-C and Objective-C++ Dialects
2165 @cindex compiler options, Objective-C and Objective-C++
2166 @cindex Objective-C and Objective-C++ options, command line
2167 @cindex options, Objective-C and Objective-C++
2168 (NOTE: This manual does not describe the Objective-C and Objective-C++
2169 languages themselves. See @xref{Standards,,Language Standards
2170 Supported by GCC}, for references.)
2172 This section describes the command-line options that are only meaningful
2173 for Objective-C and Objective-C++ programs, but you can also use most of
2174 the language-independent GNU compiler options.
2175 For example, you might compile a file @code{some_class.m} like this:
2178 gcc -g -fgnu-runtime -O -c some_class.m
2182 In this example, @option{-fgnu-runtime} is an option meant only for
2183 Objective-C and Objective-C++ programs; you can use the other options with
2184 any language supported by GCC@.
2186 Note that since Objective-C is an extension of the C language, Objective-C
2187 compilations may also use options specific to the C front-end (e.g.,
2188 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2189 C++-specific options (e.g., @option{-Wabi}).
2191 Here is a list of options that are @emph{only} for compiling Objective-C
2192 and Objective-C++ programs:
2195 @item -fconstant-string-class=@var{class-name}
2196 @opindex fconstant-string-class
2197 Use @var{class-name} as the name of the class to instantiate for each
2198 literal string specified with the syntax @code{@@"@dots{}"}. The default
2199 class name is @code{NXConstantString} if the GNU runtime is being used, and
2200 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2201 @option{-fconstant-cfstrings} option, if also present, will override the
2202 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2203 to be laid out as constant CoreFoundation strings.
2206 @opindex fgnu-runtime
2207 Generate object code compatible with the standard GNU Objective-C
2208 runtime. This is the default for most types of systems.
2210 @item -fnext-runtime
2211 @opindex fnext-runtime
2212 Generate output compatible with the NeXT runtime. This is the default
2213 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2214 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2217 @item -fno-nil-receivers
2218 @opindex fno-nil-receivers
2219 Assume that all Objective-C message dispatches (e.g.,
2220 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2221 is not @code{nil}. This allows for more efficient entry points in the runtime
2222 to be used. Currently, this option is only available in conjunction with
2223 the NeXT runtime on Mac OS X 10.3 and later.
2225 @item -fobjc-call-cxx-cdtors
2226 @opindex fobjc-call-cxx-cdtors
2227 For each Objective-C class, check if any of its instance variables is a
2228 C++ object with a non-trivial default constructor. If so, synthesize a
2229 special @code{- (id) .cxx_construct} instance method that will run
2230 non-trivial default constructors on any such instance variables, in order,
2231 and then return @code{self}. Similarly, check if any instance variable
2232 is a C++ object with a non-trivial destructor, and if so, synthesize a
2233 special @code{- (void) .cxx_destruct} method that will run
2234 all such default destructors, in reverse order.
2236 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2237 thusly generated will only operate on instance variables declared in the
2238 current Objective-C class, and not those inherited from superclasses. It
2239 is the responsibility of the Objective-C runtime to invoke all such methods
2240 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2241 will be invoked by the runtime immediately after a new object
2242 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2243 be invoked immediately before the runtime deallocates an object instance.
2245 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2246 support for invoking the @code{- (id) .cxx_construct} and
2247 @code{- (void) .cxx_destruct} methods.
2249 @item -fobjc-direct-dispatch
2250 @opindex fobjc-direct-dispatch
2251 Allow fast jumps to the message dispatcher. On Darwin this is
2252 accomplished via the comm page.
2254 @item -fobjc-exceptions
2255 @opindex fobjc-exceptions
2256 Enable syntactic support for structured exception handling in Objective-C,
2257 similar to what is offered by C++ and Java. This option is
2258 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2267 @@catch (AnObjCClass *exc) @{
2274 @@catch (AnotherClass *exc) @{
2277 @@catch (id allOthers) @{
2287 The @code{@@throw} statement may appear anywhere in an Objective-C or
2288 Objective-C++ program; when used inside of a @code{@@catch} block, the
2289 @code{@@throw} may appear without an argument (as shown above), in which case
2290 the object caught by the @code{@@catch} will be rethrown.
2292 Note that only (pointers to) Objective-C objects may be thrown and
2293 caught using this scheme. When an object is thrown, it will be caught
2294 by the nearest @code{@@catch} clause capable of handling objects of that type,
2295 analogously to how @code{catch} blocks work in C++ and Java. A
2296 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2297 any and all Objective-C exceptions not caught by previous @code{@@catch}
2300 The @code{@@finally} clause, if present, will be executed upon exit from the
2301 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2302 regardless of whether any exceptions are thrown, caught or rethrown
2303 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2304 of the @code{finally} clause in Java.
2306 There are several caveats to using the new exception mechanism:
2310 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2311 idioms provided by the @code{NSException} class, the new
2312 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2313 systems, due to additional functionality needed in the (NeXT) Objective-C
2317 As mentioned above, the new exceptions do not support handling
2318 types other than Objective-C objects. Furthermore, when used from
2319 Objective-C++, the Objective-C exception model does not interoperate with C++
2320 exceptions at this time. This means you cannot @code{@@throw} an exception
2321 from Objective-C and @code{catch} it in C++, or vice versa
2322 (i.e., @code{throw @dots{} @@catch}).
2325 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2326 blocks for thread-safe execution:
2329 @@synchronized (ObjCClass *guard) @{
2334 Upon entering the @code{@@synchronized} block, a thread of execution shall
2335 first check whether a lock has been placed on the corresponding @code{guard}
2336 object by another thread. If it has, the current thread shall wait until
2337 the other thread relinquishes its lock. Once @code{guard} becomes available,
2338 the current thread will place its own lock on it, execute the code contained in
2339 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2340 making @code{guard} available to other threads).
2342 Unlike Java, Objective-C does not allow for entire methods to be marked
2343 @code{@@synchronized}. Note that throwing exceptions out of
2344 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2345 to be unlocked properly.
2349 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2351 @item -freplace-objc-classes
2352 @opindex freplace-objc-classes
2353 Emit a special marker instructing @command{ld(1)} not to statically link in
2354 the resulting object file, and allow @command{dyld(1)} to load it in at
2355 run time instead. This is used in conjunction with the Fix-and-Continue
2356 debugging mode, where the object file in question may be recompiled and
2357 dynamically reloaded in the course of program execution, without the need
2358 to restart the program itself. Currently, Fix-and-Continue functionality
2359 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2364 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2365 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2366 compile time) with static class references that get initialized at load time,
2367 which improves run-time performance. Specifying the @option{-fzero-link} flag
2368 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2369 to be retained. This is useful in Zero-Link debugging mode, since it allows
2370 for individual class implementations to be modified during program execution.
2374 Dump interface declarations for all classes seen in the source file to a
2375 file named @file{@var{sourcename}.decl}.
2377 @item -Wassign-intercept
2378 @opindex Wassign-intercept
2379 Warn whenever an Objective-C assignment is being intercepted by the
2383 @opindex Wno-protocol
2384 If a class is declared to implement a protocol, a warning is issued for
2385 every method in the protocol that is not implemented by the class. The
2386 default behavior is to issue a warning for every method not explicitly
2387 implemented in the class, even if a method implementation is inherited
2388 from the superclass. If you use the @option{-Wno-protocol} option, then
2389 methods inherited from the superclass are considered to be implemented,
2390 and no warning is issued for them.
2394 Warn if multiple methods of different types for the same selector are
2395 found during compilation. The check is performed on the list of methods
2396 in the final stage of compilation. Additionally, a check is performed
2397 for each selector appearing in a @code{@@selector(@dots{})}
2398 expression, and a corresponding method for that selector has been found
2399 during compilation. Because these checks scan the method table only at
2400 the end of compilation, these warnings are not produced if the final
2401 stage of compilation is not reached, for example because an error is
2402 found during compilation, or because the @option{-fsyntax-only} option is
2405 @item -Wstrict-selector-match
2406 @opindex Wstrict-selector-match
2407 Warn if multiple methods with differing argument and/or return types are
2408 found for a given selector when attempting to send a message using this
2409 selector to a receiver of type @code{id} or @code{Class}. When this flag
2410 is off (which is the default behavior), the compiler will omit such warnings
2411 if any differences found are confined to types which share the same size
2414 @item -Wundeclared-selector
2415 @opindex Wundeclared-selector
2416 Warn if a @code{@@selector(@dots{})} expression referring to an
2417 undeclared selector is found. A selector is considered undeclared if no
2418 method with that name has been declared before the
2419 @code{@@selector(@dots{})} expression, either explicitly in an
2420 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2421 an @code{@@implementation} section. This option always performs its
2422 checks as soon as a @code{@@selector(@dots{})} expression is found,
2423 while @option{-Wselector} only performs its checks in the final stage of
2424 compilation. This also enforces the coding style convention
2425 that methods and selectors must be declared before being used.
2427 @item -print-objc-runtime-info
2428 @opindex print-objc-runtime-info
2429 Generate C header describing the largest structure that is passed by
2434 @node Language Independent Options
2435 @section Options to Control Diagnostic Messages Formatting
2436 @cindex options to control diagnostics formatting
2437 @cindex diagnostic messages
2438 @cindex message formatting
2440 Traditionally, diagnostic messages have been formatted irrespective of
2441 the output device's aspect (e.g.@: its width, @dots{}). The options described
2442 below can be used to control the diagnostic messages formatting
2443 algorithm, e.g.@: how many characters per line, how often source location
2444 information should be reported. Right now, only the C++ front end can
2445 honor these options. However it is expected, in the near future, that
2446 the remaining front ends would be able to digest them correctly.
2449 @item -fmessage-length=@var{n}
2450 @opindex fmessage-length
2451 Try to format error messages so that they fit on lines of about @var{n}
2452 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2453 the front ends supported by GCC@. If @var{n} is zero, then no
2454 line-wrapping will be done; each error message will appear on a single
2457 @opindex fdiagnostics-show-location
2458 @item -fdiagnostics-show-location=once
2459 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2460 reporter to emit @emph{once} source location information; that is, in
2461 case the message is too long to fit on a single physical line and has to
2462 be wrapped, the source location won't be emitted (as prefix) again,
2463 over and over, in subsequent continuation lines. This is the default
2466 @item -fdiagnostics-show-location=every-line
2467 Only meaningful in line-wrapping mode. Instructs the diagnostic
2468 messages reporter to emit the same source location information (as
2469 prefix) for physical lines that result from the process of breaking
2470 a message which is too long to fit on a single line.
2472 @item -fdiagnostics-show-option
2473 @opindex fdiagnostics-show-option
2474 This option instructs the diagnostic machinery to add text to each
2475 diagnostic emitted, which indicates which command line option directly
2476 controls that diagnostic, when such an option is known to the
2477 diagnostic machinery.
2479 @item -Wcoverage-mismatch
2480 @opindex Wcoverage-mismatch
2481 Warn if feedback profiles do not match when using the
2482 @option{-fprofile-use} option.
2483 If a source file was changed between @option{-fprofile-gen} and
2484 @option{-fprofile-use}, the files with the profile feedback can fail
2485 to match the source file and GCC can not use the profile feedback
2486 information. By default, GCC emits an error message in this case.
2487 The option @option{-Wcoverage-mismatch} emits a warning instead of an
2488 error. GCC does not use appropriate feedback profiles, so using this
2489 option can result in poorly optimized code. This option is useful
2490 only in the case of very minor changes such as bug fixes to an
2495 @node Warning Options
2496 @section Options to Request or Suppress Warnings
2497 @cindex options to control warnings
2498 @cindex warning messages
2499 @cindex messages, warning
2500 @cindex suppressing warnings
2502 Warnings are diagnostic messages that report constructions which
2503 are not inherently erroneous but which are risky or suggest there
2504 may have been an error.
2506 You can request many specific warnings with options beginning @samp{-W},
2507 for example @option{-Wimplicit} to request warnings on implicit
2508 declarations. Each of these specific warning options also has a
2509 negative form beginning @samp{-Wno-} to turn off warnings;
2510 for example, @option{-Wno-implicit}. This manual lists only one of the
2511 two forms, whichever is not the default.
2513 The following options control the amount and kinds of warnings produced
2514 by GCC; for further, language-specific options also refer to
2515 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2519 @cindex syntax checking
2521 @opindex fsyntax-only
2522 Check the code for syntax errors, but don't do anything beyond that.
2526 Issue all the warnings demanded by strict ISO C and ISO C++;
2527 reject all programs that use forbidden extensions, and some other
2528 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2529 version of the ISO C standard specified by any @option{-std} option used.
2531 Valid ISO C and ISO C++ programs should compile properly with or without
2532 this option (though a rare few will require @option{-ansi} or a
2533 @option{-std} option specifying the required version of ISO C)@. However,
2534 without this option, certain GNU extensions and traditional C and C++
2535 features are supported as well. With this option, they are rejected.
2537 @option{-pedantic} does not cause warning messages for use of the
2538 alternate keywords whose names begin and end with @samp{__}. Pedantic
2539 warnings are also disabled in the expression that follows
2540 @code{__extension__}. However, only system header files should use
2541 these escape routes; application programs should avoid them.
2542 @xref{Alternate Keywords}.
2544 Some users try to use @option{-pedantic} to check programs for strict ISO
2545 C conformance. They soon find that it does not do quite what they want:
2546 it finds some non-ISO practices, but not all---only those for which
2547 ISO C @emph{requires} a diagnostic, and some others for which
2548 diagnostics have been added.
2550 A feature to report any failure to conform to ISO C might be useful in
2551 some instances, but would require considerable additional work and would
2552 be quite different from @option{-pedantic}. We don't have plans to
2553 support such a feature in the near future.
2555 Where the standard specified with @option{-std} represents a GNU
2556 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2557 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2558 extended dialect is based. Warnings from @option{-pedantic} are given
2559 where they are required by the base standard. (It would not make sense
2560 for such warnings to be given only for features not in the specified GNU
2561 C dialect, since by definition the GNU dialects of C include all
2562 features the compiler supports with the given option, and there would be
2563 nothing to warn about.)
2565 @item -pedantic-errors
2566 @opindex pedantic-errors
2567 Like @option{-pedantic}, except that errors are produced rather than
2572 Inhibit all warning messages.
2576 Inhibit warning messages about the use of @samp{#import}.
2578 @item -Wchar-subscripts
2579 @opindex Wchar-subscripts
2580 Warn if an array subscript has type @code{char}. This is a common cause
2581 of error, as programmers often forget that this type is signed on some
2583 This warning is enabled by @option{-Wall}.
2587 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2588 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2589 This warning is enabled by @option{-Wall}.
2591 @item -Wfatal-errors
2592 @opindex Wfatal-errors
2593 This option causes the compiler to abort compilation on the first error
2594 occurred rather than trying to keep going and printing further error
2599 @opindex ffreestanding
2600 @opindex fno-builtin
2601 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2602 the arguments supplied have types appropriate to the format string
2603 specified, and that the conversions specified in the format string make
2604 sense. This includes standard functions, and others specified by format
2605 attributes (@pxref{Function Attributes}), in the @code{printf},
2606 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2607 not in the C standard) families (or other target-specific families).
2608 Which functions are checked without format attributes having been
2609 specified depends on the standard version selected, and such checks of
2610 functions without the attribute specified are disabled by
2611 @option{-ffreestanding} or @option{-fno-builtin}.
2613 The formats are checked against the format features supported by GNU
2614 libc version 2.2. These include all ISO C90 and C99 features, as well
2615 as features from the Single Unix Specification and some BSD and GNU
2616 extensions. Other library implementations may not support all these
2617 features; GCC does not support warning about features that go beyond a
2618 particular library's limitations. However, if @option{-pedantic} is used
2619 with @option{-Wformat}, warnings will be given about format features not
2620 in the selected standard version (but not for @code{strfmon} formats,
2621 since those are not in any version of the C standard). @xref{C Dialect
2622 Options,,Options Controlling C Dialect}.
2624 Since @option{-Wformat} also checks for null format arguments for
2625 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2627 @option{-Wformat} is included in @option{-Wall}. For more control over some
2628 aspects of format checking, the options @option{-Wformat-y2k},
2629 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2630 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2631 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2634 @opindex Wformat-y2k
2635 If @option{-Wformat} is specified, also warn about @code{strftime}
2636 formats which may yield only a two-digit year.
2638 @item -Wno-format-extra-args
2639 @opindex Wno-format-extra-args
2640 If @option{-Wformat} is specified, do not warn about excess arguments to a
2641 @code{printf} or @code{scanf} format function. The C standard specifies
2642 that such arguments are ignored.
2644 Where the unused arguments lie between used arguments that are
2645 specified with @samp{$} operand number specifications, normally
2646 warnings are still given, since the implementation could not know what
2647 type to pass to @code{va_arg} to skip the unused arguments. However,
2648 in the case of @code{scanf} formats, this option will suppress the
2649 warning if the unused arguments are all pointers, since the Single
2650 Unix Specification says that such unused arguments are allowed.
2652 @item -Wno-format-zero-length
2653 @opindex Wno-format-zero-length
2654 If @option{-Wformat} is specified, do not warn about zero-length formats.
2655 The C standard specifies that zero-length formats are allowed.
2657 @item -Wformat-nonliteral
2658 @opindex Wformat-nonliteral
2659 If @option{-Wformat} is specified, also warn if the format string is not a
2660 string literal and so cannot be checked, unless the format function
2661 takes its format arguments as a @code{va_list}.
2663 @item -Wformat-security
2664 @opindex Wformat-security
2665 If @option{-Wformat} is specified, also warn about uses of format
2666 functions that represent possible security problems. At present, this
2667 warns about calls to @code{printf} and @code{scanf} functions where the
2668 format string is not a string literal and there are no format arguments,
2669 as in @code{printf (foo);}. This may be a security hole if the format
2670 string came from untrusted input and contains @samp{%n}. (This is
2671 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2672 in future warnings may be added to @option{-Wformat-security} that are not
2673 included in @option{-Wformat-nonliteral}.)
2677 Enable @option{-Wformat} plus format checks not included in
2678 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2679 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2683 Warn about passing a null pointer for arguments marked as
2684 requiring a non-null value by the @code{nonnull} function attribute.
2686 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2687 can be disabled with the @option{-Wno-nonnull} option.
2689 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2691 Warn about uninitialized variables which are initialized with themselves.
2692 Note this option can only be used with the @option{-Wuninitialized} option,
2693 which in turn only works with @option{-O1} and above.
2695 For example, GCC will warn about @code{i} being uninitialized in the
2696 following snippet only when @option{-Winit-self} has been specified:
2707 @item -Wimplicit-int
2708 @opindex Wimplicit-int
2709 Warn when a declaration does not specify a type.
2710 This warning is enabled by @option{-Wall}.
2712 @item -Wimplicit-function-declaration
2713 @opindex Wimplicit-function-declaration
2714 @opindex Wno-implicit-function-declaration
2715 Give a warning whenever a function is used before being declared. In
2716 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
2717 enabled by default and it is made into an error by
2718 @option{-pedantic-errors}. This warning is also enabled by
2723 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2724 This warning is enabled by @option{-Wall}.
2728 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2729 function with external linkage, returning int, taking either zero
2730 arguments, two, or three arguments of appropriate types.
2731 This warning is enabled by @option{-Wall}.
2733 @item -Wmissing-braces
2734 @opindex Wmissing-braces
2735 Warn if an aggregate or union initializer is not fully bracketed. In
2736 the following example, the initializer for @samp{a} is not fully
2737 bracketed, but that for @samp{b} is fully bracketed.
2740 int a[2][2] = @{ 0, 1, 2, 3 @};
2741 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2744 This warning is enabled by @option{-Wall}.
2746 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2747 @opindex Wmissing-include-dirs
2748 Warn if a user-supplied include directory does not exist.
2751 @opindex Wparentheses
2752 Warn if parentheses are omitted in certain contexts, such
2753 as when there is an assignment in a context where a truth value
2754 is expected, or when operators are nested whose precedence people
2755 often get confused about.
2757 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2758 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2759 interpretation from that of ordinary mathematical notation.
2761 Also warn about constructions where there may be confusion to which
2762 @code{if} statement an @code{else} branch belongs. Here is an example of
2777 In C/C++, every @code{else} branch belongs to the innermost possible
2778 @code{if} statement, which in this example is @code{if (b)}. This is
2779 often not what the programmer expected, as illustrated in the above
2780 example by indentation the programmer chose. When there is the
2781 potential for this confusion, GCC will issue a warning when this flag
2782 is specified. To eliminate the warning, add explicit braces around
2783 the innermost @code{if} statement so there is no way the @code{else}
2784 could belong to the enclosing @code{if}. The resulting code would
2801 This warning is enabled by @option{-Wall}.
2803 @item -Wsequence-point
2804 @opindex Wsequence-point
2805 Warn about code that may have undefined semantics because of violations
2806 of sequence point rules in the C and C++ standards.
2808 The C and C++ standards defines the order in which expressions in a C/C++
2809 program are evaluated in terms of @dfn{sequence points}, which represent
2810 a partial ordering between the execution of parts of the program: those
2811 executed before the sequence point, and those executed after it. These
2812 occur after the evaluation of a full expression (one which is not part
2813 of a larger expression), after the evaluation of the first operand of a
2814 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2815 function is called (but after the evaluation of its arguments and the
2816 expression denoting the called function), and in certain other places.
2817 Other than as expressed by the sequence point rules, the order of
2818 evaluation of subexpressions of an expression is not specified. All
2819 these rules describe only a partial order rather than a total order,
2820 since, for example, if two functions are called within one expression
2821 with no sequence point between them, the order in which the functions
2822 are called is not specified. However, the standards committee have
2823 ruled that function calls do not overlap.
2825 It is not specified when between sequence points modifications to the
2826 values of objects take effect. Programs whose behavior depends on this
2827 have undefined behavior; the C and C++ standards specify that ``Between
2828 the previous and next sequence point an object shall have its stored
2829 value modified at most once by the evaluation of an expression.
2830 Furthermore, the prior value shall be read only to determine the value
2831 to be stored.''. If a program breaks these rules, the results on any
2832 particular implementation are entirely unpredictable.
2834 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2835 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2836 diagnosed by this option, and it may give an occasional false positive
2837 result, but in general it has been found fairly effective at detecting
2838 this sort of problem in programs.
2840 The standard is worded confusingly, therefore there is some debate
2841 over the precise meaning of the sequence point rules in subtle cases.
2842 Links to discussions of the problem, including proposed formal
2843 definitions, may be found on the GCC readings page, at
2844 @w{@uref{http://gcc.gnu.org/readings.html}}.
2846 This warning is enabled by @option{-Wall} for C and C++.
2849 @opindex Wreturn-type
2850 @opindex Wno-return-type
2851 Warn whenever a function is defined with a return-type that defaults
2852 to @code{int}. Also warn about any @code{return} statement with no
2853 return-value in a function whose return-type is not @code{void}
2854 (falling off the end of the function body is considered returning
2855 without a value), and about a @code{return} statement with a
2856 expression in a function whose return-type is @code{void}.
2858 Also warn if the return type of a function has a type qualifier
2859 such as @code{const}. For ISO C such a type qualifier has no effect,
2860 since the value returned by a function is not an lvalue.
2861 For C++, the warning is only emitted for scalar types or @code{void}.
2862 ISO C prohibits qualified @code{void} return types on function
2863 definitions, so such return types always receive a warning
2864 even without this option.
2866 For C++, a function without return type always produces a diagnostic
2867 message, even when @option{-Wno-return-type} is specified. The only
2868 exceptions are @samp{main} and functions defined in system headers.
2870 This warning is enabled by @option{-Wall}.
2874 Warn whenever a @code{switch} statement has an index of enumerated type
2875 and lacks a @code{case} for one or more of the named codes of that
2876 enumeration. (The presence of a @code{default} label prevents this
2877 warning.) @code{case} labels outside the enumeration range also
2878 provoke warnings when this option is used.
2879 This warning is enabled by @option{-Wall}.
2881 @item -Wswitch-default
2882 @opindex Wswitch-switch
2883 Warn whenever a @code{switch} statement does not have a @code{default}
2887 @opindex Wswitch-enum
2888 Warn whenever a @code{switch} statement has an index of enumerated type
2889 and lacks a @code{case} for one or more of the named codes of that
2890 enumeration. @code{case} labels outside the enumeration range also
2891 provoke warnings when this option is used.
2895 Warn if any trigraphs are encountered that might change the meaning of
2896 the program (trigraphs within comments are not warned about).
2897 This warning is enabled by @option{-Wall}.
2899 @item -Wunused-function
2900 @opindex Wunused-function
2901 Warn whenever a static function is declared but not defined or a
2902 non-inline static function is unused.
2903 This warning is enabled by @option{-Wall}.
2905 @item -Wunused-label
2906 @opindex Wunused-label
2907 Warn whenever a label is declared but not used.
2908 This warning is enabled by @option{-Wall}.
2910 To suppress this warning use the @samp{unused} attribute
2911 (@pxref{Variable Attributes}).
2913 @item -Wunused-parameter
2914 @opindex Wunused-parameter
2915 Warn whenever a function parameter is unused aside from its declaration.
2917 To suppress this warning use the @samp{unused} attribute
2918 (@pxref{Variable Attributes}).
2920 @item -Wunused-variable
2921 @opindex Wunused-variable
2922 Warn whenever a local variable or non-constant static variable is unused
2923 aside from its declaration.
2924 This warning is enabled by @option{-Wall}.
2926 To suppress this warning use the @samp{unused} attribute
2927 (@pxref{Variable Attributes}).
2929 @item -Wunused-value
2930 @opindex Wunused-value
2931 Warn whenever a statement computes a result that is explicitly not
2932 used. To suppress this warning cast the unused expression to
2933 @samp{void}. This includes an expression-statement or the left-hand
2934 side of a comma expression that contains no side effects. For example,
2935 an expression such as @samp{x[i,j]} will cause a warning, while
2936 @samp{x[(void)i,j]} will not.
2938 This warning is enabled by @option{-Wall}.
2942 All the above @option{-Wunused} options combined.
2944 In order to get a warning about an unused function parameter, you must
2945 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2946 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2948 @item -Wuninitialized
2949 @opindex Wuninitialized
2950 Warn if an automatic variable is used without first being initialized or
2951 if a variable may be clobbered by a @code{setjmp} call.
2953 These warnings are possible only in optimizing compilation,
2954 because they require data flow information that is computed only
2955 when optimizing. If you do not specify @option{-O}, you will not get
2956 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2957 requiring @option{-O}.
2959 If you want to warn about code which uses the uninitialized value of the
2960 variable in its own initializer, use the @option{-Winit-self} option.
2962 These warnings occur for individual uninitialized or clobbered
2963 elements of structure, union or array variables as well as for
2964 variables which are uninitialized or clobbered as a whole. They do
2965 not occur for variables or elements declared @code{volatile}. Because
2966 these warnings depend on optimization, the exact variables or elements
2967 for which there are warnings will depend on the precise optimization
2968 options and version of GCC used.
2970 Note that there may be no warning about a variable that is used only
2971 to compute a value that itself is never used, because such
2972 computations may be deleted by data flow analysis before the warnings
2975 These warnings are made optional because GCC is not smart
2976 enough to see all the reasons why the code might be correct
2977 despite appearing to have an error. Here is one example of how
2998 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2999 always initialized, but GCC doesn't know this. Here is
3000 another common case:
3005 if (change_y) save_y = y, y = new_y;
3007 if (change_y) y = save_y;
3012 This has no bug because @code{save_y} is used only if it is set.
3014 @cindex @code{longjmp} warnings
3015 This option also warns when a non-volatile automatic variable might be
3016 changed by a call to @code{longjmp}. These warnings as well are possible
3017 only in optimizing compilation.
3019 The compiler sees only the calls to @code{setjmp}. It cannot know
3020 where @code{longjmp} will be called; in fact, a signal handler could
3021 call it at any point in the code. As a result, you may get a warning
3022 even when there is in fact no problem because @code{longjmp} cannot
3023 in fact be called at the place which would cause a problem.
3025 Some spurious warnings can be avoided if you declare all the functions
3026 you use that never return as @code{noreturn}. @xref{Function
3029 This warning is enabled by @option{-Wall}.
3031 @item -Wunknown-pragmas
3032 @opindex Wunknown-pragmas
3033 @cindex warning for unknown pragmas
3034 @cindex unknown pragmas, warning
3035 @cindex pragmas, warning of unknown
3036 Warn when a #pragma directive is encountered which is not understood by
3037 GCC@. If this command line option is used, warnings will even be issued
3038 for unknown pragmas in system header files. This is not the case if
3039 the warnings were only enabled by the @option{-Wall} command line option.
3042 @opindex Wno-pragmas
3044 Do not warn about misuses of pragmas, such as incorrect parameters,
3045 invalid syntax, or conflicts between pragmas. See also
3046 @samp{-Wunknown-pragmas}.
3048 @item -Wstrict-aliasing
3049 @opindex Wstrict-aliasing
3050 This option is only active when @option{-fstrict-aliasing} is active.
3051 It warns about code which might break the strict aliasing rules that the
3052 compiler is using for optimization. The warning does not catch all
3053 cases, but does attempt to catch the more common pitfalls. It is
3054 included in @option{-Wall}.
3055 It is equivalent to -Wstrict-aliasing=3
3057 @item -Wstrict-aliasing=n
3058 @opindex Wstrict-aliasing=n
3059 This option is only active when @option{-fstrict-aliasing} is active.
3060 It warns about code which might break the strict aliasing rules that the
3061 compiler is using for optimization.
3062 Higher levels correspond to higher accuracy (fewer false positives).
3063 Higher levels also correspond to more effort, similar to the way -O works.
3064 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3067 Level 1: Most aggressive, quick, least accurate.
3068 Possibly useful when higher levels
3069 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3070 false negatives. However, it has many false positives.
3071 Warns for all pointer conversions between possibly incompatible types,
3072 even if never dereferenced. Runs in the frontend only.
3074 Level 2: Aggressive, quick, not too precise.
3075 May still have many false positives (not as many as level 1 though),
3076 and few false negatives (but possibly more than level 1).
3077 Unlike level 1, it only warns when an address is taken. Warns about
3078 incomplete types. Runs in the frontend only.
3080 Level 3 (default for @option{-Wstrict-aliasing}):
3081 Should have very few false positives and few false
3082 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3083 Takes care of the common punn+dereference pattern in the frontend:
3084 @code{*(int*)&some_float}.
3085 If optimization is enabled, it also runs in the backend, where it deals
3086 with multiple statement cases using flow-sensitive points-to information.
3087 Only warns when the converted pointer is dereferenced.
3088 Does not warn about incomplete types.
3090 @item -Wstrict-overflow
3091 @item -Wstrict-overflow=@var{n}
3092 @opindex Wstrict-overflow
3093 This option is only active when @option{-fstrict-overflow} is active.
3094 It warns about cases where the compiler optimizes based on the
3095 assumption that signed overflow does not occur. Note that it does not
3096 warn about all cases where the code might overflow: it only warns
3097 about cases where the compiler implements some optimization. Thus
3098 this warning depends on the optimization level.
3100 An optimization which assumes that signed overflow does not occur is
3101 perfectly safe if the values of the variables involved are such that
3102 overflow never does, in fact, occur. Therefore this warning can
3103 easily give a false positive: a warning about code which is not
3104 actually a problem. To help focus on important issues, several
3105 warning levels are defined. No warnings are issued for the use of
3106 undefined signed overflow when estimating how many iterations a loop
3107 will require, in particular when determining whether a loop will be
3111 @item -Wstrict-overflow=1
3112 Warn about cases which are both questionable and easy to avoid. For
3113 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3114 compiler will simplify this to @code{1}. This level of
3115 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3116 are not, and must be explicitly requested.
3118 @item -Wstrict-overflow=2
3119 Also warn about other cases where a comparison is simplified to a
3120 constant. For example: @code{abs (x) >= 0}. This can only be
3121 simplified when @option{-fstrict-overflow} is in effect, because
3122 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3123 zero. @option{-Wstrict-overflow} (with no level) is the same as
3124 @option{-Wstrict-overflow=2}.
3126 @item -Wstrict-overflow=3
3127 Also warn about other cases where a comparison is simplified. For
3128 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3130 @item -Wstrict-overflow=4
3131 Also warn about other simplifications not covered by the above cases.
3132 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3134 @item -Wstrict-overflow=5
3135 Also warn about cases where the compiler reduces the magnitude of a
3136 constant involved in a comparison. For example: @code{x + 2 > y} will
3137 be simplified to @code{x + 1 >= y}. This is reported only at the
3138 highest warning level because this simplification applies to many
3139 comparisons, so this warning level will give a very large number of
3143 @item -Warray-bounds
3144 @opindex Wno-array-bounds
3145 @opindex Warray-bounds
3146 This option is only active when @option{-ftree-vrp} is active
3147 (default for -O2 and above). It warns about subscripts to arrays
3148 that are always out of bounds. This warning is enabled by @option{-Wall}.
3152 All of the above @samp{-W} options combined. This enables all the
3153 warnings about constructions that some users consider questionable, and
3154 that are easy to avoid (or modify to prevent the warning), even in
3155 conjunction with macros. This also enables some language-specific
3156 warnings described in @ref{C++ Dialect Options} and
3157 @ref{Objective-C and Objective-C++ Dialect Options}.
3160 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
3161 Some of them warn about constructions that users generally do not
3162 consider questionable, but which occasionally you might wish to check
3163 for; others warn about constructions that are necessary or hard to avoid
3164 in some cases, and there is no simple way to modify the code to suppress
3171 (This option used to be called @option{-W}. The older name is still
3172 supported, but the newer name is more descriptive.) Print extra warning
3173 messages for these events:
3177 Warn if a comparison is always true or always false due to the limited
3178 range of the data type, but do not warn for constant expressions. For
3179 example, warn if an unsigned variable is compared against zero with
3180 @samp{<} or @samp{>=}. This warning can be independently controlled
3181 by @option{-Wtype-limits}.
3184 Storage-class specifiers like @code{static} are not the first things
3185 in a declaration. According to the C Standard, this usage is
3186 obsolescent. This warning can be independently controlled by
3187 @option{-Wold-style-declaration}.
3190 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
3194 A comparison between signed and unsigned values could produce an
3195 incorrect result when the signed value is converted to unsigned.
3196 (But don't warn if @option{-Wno-sign-compare} is also specified.)
3199 An aggregate has an initializer which does not initialize all members.
3200 This warning can be independently controlled by
3201 @option{-Wmissing-field-initializers}.
3204 An initialized field without side effects is overridden when using
3205 designated initializers (@pxref{Designated Inits, , Designated
3206 Initializers}). This warning can be independently controlled by
3207 @option{-Woverride-init}.
3210 A function parameter is declared without a type specifier in K&R-style
3211 functions. This warning can be independently controlled by
3212 @option{-Wmissing-parameter-type}.
3215 An empty body occurs in an @samp{if}, @samp{else} or
3216 @samp{do while} statement. This warning can be independently
3217 controlled by @option{-Wempty-body}.
3219 @item @r{(C++ only)}
3220 An empty body occurs in a @samp{while} or @samp{for} statement with no
3221 whitespacing before the semicolon. This warning can be independently
3222 controlled by @option{-Wempty-body}.
3225 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3226 @samp{>}, or @samp{>=}.
3229 A variable might be changed by @samp{longjmp} or @samp{vfork}.
3230 This warning can be independently controlled by @option{-Wclobbered}.
3232 @item @r{(C++ only)}
3233 An enumerator and a non-enumerator both appear in a conditional expression.
3235 @item @r{(C++ only)}
3236 A non-static reference or non-static @samp{const} member appears in a
3237 class without constructors.
3239 @item @r{(C++ only)}
3240 Ambiguous virtual bases.
3242 @item @r{(C++ only)}
3243 Subscripting an array which has been declared @samp{register}.
3245 @item @r{(C++ only)}
3246 Taking the address of a variable which has been declared @samp{register}.
3248 @item @r{(C++ only)}
3249 A base class is not initialized in a derived class' copy constructor.
3252 @item -Wno-div-by-zero
3253 @opindex Wno-div-by-zero
3254 @opindex Wdiv-by-zero
3255 Do not warn about compile-time integer division by zero. Floating point
3256 division by zero is not warned about, as it can be a legitimate way of
3257 obtaining infinities and NaNs.
3259 @item -Wsystem-headers
3260 @opindex Wsystem-headers
3261 @cindex warnings from system headers
3262 @cindex system headers, warnings from
3263 Print warning messages for constructs found in system header files.
3264 Warnings from system headers are normally suppressed, on the assumption
3265 that they usually do not indicate real problems and would only make the
3266 compiler output harder to read. Using this command line option tells
3267 GCC to emit warnings from system headers as if they occurred in user
3268 code. However, note that using @option{-Wall} in conjunction with this
3269 option will @emph{not} warn about unknown pragmas in system
3270 headers---for that, @option{-Wunknown-pragmas} must also be used.
3273 @opindex Wfloat-equal
3274 Warn if floating point values are used in equality comparisons.
3276 The idea behind this is that sometimes it is convenient (for the
3277 programmer) to consider floating-point values as approximations to
3278 infinitely precise real numbers. If you are doing this, then you need
3279 to compute (by analyzing the code, or in some other way) the maximum or
3280 likely maximum error that the computation introduces, and allow for it
3281 when performing comparisons (and when producing output, but that's a
3282 different problem). In particular, instead of testing for equality, you
3283 would check to see whether the two values have ranges that overlap; and
3284 this is done with the relational operators, so equality comparisons are
3287 @item -Wtraditional @r{(C only)}
3288 @opindex Wtraditional
3289 Warn about certain constructs that behave differently in traditional and
3290 ISO C@. Also warn about ISO C constructs that have no traditional C
3291 equivalent, and/or problematic constructs which should be avoided.
3295 Macro parameters that appear within string literals in the macro body.
3296 In traditional C macro replacement takes place within string literals,
3297 but does not in ISO C@.
3300 In traditional C, some preprocessor directives did not exist.
3301 Traditional preprocessors would only consider a line to be a directive
3302 if the @samp{#} appeared in column 1 on the line. Therefore
3303 @option{-Wtraditional} warns about directives that traditional C
3304 understands but would ignore because the @samp{#} does not appear as the
3305 first character on the line. It also suggests you hide directives like
3306 @samp{#pragma} not understood by traditional C by indenting them. Some
3307 traditional implementations would not recognize @samp{#elif}, so it
3308 suggests avoiding it altogether.
3311 A function-like macro that appears without arguments.
3314 The unary plus operator.
3317 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3318 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3319 constants.) Note, these suffixes appear in macros defined in the system
3320 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3321 Use of these macros in user code might normally lead to spurious
3322 warnings, however GCC's integrated preprocessor has enough context to
3323 avoid warning in these cases.
3326 A function declared external in one block and then used after the end of
3330 A @code{switch} statement has an operand of type @code{long}.
3333 A non-@code{static} function declaration follows a @code{static} one.
3334 This construct is not accepted by some traditional C compilers.
3337 The ISO type of an integer constant has a different width or
3338 signedness from its traditional type. This warning is only issued if
3339 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3340 typically represent bit patterns, are not warned about.
3343 Usage of ISO string concatenation is detected.
3346 Initialization of automatic aggregates.
3349 Identifier conflicts with labels. Traditional C lacks a separate
3350 namespace for labels.
3353 Initialization of unions. If the initializer is zero, the warning is
3354 omitted. This is done under the assumption that the zero initializer in
3355 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3356 initializer warnings and relies on default initialization to zero in the
3360 Conversions by prototypes between fixed/floating point values and vice
3361 versa. The absence of these prototypes when compiling with traditional
3362 C would cause serious problems. This is a subset of the possible
3363 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3366 Use of ISO C style function definitions. This warning intentionally is
3367 @emph{not} issued for prototype declarations or variadic functions
3368 because these ISO C features will appear in your code when using
3369 libiberty's traditional C compatibility macros, @code{PARAMS} and
3370 @code{VPARAMS}. This warning is also bypassed for nested functions
3371 because that feature is already a GCC extension and thus not relevant to
3372 traditional C compatibility.
3375 @item -Wtraditional-conversion @r{(C only)}
3376 @opindex Wtraditional-conversion
3377 Warn if a prototype causes a type conversion that is different from what
3378 would happen to the same argument in the absence of a prototype. This
3379 includes conversions of fixed point to floating and vice versa, and
3380 conversions changing the width or signedness of a fixed point argument
3381 except when the same as the default promotion.
3383 @item -Wdeclaration-after-statement @r{(C only)}
3384 @opindex Wdeclaration-after-statement
3385 Warn when a declaration is found after a statement in a block. This
3386 construct, known from C++, was introduced with ISO C99 and is by default
3387 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3388 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3392 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3394 @item -Wno-endif-labels
3395 @opindex Wno-endif-labels
3396 @opindex Wendif-labels
3397 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3401 Warn whenever a local variable shadows another local variable, parameter or
3402 global variable or whenever a built-in function is shadowed.
3404 @item -Wlarger-than-@var{len}
3405 @opindex Wlarger-than
3406 Warn whenever an object of larger than @var{len} bytes is defined.
3408 @item -Wunsafe-loop-optimizations
3409 @opindex Wunsafe-loop-optimizations
3410 Warn if the loop cannot be optimized because the compiler could not
3411 assume anything on the bounds of the loop indices. With
3412 @option{-funsafe-loop-optimizations} warn if the compiler made
3415 @item -Wpointer-arith
3416 @opindex Wpointer-arith
3417 Warn about anything that depends on the ``size of'' a function type or
3418 of @code{void}. GNU C assigns these types a size of 1, for
3419 convenience in calculations with @code{void *} pointers and pointers
3420 to functions. In C++, warn also when an arithmetic operation involves
3421 @code{NULL}. This warning is also enabled by @option{-pedantic}.
3424 @opindex Wtype-limits
3425 @opindex Wno-type-limits
3426 Warn if a comparison is always true or always false due to the limited
3427 range of the data type, but do not warn for constant expressions. For
3428 example, warn if an unsigned variable is compared against zero with
3429 @samp{<} or @samp{>=}. This warning is also enabled by
3432 @item -Wbad-function-cast @r{(C only)}
3433 @opindex Wbad-function-cast
3434 Warn whenever a function call is cast to a non-matching type.
3435 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3438 Warn about ISO C constructs that are outside of the common subset of
3439 ISO C and ISO C++, e.g.@: request for implicit conversion from
3440 @code{void *} to a pointer to non-@code{void} type.
3442 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
3443 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
3444 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
3445 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
3449 Warn whenever a pointer is cast so as to remove a type qualifier from
3450 the target type. For example, warn if a @code{const char *} is cast
3451 to an ordinary @code{char *}.
3454 @opindex Wcast-align
3455 Warn whenever a pointer is cast such that the required alignment of the
3456 target is increased. For example, warn if a @code{char *} is cast to
3457 an @code{int *} on machines where integers can only be accessed at
3458 two- or four-byte boundaries.
3460 @item -Wwrite-strings
3461 @opindex Wwrite-strings
3462 When compiling C, give string constants the type @code{const
3463 char[@var{length}]} so that
3464 copying the address of one into a non-@code{const} @code{char *}
3465 pointer will get a warning; when compiling C++, warn about the
3466 deprecated conversion from string literals to @code{char *}. This
3467 warning, by default, is enabled for C++ programs.
3468 These warnings will help you find at
3469 compile time code that can try to write into a string constant, but
3470 only if you have been very careful about using @code{const} in
3471 declarations and prototypes. Otherwise, it will just be a nuisance;
3472 this is why we did not make @option{-Wall} request these warnings.
3476 Warn for variables that might be changed by @samp{longjmp} or
3477 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3480 @opindex Wconversion
3481 @opindex Wno-conversion
3482 Warn for implicit conversions that may alter a value. This includes
3483 conversions between real and integer, like @code{abs (x)} when
3484 @code{x} is @code{double}; conversions between signed and unsigned,
3485 like @code{unsigned ui = -1}; and conversions to smaller types, like
3486 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3487 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3488 changed by the conversion like in @code{abs (2.0)}. Warnings about
3489 conversions between signed and unsigned integers can be disabled by
3490 using @option{-Wno-sign-conversion}.
3492 For C++, also warn for conversions between @code{NULL} and non-pointer
3493 types; confusing overload resolution for user-defined conversions; and
3494 conversions that will never use a type conversion operator:
3495 conversions to @code{void}, the same type, a base class or a reference
3496 to them. Warnings about conversions between signed and unsigned
3497 integers are disabled by default in C++ unless
3498 @option{-Wsign-conversion} is explicitly enabled.
3501 @opindex Wempty-body
3502 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
3503 while} statement. Additionally, in C++, warn when an empty body occurs
3504 in a @samp{while} or @samp{for} statement with no whitespacing before
3505 the semicolon. This warning is also enabled by @option{-Wextra}.
3507 @item -Wsign-compare
3508 @opindex Wsign-compare
3509 @cindex warning for comparison of signed and unsigned values
3510 @cindex comparison of signed and unsigned values, warning
3511 @cindex signed and unsigned values, comparison warning
3512 Warn when a comparison between signed and unsigned values could produce
3513 an incorrect result when the signed value is converted to unsigned.
3514 This warning is also enabled by @option{-Wextra}; to get the other warnings
3515 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3517 @item -Wsign-conversion
3518 @opindex Wsign-conversion
3519 @opindex Wno-sign-conversion
3520 Warn for implicit conversions that may change the sign of an integer
3521 value, like assigning a signed integer expression to an unsigned
3522 integer variable. An explicit cast silences the warning. In C, this
3523 option is enabled also by @option{-Wconversion}.
3527 @opindex Wno-address
3528 Warn about suspicious uses of memory addresses. These include using
3529 the address of a function in a conditional expression, such as
3530 @code{void func(void); if (func)}, and comparisons against the memory
3531 address of a string literal, such as @code{if (x == "abc")}. Such
3532 uses typically indicate a programmer error: the address of a function
3533 always evaluates to true, so their use in a conditional usually
3534 indicate that the programmer forgot the parentheses in a function
3535 call; and comparisons against string literals result in unspecified
3536 behavior and are not portable in C, so they usually indicate that the
3537 programmer intended to use @code{strcmp}. This warning is enabled by
3541 @opindex Wlogical-op
3542 @opindex Wno-logical-op
3543 Warn about suspicious uses of logical operators in expressions.
3544 This includes using logical operators in contexts where a
3545 bit-wise operator is likely to be expected.
3547 @item -Waggregate-return
3548 @opindex Waggregate-return
3549 Warn if any functions that return structures or unions are defined or
3550 called. (In languages where you can return an array, this also elicits
3553 @item -Wno-attributes
3554 @opindex Wno-attributes
3555 @opindex Wattributes
3556 Do not warn if an unexpected @code{__attribute__} is used, such as
3557 unrecognized attributes, function attributes applied to variables,
3558 etc. This will not stop errors for incorrect use of supported
3561 @item -Wstrict-prototypes @r{(C only)}
3562 @opindex Wstrict-prototypes
3563 Warn if a function is declared or defined without specifying the
3564 argument types. (An old-style function definition is permitted without
3565 a warning if preceded by a declaration which specifies the argument
3568 @item -Wold-style-declaration @r{(C only)}
3569 @opindex Wold-style-declaration
3570 Warn for obsolescent usages, according to the C Standard, in a
3571 declaration. For example, warn if storage-class specifiers like
3572 @code{static} are not the first things in a declaration. This warning
3573 is also enabled by @option{-Wextra}.
3575 @item -Wold-style-definition @r{(C only)}
3576 @opindex Wold-style-definition
3577 Warn if an old-style function definition is used. A warning is given
3578 even if there is a previous prototype.
3580 @item -Wmissing-parameter-type @r{(C only)}
3581 @opindex Wmissing-parameter-type
3582 A function parameter is declared without a type specifier in K&R-style
3589 This warning is also enabled by @option{-Wextra}.
3591 @item -Wmissing-prototypes @r{(C only)}
3592 @opindex Wmissing-prototypes
3593 Warn if a global function is defined without a previous prototype
3594 declaration. This warning is issued even if the definition itself
3595 provides a prototype. The aim is to detect global functions that fail
3596 to be declared in header files.
3598 @item -Wmissing-declarations @r{(C and C++ only)}
3599 @opindex Wmissing-declarations
3600 Warn if a global function is defined without a previous declaration.
3601 Do so even if the definition itself provides a prototype.
3602 Use this option to detect global functions that are not declared in
3603 header files. In C++, no warnings are issued for function templates,
3604 or for inline functions, or for functions in anonymous namespaces.
3606 @item -Wmissing-field-initializers
3607 @opindex Wmissing-field-initializers
3610 Warn if a structure's initializer has some fields missing. For
3611 example, the following code would cause such a warning, because
3612 @code{x.h} is implicitly zero:
3615 struct s @{ int f, g, h; @};
3616 struct s x = @{ 3, 4 @};
3619 This option does not warn about designated initializers, so the following
3620 modification would not trigger a warning:
3623 struct s @{ int f, g, h; @};
3624 struct s x = @{ .f = 3, .g = 4 @};
3627 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3628 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3630 @item -Wmissing-noreturn
3631 @opindex Wmissing-noreturn
3632 Warn about functions which might be candidates for attribute @code{noreturn}.
3633 Note these are only possible candidates, not absolute ones. Care should
3634 be taken to manually verify functions actually do not ever return before
3635 adding the @code{noreturn} attribute, otherwise subtle code generation
3636 bugs could be introduced. You will not get a warning for @code{main} in
3637 hosted C environments.
3639 @item -Wmissing-format-attribute
3640 @opindex Wmissing-format-attribute
3642 Warn about function pointers which might be candidates for @code{format}
3643 attributes. Note these are only possible candidates, not absolute ones.
3644 GCC will guess that function pointers with @code{format} attributes that
3645 are used in assignment, initialization, parameter passing or return
3646 statements should have a corresponding @code{format} attribute in the
3647 resulting type. I.e.@: the left-hand side of the assignment or
3648 initialization, the type of the parameter variable, or the return type
3649 of the containing function respectively should also have a @code{format}
3650 attribute to avoid the warning.
3652 GCC will also warn about function definitions which might be
3653 candidates for @code{format} attributes. Again, these are only
3654 possible candidates. GCC will guess that @code{format} attributes
3655 might be appropriate for any function that calls a function like
3656 @code{vprintf} or @code{vscanf}, but this might not always be the
3657 case, and some functions for which @code{format} attributes are
3658 appropriate may not be detected.
3660 @item -Wno-multichar
3661 @opindex Wno-multichar
3663 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3664 Usually they indicate a typo in the user's code, as they have
3665 implementation-defined values, and should not be used in portable code.
3667 @item -Wnormalized=<none|id|nfc|nfkc>
3668 @opindex Wnormalized
3671 @cindex character set, input normalization
3672 In ISO C and ISO C++, two identifiers are different if they are
3673 different sequences of characters. However, sometimes when characters
3674 outside the basic ASCII character set are used, you can have two
3675 different character sequences that look the same. To avoid confusion,
3676 the ISO 10646 standard sets out some @dfn{normalization rules} which
3677 when applied ensure that two sequences that look the same are turned into
3678 the same sequence. GCC can warn you if you are using identifiers which
3679 have not been normalized; this option controls that warning.
3681 There are four levels of warning that GCC supports. The default is
3682 @option{-Wnormalized=nfc}, which warns about any identifier which is
3683 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3684 recommended form for most uses.
3686 Unfortunately, there are some characters which ISO C and ISO C++ allow
3687 in identifiers that when turned into NFC aren't allowable as
3688 identifiers. That is, there's no way to use these symbols in portable
3689 ISO C or C++ and have all your identifiers in NFC.
3690 @option{-Wnormalized=id} suppresses the warning for these characters.
3691 It is hoped that future versions of the standards involved will correct
3692 this, which is why this option is not the default.
3694 You can switch the warning off for all characters by writing
3695 @option{-Wnormalized=none}. You would only want to do this if you
3696 were using some other normalization scheme (like ``D''), because
3697 otherwise you can easily create bugs that are literally impossible to see.
3699 Some characters in ISO 10646 have distinct meanings but look identical
3700 in some fonts or display methodologies, especially once formatting has
3701 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3702 LETTER N'', will display just like a regular @code{n} which has been
3703 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3704 normalization scheme to convert all these into a standard form as
3705 well, and GCC will warn if your code is not in NFKC if you use
3706 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3707 about every identifier that contains the letter O because it might be
3708 confused with the digit 0, and so is not the default, but may be
3709 useful as a local coding convention if the programming environment is
3710 unable to be fixed to display these characters distinctly.
3712 @item -Wno-deprecated-declarations
3713 @opindex Wno-deprecated-declarations
3714 Do not warn about uses of functions (@pxref{Function Attributes}),
3715 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3716 Attributes}) marked as deprecated by using the @code{deprecated}
3720 @opindex Wno-overflow
3721 Do not warn about compile-time overflow in constant expressions.
3723 @item -Woverride-init
3724 @opindex Woverride-init
3727 Warn if an initialized field without side effects is overridden when
3728 using designated initializers (@pxref{Designated Inits, , Designated
3731 This warning is included in @option{-Wextra}. To get other
3732 @option{-Wextra} warnings without this one, use @samp{-Wextra
3733 -Wno-override-init}.
3737 Warn if a structure is given the packed attribute, but the packed
3738 attribute has no effect on the layout or size of the structure.
3739 Such structures may be mis-aligned for little benefit. For
3740 instance, in this code, the variable @code{f.x} in @code{struct bar}
3741 will be misaligned even though @code{struct bar} does not itself
3742 have the packed attribute:
3749 @} __attribute__((packed));
3759 Warn if padding is included in a structure, either to align an element
3760 of the structure or to align the whole structure. Sometimes when this
3761 happens it is possible to rearrange the fields of the structure to
3762 reduce the padding and so make the structure smaller.
3764 @item -Wredundant-decls
3765 @opindex Wredundant-decls
3766 Warn if anything is declared more than once in the same scope, even in
3767 cases where multiple declaration is valid and changes nothing.
3769 @item -Wnested-externs @r{(C only)}
3770 @opindex Wnested-externs
3771 Warn if an @code{extern} declaration is encountered within a function.
3773 @item -Wunreachable-code
3774 @opindex Wunreachable-code
3775 Warn if the compiler detects that code will never be executed.
3777 This option is intended to warn when the compiler detects that at
3778 least a whole line of source code will never be executed, because
3779 some condition is never satisfied or because it is after a
3780 procedure that never returns.
3782 It is possible for this option to produce a warning even though there
3783 are circumstances under which part of the affected line can be executed,
3784 so care should be taken when removing apparently-unreachable code.
3786 For instance, when a function is inlined, a warning may mean that the
3787 line is unreachable in only one inlined copy of the function.
3789 This option is not made part of @option{-Wall} because in a debugging
3790 version of a program there is often substantial code which checks
3791 correct functioning of the program and is, hopefully, unreachable
3792 because the program does work. Another common use of unreachable
3793 code is to provide behavior which is selectable at compile-time.
3797 Warn if a function can not be inlined and it was declared as inline.
3798 Even with this option, the compiler will not warn about failures to
3799 inline functions declared in system headers.
3801 The compiler uses a variety of heuristics to determine whether or not
3802 to inline a function. For example, the compiler takes into account
3803 the size of the function being inlined and the amount of inlining
3804 that has already been done in the current function. Therefore,
3805 seemingly insignificant changes in the source program can cause the
3806 warnings produced by @option{-Winline} to appear or disappear.
3808 @item -Wno-invalid-offsetof @r{(C++ only)}
3809 @opindex Wno-invalid-offsetof
3810 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3811 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3812 to a non-POD type is undefined. In existing C++ implementations,
3813 however, @samp{offsetof} typically gives meaningful results even when
3814 applied to certain kinds of non-POD types. (Such as a simple
3815 @samp{struct} that fails to be a POD type only by virtue of having a
3816 constructor.) This flag is for users who are aware that they are
3817 writing nonportable code and who have deliberately chosen to ignore the
3820 The restrictions on @samp{offsetof} may be relaxed in a future version
3821 of the C++ standard.
3823 @item -Wno-int-to-pointer-cast @r{(C only)}
3824 @opindex Wno-int-to-pointer-cast
3825 Suppress warnings from casts to pointer type of an integer of a
3828 @item -Wno-pointer-to-int-cast @r{(C only)}
3829 @opindex Wno-pointer-to-int-cast
3830 Suppress warnings from casts from a pointer to an integer type of a
3834 @opindex Winvalid-pch
3835 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3836 the search path but can't be used.
3840 @opindex Wno-long-long
3841 Warn if @samp{long long} type is used. This is default. To inhibit
3842 the warning messages, use @option{-Wno-long-long}. Flags
3843 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3844 only when @option{-pedantic} flag is used.
3846 @item -Wvariadic-macros
3847 @opindex Wvariadic-macros
3848 @opindex Wno-variadic-macros
3849 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3850 alternate syntax when in pedantic ISO C99 mode. This is default.
3851 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3856 Warn if variable length array is used in the code.
3857 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
3858 the variable length array.
3860 @item -Wvolatile-register-var
3861 @opindex Wvolatile-register-var
3862 @opindex Wno-volatile-register-var
3863 Warn if a register variable is declared volatile. The volatile
3864 modifier does not inhibit all optimizations that may eliminate reads
3865 and/or writes to register variables.
3867 @item -Wdisabled-optimization
3868 @opindex Wdisabled-optimization
3869 Warn if a requested optimization pass is disabled. This warning does
3870 not generally indicate that there is anything wrong with your code; it
3871 merely indicates that GCC's optimizers were unable to handle the code
3872 effectively. Often, the problem is that your code is too big or too
3873 complex; GCC will refuse to optimize programs when the optimization
3874 itself is likely to take inordinate amounts of time.
3876 @item -Wpointer-sign
3877 @opindex Wpointer-sign
3878 @opindex Wno-pointer-sign
3879 Warn for pointer argument passing or assignment with different signedness.
3880 This option is only supported for C and Objective-C@. It is implied by
3881 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3882 @option{-Wno-pointer-sign}.
3886 Make all warnings into errors.
3890 Make the specified warning into an errors. The specifier for a
3891 warning is appended, for example @option{-Werror=switch} turns the
3892 warnings controlled by @option{-Wswitch} into errors. This switch
3893 takes a negative form, to be used to negate @option{-Werror} for
3894 specific warnings, for example @option{-Wno-error=switch} makes
3895 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3896 is in effect. You can use the @option{-fdiagnostics-show-option}
3897 option to have each controllable warning amended with the option which
3898 controls it, to determine what to use with this option.
3900 Note that specifying @option{-Werror=}@var{foo} automatically implies
3901 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3904 @item -Wstack-protector
3905 @opindex Wstack-protector
3906 This option is only active when @option{-fstack-protector} is active. It
3907 warns about functions that will not be protected against stack smashing.
3909 @item -Woverlength-strings
3910 @opindex Woverlength-strings
3911 Warn about string constants which are longer than the ``minimum
3912 maximum'' length specified in the C standard. Modern compilers
3913 generally allow string constants which are much longer than the
3914 standard's minimum limit, but very portable programs should avoid
3915 using longer strings.
3917 The limit applies @emph{after} string constant concatenation, and does
3918 not count the trailing NUL@. In C89, the limit was 509 characters; in
3919 C99, it was raised to 4095. C++98 does not specify a normative
3920 minimum maximum, so we do not diagnose overlength strings in C++@.
3922 This option is implied by @option{-pedantic}, and can be disabled with
3923 @option{-Wno-overlength-strings}.
3926 @node Debugging Options
3927 @section Options for Debugging Your Program or GCC
3928 @cindex options, debugging
3929 @cindex debugging information options
3931 GCC has various special options that are used for debugging
3932 either your program or GCC:
3937 Produce debugging information in the operating system's native format
3938 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3941 On most systems that use stabs format, @option{-g} enables use of extra
3942 debugging information that only GDB can use; this extra information
3943 makes debugging work better in GDB but will probably make other debuggers
3945 refuse to read the program. If you want to control for certain whether
3946 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3947 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3949 GCC allows you to use @option{-g} with
3950 @option{-O}. The shortcuts taken by optimized code may occasionally
3951 produce surprising results: some variables you declared may not exist
3952 at all; flow of control may briefly move where you did not expect it;
3953 some statements may not be executed because they compute constant
3954 results or their values were already at hand; some statements may
3955 execute in different places because they were moved out of loops.
3957 Nevertheless it proves possible to debug optimized output. This makes
3958 it reasonable to use the optimizer for programs that might have bugs.
3960 The following options are useful when GCC is generated with the
3961 capability for more than one debugging format.
3965 Produce debugging information for use by GDB@. This means to use the
3966 most expressive format available (DWARF 2, stabs, or the native format
3967 if neither of those are supported), including GDB extensions if at all
3972 Produce debugging information in stabs format (if that is supported),
3973 without GDB extensions. This is the format used by DBX on most BSD
3974 systems. On MIPS, Alpha and System V Release 4 systems this option
3975 produces stabs debugging output which is not understood by DBX or SDB@.
3976 On System V Release 4 systems this option requires the GNU assembler.
3978 @item -feliminate-unused-debug-symbols
3979 @opindex feliminate-unused-debug-symbols
3980 Produce debugging information in stabs format (if that is supported),
3981 for only symbols that are actually used.
3983 @item -femit-class-debug-always
3984 Instead of emitting debugging information for a C++ class in only one
3985 object file, emit it in all object files using the class. This option
3986 should be used only with debuggers that are unable to handle the way GCC
3987 normally emits debugging information for classes because using this
3988 option will increase the size of debugging information by as much as a
3993 Produce debugging information in stabs format (if that is supported),
3994 using GNU extensions understood only by the GNU debugger (GDB)@. The
3995 use of these extensions is likely to make other debuggers crash or
3996 refuse to read the program.
4000 Produce debugging information in COFF format (if that is supported).
4001 This is the format used by SDB on most System V systems prior to
4006 Produce debugging information in XCOFF format (if that is supported).
4007 This is the format used by the DBX debugger on IBM RS/6000 systems.
4011 Produce debugging information in XCOFF format (if that is supported),
4012 using GNU extensions understood only by the GNU debugger (GDB)@. The
4013 use of these extensions is likely to make other debuggers crash or
4014 refuse to read the program, and may cause assemblers other than the GNU
4015 assembler (GAS) to fail with an error.
4019 Produce debugging information in DWARF version 2 format (if that is
4020 supported). This is the format used by DBX on IRIX 6. With this
4021 option, GCC uses features of DWARF version 3 when they are useful;
4022 version 3 is upward compatible with version 2, but may still cause
4023 problems for older debuggers.
4027 Produce debugging information in VMS debug format (if that is
4028 supported). This is the format used by DEBUG on VMS systems.
4031 @itemx -ggdb@var{level}
4032 @itemx -gstabs@var{level}
4033 @itemx -gcoff@var{level}
4034 @itemx -gxcoff@var{level}
4035 @itemx -gvms@var{level}
4036 Request debugging information and also use @var{level} to specify how
4037 much information. The default level is 2.
4039 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4042 Level 1 produces minimal information, enough for making backtraces in
4043 parts of the program that you don't plan to debug. This includes
4044 descriptions of functions and external variables, but no information
4045 about local variables and no line numbers.
4047 Level 3 includes extra information, such as all the macro definitions
4048 present in the program. Some debuggers support macro expansion when
4049 you use @option{-g3}.
4051 @option{-gdwarf-2} does not accept a concatenated debug level, because
4052 GCC used to support an option @option{-gdwarf} that meant to generate
4053 debug information in version 1 of the DWARF format (which is very
4054 different from version 2), and it would have been too confusing. That
4055 debug format is long obsolete, but the option cannot be changed now.
4056 Instead use an additional @option{-g@var{level}} option to change the
4057 debug level for DWARF2.
4059 @item -feliminate-dwarf2-dups
4060 @opindex feliminate-dwarf2-dups
4061 Compress DWARF2 debugging information by eliminating duplicated
4062 information about each symbol. This option only makes sense when
4063 generating DWARF2 debugging information with @option{-gdwarf-2}.
4065 @item -femit-struct-debug-baseonly
4066 Emit debug information for struct-like types
4067 only when the base name of the compilation source file
4068 matches the base name of file in which the struct was defined.
4070 This option substantially reduces the size of debugging information,
4071 but at significant potential loss in type information to the debugger.
4072 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4073 See @option{-femit-struct-debug-detailed} for more detailed control.
4075 This option works only with DWARF 2.
4077 @item -femit-struct-debug-reduced
4078 Emit debug information for struct-like types
4079 only when the base name of the compilation source file
4080 matches the base name of file in which the type was defined,
4081 unless the struct is a template or defined in a system header.
4083 This option significantly reduces the size of debugging information,
4084 with some potential loss in type information to the debugger.
4085 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4086 See @option{-femit-struct-debug-detailed} for more detailed control.
4088 This option works only with DWARF 2.
4090 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4091 Specify the struct-like types
4092 for which the compiler will generate debug information.
4093 The intent is to reduce duplicate struct debug information
4094 between different object files within the same program.
4096 This option is a detailed version of
4097 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4098 which will serve for most needs.
4100 A specification has the syntax
4101 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4103 The optional first word limits the specification to
4104 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4105 A struct type is used directly when it is the type of a variable, member.
4106 Indirect uses arise through pointers to structs.
4107 That is, when use of an incomplete struct would be legal, the use is indirect.
4109 @samp{struct one direct; struct two * indirect;}.
4111 The optional second word limits the specification to
4112 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4113 Generic structs are a bit complicated to explain.
4114 For C++, these are non-explicit specializations of template classes,
4115 or non-template classes within the above.
4116 Other programming languages have generics,
4117 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4119 The third word specifies the source files for those
4120 structs for which the compiler will emit debug information.
4121 The values @samp{none} and @samp{any} have the normal meaning.
4122 The value @samp{base} means that
4123 the base of name of the file in which the type declaration appears
4124 must match the base of the name of the main compilation file.
4125 In practice, this means that
4126 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4127 but types declared in other header will not.
4128 The value @samp{sys} means those types satisfying @samp{base}
4129 or declared in system or compiler headers.
4131 You may need to experiment to determine the best settings for your application.
4133 The default is @samp{-femit-struct-debug-detailed=all}.
4135 This option works only with DWARF 2.
4137 @cindex @command{prof}
4140 Generate extra code to write profile information suitable for the
4141 analysis program @command{prof}. You must use this option when compiling
4142 the source files you want data about, and you must also use it when
4145 @cindex @command{gprof}
4148 Generate extra code to write profile information suitable for the
4149 analysis program @command{gprof}. You must use this option when compiling
4150 the source files you want data about, and you must also use it when
4155 Makes the compiler print out each function name as it is compiled, and
4156 print some statistics about each pass when it finishes.
4159 @opindex ftime-report
4160 Makes the compiler print some statistics about the time consumed by each
4161 pass when it finishes.
4164 @opindex fmem-report
4165 Makes the compiler print some statistics about permanent memory
4166 allocation when it finishes.
4168 @item -fpre-ipa-mem-report
4169 @opindex fpre-ipa-mem-report
4170 @item -fpost-ipa-mem-report
4171 @opindex fpost-ipa-mem-report
4172 Makes the compiler print some statistics about permanent memory
4173 allocation before or after interprocedural optimization.
4175 @item -fprofile-arcs
4176 @opindex fprofile-arcs
4177 Add code so that program flow @dfn{arcs} are instrumented. During
4178 execution the program records how many times each branch and call is
4179 executed and how many times it is taken or returns. When the compiled
4180 program exits it saves this data to a file called
4181 @file{@var{auxname}.gcda} for each source file. The data may be used for
4182 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4183 test coverage analysis (@option{-ftest-coverage}). Each object file's
4184 @var{auxname} is generated from the name of the output file, if
4185 explicitly specified and it is not the final executable, otherwise it is
4186 the basename of the source file. In both cases any suffix is removed
4187 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4188 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4189 @xref{Cross-profiling}.
4191 @cindex @command{gcov}
4195 This option is used to compile and link code instrumented for coverage
4196 analysis. The option is a synonym for @option{-fprofile-arcs}
4197 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4198 linking). See the documentation for those options for more details.
4203 Compile the source files with @option{-fprofile-arcs} plus optimization
4204 and code generation options. For test coverage analysis, use the
4205 additional @option{-ftest-coverage} option. You do not need to profile
4206 every source file in a program.
4209 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4210 (the latter implies the former).
4213 Run the program on a representative workload to generate the arc profile
4214 information. This may be repeated any number of times. You can run
4215 concurrent instances of your program, and provided that the file system
4216 supports locking, the data files will be correctly updated. Also
4217 @code{fork} calls are detected and correctly handled (double counting
4221 For profile-directed optimizations, compile the source files again with
4222 the same optimization and code generation options plus
4223 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
4224 Control Optimization}).
4227 For test coverage analysis, use @command{gcov} to produce human readable
4228 information from the @file{.gcno} and @file{.gcda} files. Refer to the
4229 @command{gcov} documentation for further information.
4233 With @option{-fprofile-arcs}, for each function of your program GCC
4234 creates a program flow graph, then finds a spanning tree for the graph.
4235 Only arcs that are not on the spanning tree have to be instrumented: the
4236 compiler adds code to count the number of times that these arcs are
4237 executed. When an arc is the only exit or only entrance to a block, the
4238 instrumentation code can be added to the block; otherwise, a new basic
4239 block must be created to hold the instrumentation code.
4242 @item -ftest-coverage
4243 @opindex ftest-coverage
4244 Produce a notes file that the @command{gcov} code-coverage utility
4245 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
4246 show program coverage. Each source file's note file is called
4247 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
4248 above for a description of @var{auxname} and instructions on how to
4249 generate test coverage data. Coverage data will match the source files
4250 more closely, if you do not optimize.
4252 @item -fdbg-cnt-list
4253 @opindex fdbg-cnt-list
4254 Print the name and the counter upperbound for all debug counters.
4256 @item -fdbg-cnt=@var{counter-value-list}
4258 Set the internal debug counter upperbound. @var{counter-value-list}
4259 is a comma-separated list of @var{name}:@var{value} pairs
4260 which sets the upperbound of each debug counter @var{name} to @var{value}.
4261 All debug counters have the initial upperbound of @var{UINT_MAX},
4262 thus dbg_cnt() returns true always unless the upperbound is set by this option.
4263 e.g. With -fdbg-cnt=dce:10,tail_call:0
4264 dbg_cnt(dce) will return true only for first 10 invocations
4265 and dbg_cnt(tail_call) will return false always.
4267 @item -d@var{letters}
4268 @item -fdump-rtl-@var{pass}
4270 Says to make debugging dumps during compilation at times specified by
4271 @var{letters}. This is used for debugging the RTL-based passes of the
4272 compiler. The file names for most of the dumps are made by appending a
4273 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
4274 from the name of the output file, if explicitly specified and it is not
4275 an executable, otherwise it is the basename of the source file. These
4276 switches may have different effects when @option{-E} is used for
4279 Most debug dumps can be enabled either passing a letter to the @option{-d}
4280 option, or with a long @option{-fdump-rtl} switch; here are the possible
4281 letters for use in @var{letters} and @var{pass}, and their meanings:
4286 Annotate the assembler output with miscellaneous debugging information.
4289 @itemx -fdump-rtl-bbro
4291 @opindex fdump-rtl-bbro
4292 Dump after block reordering, to @file{@var{file}.148r.bbro}.
4295 @itemx -fdump-rtl-combine
4297 @opindex fdump-rtl-combine
4298 Dump after the RTL instruction combination pass, to the file
4299 @file{@var{file}.129r.combine}.
4302 @itemx -fdump-rtl-ce1
4303 @itemx -fdump-rtl-ce2
4305 @opindex fdump-rtl-ce1
4306 @opindex fdump-rtl-ce2
4307 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
4308 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
4309 and @option{-fdump-rtl-ce2} enable dumping after the second if
4310 conversion, to the file @file{@var{file}.130r.ce2}.
4313 @itemx -fdump-rtl-btl
4314 @itemx -fdump-rtl-dbr
4316 @opindex fdump-rtl-btl
4317 @opindex fdump-rtl-dbr
4318 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
4319 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
4320 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
4321 scheduling, to @file{@var{file}.36.dbr}.
4325 Dump all macro definitions, at the end of preprocessing, in addition to
4329 @itemx -fdump-rtl-ce3
4331 @opindex fdump-rtl-ce3
4332 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
4335 @itemx -fdump-rtl-cfg
4336 @itemx -fdump-rtl-life
4338 @opindex fdump-rtl-cfg
4339 @opindex fdump-rtl-life
4340 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
4341 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
4342 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
4343 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
4346 @itemx -fdump-rtl-greg
4348 @opindex fdump-rtl-greg
4349 Dump after global register allocation, to @file{@var{file}.139r.greg}.
4352 @itemx -fdump-rtl-gcse
4353 @itemx -fdump-rtl-bypass
4355 @opindex fdump-rtl-gcse
4356 @opindex fdump-rtl-bypass
4357 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
4358 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
4359 enable dumping after jump bypassing and control flow optimizations, to
4360 @file{@var{file}.115r.bypass}.
4363 @itemx -fdump-rtl-eh
4365 @opindex fdump-rtl-eh
4366 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
4369 @itemx -fdump-rtl-sibling
4371 @opindex fdump-rtl-sibling
4372 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
4375 @itemx -fdump-rtl-jump
4377 @opindex fdump-rtl-jump
4378 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
4381 @itemx -fdump-rtl-stack
4383 @opindex fdump-rtl-stack
4384 Dump after conversion from GCC's "flat register file" registers to the
4385 x87's stack-like registers, to @file{@var{file}.152r.stack}.
4388 @itemx -fdump-rtl-lreg
4390 @opindex fdump-rtl-lreg
4391 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
4394 @itemx -fdump-rtl-loop2
4396 @opindex fdump-rtl-loop2
4397 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
4398 loop optimization pass, to @file{@var{file}.119r.loop2},
4399 @file{@var{file}.120r.loop2_init},
4400 @file{@var{file}.121r.loop2_invariant}, and
4401 @file{@var{file}.125r.loop2_done}.
4404 @itemx -fdump-rtl-sms
4406 @opindex fdump-rtl-sms
4407 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
4410 @itemx -fdump-rtl-mach
4412 @opindex fdump-rtl-mach
4413 Dump after performing the machine dependent reorganization pass, to
4414 @file{@var{file}.155r.mach} if that pass exists.
4417 @itemx -fdump-rtl-rnreg
4419 @opindex fdump-rtl-rnreg
4420 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
4423 @itemx -fdump-rtl-regmove
4425 @opindex fdump-rtl-regmove
4426 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
4429 @itemx -fdump-rtl-postreload
4431 @opindex fdump-rtl-postreload
4432 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4435 @itemx -fdump-rtl-expand
4437 @opindex fdump-rtl-expand
4438 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4441 @itemx -fdump-rtl-sched2
4443 @opindex fdump-rtl-sched2
4444 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4447 @itemx -fdump-rtl-cse
4449 @opindex fdump-rtl-cse
4450 Dump after CSE (including the jump optimization that sometimes follows
4451 CSE), to @file{@var{file}.113r.cse}.
4454 @itemx -fdump-rtl-sched1
4456 @opindex fdump-rtl-sched1
4457 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4460 @itemx -fdump-rtl-cse2
4462 @opindex fdump-rtl-cse2
4463 Dump after the second CSE pass (including the jump optimization that
4464 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4467 @itemx -fdump-rtl-tracer
4469 @opindex fdump-rtl-tracer
4470 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4473 @itemx -fdump-rtl-vpt
4474 @itemx -fdump-rtl-vartrack
4476 @opindex fdump-rtl-vpt
4477 @opindex fdump-rtl-vartrack
4478 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4479 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4480 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4481 to @file{@var{file}.154r.vartrack}.
4484 @itemx -fdump-rtl-flow2
4486 @opindex fdump-rtl-flow2
4487 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4490 @itemx -fdump-rtl-peephole2
4492 @opindex fdump-rtl-peephole2
4493 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4496 @itemx -fdump-rtl-web
4498 @opindex fdump-rtl-web
4499 Dump after live range splitting, to @file{@var{file}.126r.web}.
4502 @itemx -fdump-rtl-all
4504 @opindex fdump-rtl-all
4505 Produce all the dumps listed above.
4509 Produce a core dump whenever an error occurs.
4513 Print statistics on memory usage, at the end of the run, to
4518 Annotate the assembler output with a comment indicating which
4519 pattern and alternative was used. The length of each instruction is
4524 Dump the RTL in the assembler output as a comment before each instruction.
4525 Also turns on @option{-dp} annotation.
4529 For each of the other indicated dump files (either with @option{-d} or
4530 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4531 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4535 Just generate RTL for a function instead of compiling it. Usually used
4536 with @samp{r} (@option{-fdump-rtl-expand}).
4540 Dump debugging information during parsing, to standard error.
4544 @opindex fdump-noaddr
4545 When doing debugging dumps (see @option{-d} option above), suppress
4546 address output. This makes it more feasible to use diff on debugging
4547 dumps for compiler invocations with different compiler binaries and/or
4548 different text / bss / data / heap / stack / dso start locations.
4550 @item -fdump-unnumbered
4551 @opindex fdump-unnumbered
4552 When doing debugging dumps (see @option{-d} option above), suppress instruction
4553 numbers and address output. This makes it more feasible to
4554 use diff on debugging dumps for compiler invocations with different
4555 options, in particular with and without @option{-g}.
4557 @item -fdump-translation-unit @r{(C++ only)}
4558 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4559 @opindex fdump-translation-unit
4560 Dump a representation of the tree structure for the entire translation
4561 unit to a file. The file name is made by appending @file{.tu} to the
4562 source file name. If the @samp{-@var{options}} form is used, @var{options}
4563 controls the details of the dump as described for the
4564 @option{-fdump-tree} options.
4566 @item -fdump-class-hierarchy @r{(C++ only)}
4567 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4568 @opindex fdump-class-hierarchy
4569 Dump a representation of each class's hierarchy and virtual function
4570 table layout to a file. The file name is made by appending @file{.class}
4571 to the source file name. If the @samp{-@var{options}} form is used,
4572 @var{options} controls the details of the dump as described for the
4573 @option{-fdump-tree} options.
4575 @item -fdump-ipa-@var{switch}
4577 Control the dumping at various stages of inter-procedural analysis
4578 language tree to a file. The file name is generated by appending a switch
4579 specific suffix to the source file name. The following dumps are possible:
4583 Enables all inter-procedural analysis dumps; currently the only produced
4584 dump is the @samp{cgraph} dump.
4587 Dumps information about call-graph optimization, unused function removal,
4588 and inlining decisions.
4591 @item -fdump-tree-@var{switch}
4592 @itemx -fdump-tree-@var{switch}-@var{options}
4594 Control the dumping at various stages of processing the intermediate
4595 language tree to a file. The file name is generated by appending a switch
4596 specific suffix to the source file name. If the @samp{-@var{options}}
4597 form is used, @var{options} is a list of @samp{-} separated options that
4598 control the details of the dump. Not all options are applicable to all
4599 dumps, those which are not meaningful will be ignored. The following
4600 options are available
4604 Print the address of each node. Usually this is not meaningful as it
4605 changes according to the environment and source file. Its primary use
4606 is for tying up a dump file with a debug environment.
4608 Inhibit dumping of members of a scope or body of a function merely
4609 because that scope has been reached. Only dump such items when they
4610 are directly reachable by some other path. When dumping pretty-printed
4611 trees, this option inhibits dumping the bodies of control structures.
4613 Print a raw representation of the tree. By default, trees are
4614 pretty-printed into a C-like representation.
4616 Enable more detailed dumps (not honored by every dump option).
4618 Enable dumping various statistics about the pass (not honored by every dump
4621 Enable showing basic block boundaries (disabled in raw dumps).
4623 Enable showing virtual operands for every statement.
4625 Enable showing line numbers for statements.
4627 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4629 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4632 The following tree dumps are possible:
4636 Dump before any tree based optimization, to @file{@var{file}.original}.
4639 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4642 Dump after function inlining, to @file{@var{file}.inlined}.
4645 @opindex fdump-tree-gimple
4646 Dump each function before and after the gimplification pass to a file. The
4647 file name is made by appending @file{.gimple} to the source file name.
4650 @opindex fdump-tree-cfg
4651 Dump the control flow graph of each function to a file. The file name is
4652 made by appending @file{.cfg} to the source file name.
4655 @opindex fdump-tree-vcg
4656 Dump the control flow graph of each function to a file in VCG format. The
4657 file name is made by appending @file{.vcg} to the source file name. Note
4658 that if the file contains more than one function, the generated file cannot
4659 be used directly by VCG@. You will need to cut and paste each function's
4660 graph into its own separate file first.
4663 @opindex fdump-tree-ch
4664 Dump each function after copying loop headers. The file name is made by
4665 appending @file{.ch} to the source file name.
4668 @opindex fdump-tree-ssa
4669 Dump SSA related information to a file. The file name is made by appending
4670 @file{.ssa} to the source file name.
4673 @opindex fdump-tree-salias
4674 Dump structure aliasing variable information to a file. This file name
4675 is made by appending @file{.salias} to the source file name.
4678 @opindex fdump-tree-alias
4679 Dump aliasing information for each function. The file name is made by
4680 appending @file{.alias} to the source file name.
4683 @opindex fdump-tree-ccp
4684 Dump each function after CCP@. The file name is made by appending
4685 @file{.ccp} to the source file name.
4688 @opindex fdump-tree-storeccp
4689 Dump each function after STORE-CCP. The file name is made by appending
4690 @file{.storeccp} to the source file name.
4693 @opindex fdump-tree-pre
4694 Dump trees after partial redundancy elimination. The file name is made
4695 by appending @file{.pre} to the source file name.
4698 @opindex fdump-tree-fre
4699 Dump trees after full redundancy elimination. The file name is made
4700 by appending @file{.fre} to the source file name.
4703 @opindex fdump-tree-copyprop
4704 Dump trees after copy propagation. The file name is made
4705 by appending @file{.copyprop} to the source file name.
4707 @item store_copyprop
4708 @opindex fdump-tree-store_copyprop
4709 Dump trees after store copy-propagation. The file name is made
4710 by appending @file{.store_copyprop} to the source file name.
4713 @opindex fdump-tree-dce
4714 Dump each function after dead code elimination. The file name is made by
4715 appending @file{.dce} to the source file name.
4718 @opindex fdump-tree-mudflap
4719 Dump each function after adding mudflap instrumentation. The file name is
4720 made by appending @file{.mudflap} to the source file name.
4723 @opindex fdump-tree-sra
4724 Dump each function after performing scalar replacement of aggregates. The
4725 file name is made by appending @file{.sra} to the source file name.
4728 @opindex fdump-tree-sink
4729 Dump each function after performing code sinking. The file name is made
4730 by appending @file{.sink} to the source file name.
4733 @opindex fdump-tree-dom
4734 Dump each function after applying dominator tree optimizations. The file
4735 name is made by appending @file{.dom} to the source file name.
4738 @opindex fdump-tree-dse
4739 Dump each function after applying dead store elimination. The file
4740 name is made by appending @file{.dse} to the source file name.
4743 @opindex fdump-tree-phiopt
4744 Dump each function after optimizing PHI nodes into straightline code. The file
4745 name is made by appending @file{.phiopt} to the source file name.
4748 @opindex fdump-tree-forwprop
4749 Dump each function after forward propagating single use variables. The file
4750 name is made by appending @file{.forwprop} to the source file name.
4753 @opindex fdump-tree-copyrename
4754 Dump each function after applying the copy rename optimization. The file
4755 name is made by appending @file{.copyrename} to the source file name.
4758 @opindex fdump-tree-nrv
4759 Dump each function after applying the named return value optimization on
4760 generic trees. The file name is made by appending @file{.nrv} to the source
4764 @opindex fdump-tree-vect
4765 Dump each function after applying vectorization of loops. The file name is
4766 made by appending @file{.vect} to the source file name.
4769 @opindex fdump-tree-vrp
4770 Dump each function after Value Range Propagation (VRP). The file name
4771 is made by appending @file{.vrp} to the source file name.
4774 @opindex fdump-tree-all
4775 Enable all the available tree dumps with the flags provided in this option.
4778 @item -ftree-vectorizer-verbose=@var{n}
4779 @opindex ftree-vectorizer-verbose
4780 This option controls the amount of debugging output the vectorizer prints.
4781 This information is written to standard error, unless
4782 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4783 in which case it is output to the usual dump listing file, @file{.vect}.
4784 For @var{n}=0 no diagnostic information is reported.
4785 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4786 and the total number of loops that got vectorized.
4787 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4788 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4789 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4790 level that @option{-fdump-tree-vect-stats} uses.
4791 Higher verbosity levels mean either more information dumped for each
4792 reported loop, or same amount of information reported for more loops:
4793 If @var{n}=3, alignment related information is added to the reports.
4794 If @var{n}=4, data-references related information (e.g. memory dependences,
4795 memory access-patterns) is added to the reports.
4796 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4797 that did not pass the first analysis phase (i.e. may not be countable, or
4798 may have complicated control-flow).
4799 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4800 For @var{n}=7, all the information the vectorizer generates during its
4801 analysis and transformation is reported. This is the same verbosity level
4802 that @option{-fdump-tree-vect-details} uses.
4804 @item -frandom-seed=@var{string}
4805 @opindex frandom-string
4806 This option provides a seed that GCC uses when it would otherwise use
4807 random numbers. It is used to generate certain symbol names
4808 that have to be different in every compiled file. It is also used to
4809 place unique stamps in coverage data files and the object files that
4810 produce them. You can use the @option{-frandom-seed} option to produce
4811 reproducibly identical object files.
4813 The @var{string} should be different for every file you compile.
4815 @item -fsched-verbose=@var{n}
4816 @opindex fsched-verbose
4817 On targets that use instruction scheduling, this option controls the
4818 amount of debugging output the scheduler prints. This information is
4819 written to standard error, unless @option{-dS} or @option{-dR} is
4820 specified, in which case it is output to the usual dump
4821 listing file, @file{.sched} or @file{.sched2} respectively. However
4822 for @var{n} greater than nine, the output is always printed to standard
4825 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4826 same information as @option{-dRS}. For @var{n} greater than one, it
4827 also output basic block probabilities, detailed ready list information
4828 and unit/insn info. For @var{n} greater than two, it includes RTL
4829 at abort point, control-flow and regions info. And for @var{n} over
4830 four, @option{-fsched-verbose} also includes dependence info.
4834 Store the usual ``temporary'' intermediate files permanently; place them
4835 in the current directory and name them based on the source file. Thus,
4836 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4837 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4838 preprocessed @file{foo.i} output file even though the compiler now
4839 normally uses an integrated preprocessor.
4841 When used in combination with the @option{-x} command line option,
4842 @option{-save-temps} is sensible enough to avoid over writing an
4843 input source file with the same extension as an intermediate file.
4844 The corresponding intermediate file may be obtained by renaming the
4845 source file before using @option{-save-temps}.
4849 Report the CPU time taken by each subprocess in the compilation
4850 sequence. For C source files, this is the compiler proper and assembler
4851 (plus the linker if linking is done). The output looks like this:
4858 The first number on each line is the ``user time'', that is time spent
4859 executing the program itself. The second number is ``system time'',
4860 time spent executing operating system routines on behalf of the program.
4861 Both numbers are in seconds.
4863 @item -fvar-tracking
4864 @opindex fvar-tracking
4865 Run variable tracking pass. It computes where variables are stored at each
4866 position in code. Better debugging information is then generated
4867 (if the debugging information format supports this information).
4869 It is enabled by default when compiling with optimization (@option{-Os},
4870 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4871 the debug info format supports it.
4873 @item -print-file-name=@var{library}
4874 @opindex print-file-name
4875 Print the full absolute name of the library file @var{library} that
4876 would be used when linking---and don't do anything else. With this
4877 option, GCC does not compile or link anything; it just prints the
4880 @item -print-multi-directory
4881 @opindex print-multi-directory
4882 Print the directory name corresponding to the multilib selected by any
4883 other switches present in the command line. This directory is supposed
4884 to exist in @env{GCC_EXEC_PREFIX}.
4886 @item -print-multi-lib
4887 @opindex print-multi-lib
4888 Print the mapping from multilib directory names to compiler switches
4889 that enable them. The directory name is separated from the switches by
4890 @samp{;}, and each switch starts with an @samp{@@} instead of the
4891 @samp{-}, without spaces between multiple switches. This is supposed to
4892 ease shell-processing.
4894 @item -print-prog-name=@var{program}
4895 @opindex print-prog-name
4896 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4898 @item -print-libgcc-file-name
4899 @opindex print-libgcc-file-name
4900 Same as @option{-print-file-name=libgcc.a}.
4902 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4903 but you do want to link with @file{libgcc.a}. You can do
4906 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4909 @item -print-search-dirs
4910 @opindex print-search-dirs
4911 Print the name of the configured installation directory and a list of
4912 program and library directories @command{gcc} will search---and don't do anything else.
4914 This is useful when @command{gcc} prints the error message
4915 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4916 To resolve this you either need to put @file{cpp0} and the other compiler
4917 components where @command{gcc} expects to find them, or you can set the environment
4918 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4919 Don't forget the trailing @samp{/}.
4920 @xref{Environment Variables}.
4922 @item -print-sysroot-headers-suffix
4923 @opindex print-sysroot-headers-suffix
4924 Print the suffix added to the target sysroot when searching for
4925 headers, or give an error if the compiler is not configured with such
4926 a suffix---and don't do anything else.
4929 @opindex dumpmachine
4930 Print the compiler's target machine (for example,
4931 @samp{i686-pc-linux-gnu})---and don't do anything else.
4934 @opindex dumpversion
4935 Print the compiler version (for example, @samp{3.0})---and don't do
4940 Print the compiler's built-in specs---and don't do anything else. (This
4941 is used when GCC itself is being built.) @xref{Spec Files}.
4943 @item -feliminate-unused-debug-types
4944 @opindex feliminate-unused-debug-types
4945 Normally, when producing DWARF2 output, GCC will emit debugging
4946 information for all types declared in a compilation
4947 unit, regardless of whether or not they are actually used
4948 in that compilation unit. Sometimes this is useful, such as
4949 if, in the debugger, you want to cast a value to a type that is
4950 not actually used in your program (but is declared). More often,
4951 however, this results in a significant amount of wasted space.
4952 With this option, GCC will avoid producing debug symbol output
4953 for types that are nowhere used in the source file being compiled.
4956 @node Optimize Options
4957 @section Options That Control Optimization
4958 @cindex optimize options
4959 @cindex options, optimization
4961 These options control various sorts of optimizations.
4963 Without any optimization option, the compiler's goal is to reduce the
4964 cost of compilation and to make debugging produce the expected
4965 results. Statements are independent: if you stop the program with a
4966 breakpoint between statements, you can then assign a new value to any
4967 variable or change the program counter to any other statement in the
4968 function and get exactly the results you would expect from the source
4971 Turning on optimization flags makes the compiler attempt to improve
4972 the performance and/or code size at the expense of compilation time
4973 and possibly the ability to debug the program.
4975 The compiler performs optimization based on the knowledge it has of
4976 the program. Optimization levels @option{-O} and above, in
4977 particular, enable @emph{unit-at-a-time} mode, which allows the
4978 compiler to consider information gained from later functions in
4979 the file when compiling a function. Compiling multiple files at
4980 once to a single output file in @emph{unit-at-a-time} mode allows
4981 the compiler to use information gained from all of the files when
4982 compiling each of them.
4984 Not all optimizations are controlled directly by a flag. Only
4985 optimizations that have a flag are listed.
4992 Optimize. Optimizing compilation takes somewhat more time, and a lot
4993 more memory for a large function.
4995 With @option{-O}, the compiler tries to reduce code size and execution
4996 time, without performing any optimizations that take a great deal of
4999 @option{-O} turns on the following optimization flags:
5000 @gccoptlist{-fdefer-pop @gol
5001 -fdelayed-branch @gol
5002 -fguess-branch-probability @gol
5003 -fcprop-registers @gol
5004 -fif-conversion @gol
5005 -fif-conversion2 @gol
5006 -fsplit-wide-types @gol
5009 -ftree-dominator-opts @gol
5013 -ftree-copyrename @gol
5016 -funit-at-a-time @gol
5019 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
5020 where doing so does not interfere with debugging.
5024 Optimize even more. GCC performs nearly all supported optimizations
5025 that do not involve a space-speed tradeoff. The compiler does not
5026 perform loop unrolling or function inlining when you specify @option{-O2}.
5027 As compared to @option{-O}, this option increases both compilation time
5028 and the performance of the generated code.
5030 @option{-O2} turns on all optimization flags specified by @option{-O}. It
5031 also turns on the following optimization flags:
5032 @gccoptlist{-fthread-jumps @gol
5034 -foptimize-sibling-calls @gol
5035 -fcse-follow-jumps -fcse-skip-blocks @gol
5036 -fgcse -fgcse-lm @gol
5037 -fexpensive-optimizations @gol
5038 -frerun-cse-after-loop @gol
5041 -fschedule-insns -fschedule-insns2 @gol
5042 -fsched-interblock -fsched-spec @gol
5044 -fstrict-aliasing -fstrict-overflow @gol
5045 -fdelete-null-pointer-checks @gol
5046 -freorder-blocks -freorder-functions @gol
5047 -falign-functions -falign-jumps @gol
5048 -falign-loops -falign-labels @gol
5052 Please note the warning under @option{-fgcse} about
5053 invoking @option{-O2} on programs that use computed gotos.
5057 Optimize yet more. @option{-O3} turns on all optimizations specified by
5058 @option{-O2} and also turns on the @option{-finline-functions},
5059 @option{-funswitch-loops}, @option{-fpredictive-commoning} and
5060 @option{-fgcse-after-reload} options.
5064 Reduce compilation time and make debugging produce the expected
5065 results. This is the default.
5069 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
5070 do not typically increase code size. It also performs further
5071 optimizations designed to reduce code size.
5073 @option{-Os} disables the following optimization flags:
5074 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
5075 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
5076 -fprefetch-loop-arrays -ftree-vect-loop-version}
5078 If you use multiple @option{-O} options, with or without level numbers,
5079 the last such option is the one that is effective.
5082 Options of the form @option{-f@var{flag}} specify machine-independent
5083 flags. Most flags have both positive and negative forms; the negative
5084 form of @option{-ffoo} would be @option{-fno-foo}. In the table
5085 below, only one of the forms is listed---the one you typically will
5086 use. You can figure out the other form by either removing @samp{no-}
5089 The following options control specific optimizations. They are either
5090 activated by @option{-O} options or are related to ones that are. You
5091 can use the following flags in the rare cases when ``fine-tuning'' of
5092 optimizations to be performed is desired.
5095 @item -fno-default-inline
5096 @opindex fno-default-inline
5097 Do not make member functions inline by default merely because they are
5098 defined inside the class scope (C++ only). Otherwise, when you specify
5099 @w{@option{-O}}, member functions defined inside class scope are compiled
5100 inline by default; i.e., you don't need to add @samp{inline} in front of
5101 the member function name.
5103 @item -fno-defer-pop
5104 @opindex fno-defer-pop
5105 Always pop the arguments to each function call as soon as that function
5106 returns. For machines which must pop arguments after a function call,
5107 the compiler normally lets arguments accumulate on the stack for several
5108 function calls and pops them all at once.
5110 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5113 @opindex fforce-addr
5114 Force memory address constants to be copied into registers before
5115 doing arithmetic on them.
5117 @item -fforward-propagate
5118 @opindex fforward-propagate
5119 Perform a forward propagation pass on RTL. The pass tries to combine two
5120 instructions and checks if the result can be simplified. If loop unrolling
5121 is active, two passes are performed and the second is scheduled after
5124 This option is enabled by default at optimization levels @option{-O2},
5125 @option{-O3}, @option{-Os}.
5127 @item -fomit-frame-pointer
5128 @opindex fomit-frame-pointer
5129 Don't keep the frame pointer in a register for functions that
5130 don't need one. This avoids the instructions to save, set up and
5131 restore frame pointers; it also makes an extra register available
5132 in many functions. @strong{It also makes debugging impossible on
5135 On some machines, such as the VAX, this flag has no effect, because
5136 the standard calling sequence automatically handles the frame pointer
5137 and nothing is saved by pretending it doesn't exist. The
5138 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
5139 whether a target machine supports this flag. @xref{Registers,,Register
5140 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
5142 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5144 @item -foptimize-sibling-calls
5145 @opindex foptimize-sibling-calls
5146 Optimize sibling and tail recursive calls.
5148 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5152 Don't pay attention to the @code{inline} keyword. Normally this option
5153 is used to keep the compiler from expanding any functions inline.
5154 Note that if you are not optimizing, no functions can be expanded inline.
5156 @item -finline-functions
5157 @opindex finline-functions
5158 Integrate all simple functions into their callers. The compiler
5159 heuristically decides which functions are simple enough to be worth
5160 integrating in this way.
5162 If all calls to a given function are integrated, and the function is
5163 declared @code{static}, then the function is normally not output as
5164 assembler code in its own right.
5166 Enabled at level @option{-O3}.
5168 @item -finline-functions-called-once
5169 @opindex finline-functions-called-once
5170 Consider all @code{static} functions called once for inlining into their
5171 caller even if they are not marked @code{inline}. If a call to a given
5172 function is integrated, then the function is not output as assembler code
5175 Enabled if @option{-funit-at-a-time} is enabled.
5177 @item -fearly-inlining
5178 @opindex fearly-inlining
5179 Inline functions marked by @code{always_inline} and functions whose body seems
5180 smaller than the function call overhead early before doing
5181 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
5182 makes profiling significantly cheaper and usually inlining faster on programs
5183 having large chains of nested wrapper functions.
5187 @item -finline-limit=@var{n}
5188 @opindex finline-limit
5189 By default, GCC limits the size of functions that can be inlined. This flag
5190 allows the control of this limit for functions that are explicitly marked as
5191 inline (i.e., marked with the inline keyword or defined within the class
5192 definition in c++). @var{n} is the size of functions that can be inlined in
5193 number of pseudo instructions (not counting parameter handling). The default
5194 value of @var{n} is 600.
5195 Increasing this value can result in more inlined code at
5196 the cost of compilation time and memory consumption. Decreasing usually makes
5197 the compilation faster and less code will be inlined (which presumably
5198 means slower programs). This option is particularly useful for programs that
5199 use inlining heavily such as those based on recursive templates with C++.
5201 Inlining is actually controlled by a number of parameters, which may be
5202 specified individually by using @option{--param @var{name}=@var{value}}.
5203 The @option{-finline-limit=@var{n}} option sets some of these parameters
5207 @item max-inline-insns-single
5208 is set to @var{n}/2.
5209 @item max-inline-insns-auto
5210 is set to @var{n}/2.
5211 @item min-inline-insns
5212 is set to 130 or @var{n}/4, whichever is smaller.
5213 @item max-inline-insns-rtl
5217 See below for a documentation of the individual
5218 parameters controlling inlining.
5220 @emph{Note:} pseudo instruction represents, in this particular context, an
5221 abstract measurement of function's size. In no way does it represent a count
5222 of assembly instructions and as such its exact meaning might change from one
5223 release to an another.
5225 @item -fkeep-inline-functions
5226 @opindex fkeep-inline-functions
5227 In C, emit @code{static} functions that are declared @code{inline}
5228 into the object file, even if the function has been inlined into all
5229 of its callers. This switch does not affect functions using the
5230 @code{extern inline} extension in GNU C89@. In C++, emit any and all
5231 inline functions into the object file.
5233 @item -fkeep-static-consts
5234 @opindex fkeep-static-consts
5235 Emit variables declared @code{static const} when optimization isn't turned
5236 on, even if the variables aren't referenced.
5238 GCC enables this option by default. If you want to force the compiler to
5239 check if the variable was referenced, regardless of whether or not
5240 optimization is turned on, use the @option{-fno-keep-static-consts} option.
5242 @item -fmerge-constants
5243 Attempt to merge identical constants (string constants and floating point
5244 constants) across compilation units.
5246 This option is the default for optimized compilation if the assembler and
5247 linker support it. Use @option{-fno-merge-constants} to inhibit this
5250 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5252 @item -fmerge-all-constants
5253 Attempt to merge identical constants and identical variables.
5255 This option implies @option{-fmerge-constants}. In addition to
5256 @option{-fmerge-constants} this considers e.g.@: even constant initialized
5257 arrays or initialized constant variables with integral or floating point
5258 types. Languages like C or C++ require each non-automatic variable to
5259 have distinct location, so using this option will result in non-conforming
5262 @item -fmodulo-sched
5263 @opindex fmodulo-sched
5264 Perform swing modulo scheduling immediately before the first scheduling
5265 pass. This pass looks at innermost loops and reorders their
5266 instructions by overlapping different iterations.
5268 @item -fno-branch-count-reg
5269 @opindex fno-branch-count-reg
5270 Do not use ``decrement and branch'' instructions on a count register,
5271 but instead generate a sequence of instructions that decrement a
5272 register, compare it against zero, then branch based upon the result.
5273 This option is only meaningful on architectures that support such
5274 instructions, which include x86, PowerPC, IA-64 and S/390.
5276 The default is @option{-fbranch-count-reg}.
5278 @item -fno-function-cse
5279 @opindex fno-function-cse
5280 Do not put function addresses in registers; make each instruction that
5281 calls a constant function contain the function's address explicitly.
5283 This option results in less efficient code, but some strange hacks
5284 that alter the assembler output may be confused by the optimizations
5285 performed when this option is not used.
5287 The default is @option{-ffunction-cse}
5289 @item -fno-zero-initialized-in-bss
5290 @opindex fno-zero-initialized-in-bss
5291 If the target supports a BSS section, GCC by default puts variables that
5292 are initialized to zero into BSS@. This can save space in the resulting
5295 This option turns off this behavior because some programs explicitly
5296 rely on variables going to the data section. E.g., so that the
5297 resulting executable can find the beginning of that section and/or make
5298 assumptions based on that.
5300 The default is @option{-fzero-initialized-in-bss}.
5302 @item -fbounds-check
5303 @opindex fbounds-check
5304 For front-ends that support it, generate additional code to check that
5305 indices used to access arrays are within the declared range. This is
5306 currently only supported by the Java and Fortran front-ends, where
5307 this option defaults to true and false respectively.
5309 @item -fmudflap -fmudflapth -fmudflapir
5313 @cindex bounds checking
5315 For front-ends that support it (C and C++), instrument all risky
5316 pointer/array dereferencing operations, some standard library
5317 string/heap functions, and some other associated constructs with
5318 range/validity tests. Modules so instrumented should be immune to
5319 buffer overflows, invalid heap use, and some other classes of C/C++
5320 programming errors. The instrumentation relies on a separate runtime
5321 library (@file{libmudflap}), which will be linked into a program if
5322 @option{-fmudflap} is given at link time. Run-time behavior of the
5323 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
5324 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
5327 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
5328 link if your program is multi-threaded. Use @option{-fmudflapir}, in
5329 addition to @option{-fmudflap} or @option{-fmudflapth}, if
5330 instrumentation should ignore pointer reads. This produces less
5331 instrumentation (and therefore faster execution) and still provides
5332 some protection against outright memory corrupting writes, but allows
5333 erroneously read data to propagate within a program.
5335 @item -fthread-jumps
5336 @opindex fthread-jumps
5337 Perform optimizations where we check to see if a jump branches to a
5338 location where another comparison subsumed by the first is found. If
5339 so, the first branch is redirected to either the destination of the
5340 second branch or a point immediately following it, depending on whether
5341 the condition is known to be true or false.
5343 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5345 @item -fsplit-wide-types
5346 @opindex fsplit-wide-types
5347 When using a type that occupies multiple registers, such as @code{long
5348 long} on a 32-bit system, split the registers apart and allocate them
5349 independently. This normally generates better code for those types,
5350 but may make debugging more difficult.
5352 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
5355 @item -fcse-follow-jumps
5356 @opindex fcse-follow-jumps
5357 In common subexpression elimination, scan through jump instructions
5358 when the target of the jump is not reached by any other path. For
5359 example, when CSE encounters an @code{if} statement with an
5360 @code{else} clause, CSE will follow the jump when the condition
5363 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5365 @item -fcse-skip-blocks
5366 @opindex fcse-skip-blocks
5367 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
5368 follow jumps which conditionally skip over blocks. When CSE
5369 encounters a simple @code{if} statement with no else clause,
5370 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
5371 body of the @code{if}.
5373 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5375 @item -frerun-cse-after-loop
5376 @opindex frerun-cse-after-loop
5377 Re-run common subexpression elimination after loop optimizations has been
5380 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5384 Perform a global common subexpression elimination pass.
5385 This pass also performs global constant and copy propagation.
5387 @emph{Note:} When compiling a program using computed gotos, a GCC
5388 extension, you may get better runtime performance if you disable
5389 the global common subexpression elimination pass by adding
5390 @option{-fno-gcse} to the command line.
5392 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5396 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
5397 attempt to move loads which are only killed by stores into themselves. This
5398 allows a loop containing a load/store sequence to be changed to a load outside
5399 the loop, and a copy/store within the loop.
5401 Enabled by default when gcse is enabled.
5405 When @option{-fgcse-sm} is enabled, a store motion pass is run after
5406 global common subexpression elimination. This pass will attempt to move
5407 stores out of loops. When used in conjunction with @option{-fgcse-lm},
5408 loops containing a load/store sequence can be changed to a load before
5409 the loop and a store after the loop.
5411 Not enabled at any optimization level.
5415 When @option{-fgcse-las} is enabled, the global common subexpression
5416 elimination pass eliminates redundant loads that come after stores to the
5417 same memory location (both partial and full redundancies).
5419 Not enabled at any optimization level.
5421 @item -fgcse-after-reload
5422 @opindex fgcse-after-reload
5423 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
5424 pass is performed after reload. The purpose of this pass is to cleanup
5427 @item -funsafe-loop-optimizations
5428 @opindex funsafe-loop-optimizations
5429 If given, the loop optimizer will assume that loop indices do not
5430 overflow, and that the loops with nontrivial exit condition are not
5431 infinite. This enables a wider range of loop optimizations even if
5432 the loop optimizer itself cannot prove that these assumptions are valid.
5433 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
5434 if it finds this kind of loop.
5436 @item -fcrossjumping
5437 @opindex crossjumping
5438 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
5439 resulting code may or may not perform better than without cross-jumping.
5441 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5443 @item -fif-conversion
5444 @opindex if-conversion
5445 Attempt to transform conditional jumps into branch-less equivalents. This
5446 include use of conditional moves, min, max, set flags and abs instructions, and
5447 some tricks doable by standard arithmetics. The use of conditional execution
5448 on chips where it is available is controlled by @code{if-conversion2}.
5450 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5452 @item -fif-conversion2
5453 @opindex if-conversion2
5454 Use conditional execution (where available) to transform conditional jumps into
5455 branch-less equivalents.
5457 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5459 @item -fdelete-null-pointer-checks
5460 @opindex fdelete-null-pointer-checks
5461 Use global dataflow analysis to identify and eliminate useless checks
5462 for null pointers. The compiler assumes that dereferencing a null
5463 pointer would have halted the program. If a pointer is checked after
5464 it has already been dereferenced, it cannot be null.
5466 In some environments, this assumption is not true, and programs can
5467 safely dereference null pointers. Use
5468 @option{-fno-delete-null-pointer-checks} to disable this optimization
5469 for programs which depend on that behavior.
5471 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5473 @item -fexpensive-optimizations
5474 @opindex fexpensive-optimizations
5475 Perform a number of minor optimizations that are relatively expensive.
5477 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5479 @item -foptimize-register-move
5481 @opindex foptimize-register-move
5483 Attempt to reassign register numbers in move instructions and as
5484 operands of other simple instructions in order to maximize the amount of
5485 register tying. This is especially helpful on machines with two-operand
5488 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5491 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5493 @item -fdelayed-branch
5494 @opindex fdelayed-branch
5495 If supported for the target machine, attempt to reorder instructions
5496 to exploit instruction slots available after delayed branch
5499 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5501 @item -fschedule-insns
5502 @opindex fschedule-insns
5503 If supported for the target machine, attempt to reorder instructions to
5504 eliminate execution stalls due to required data being unavailable. This
5505 helps machines that have slow floating point or memory load instructions
5506 by allowing other instructions to be issued until the result of the load
5507 or floating point instruction is required.
5509 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5511 @item -fschedule-insns2
5512 @opindex fschedule-insns2
5513 Similar to @option{-fschedule-insns}, but requests an additional pass of
5514 instruction scheduling after register allocation has been done. This is
5515 especially useful on machines with a relatively small number of
5516 registers and where memory load instructions take more than one cycle.
5518 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5520 @item -fno-sched-interblock
5521 @opindex fno-sched-interblock
5522 Don't schedule instructions across basic blocks. This is normally
5523 enabled by default when scheduling before register allocation, i.e.@:
5524 with @option{-fschedule-insns} or at @option{-O2} or higher.
5526 @item -fno-sched-spec
5527 @opindex fno-sched-spec
5528 Don't allow speculative motion of non-load instructions. This is normally
5529 enabled by default when scheduling before register allocation, i.e.@:
5530 with @option{-fschedule-insns} or at @option{-O2} or higher.
5532 @item -fsched-spec-load
5533 @opindex fsched-spec-load
5534 Allow speculative motion of some load instructions. This only makes
5535 sense when scheduling before register allocation, i.e.@: with
5536 @option{-fschedule-insns} or at @option{-O2} or higher.
5538 @item -fsched-spec-load-dangerous
5539 @opindex fsched-spec-load-dangerous
5540 Allow speculative motion of more load instructions. This only makes
5541 sense when scheduling before register allocation, i.e.@: with
5542 @option{-fschedule-insns} or at @option{-O2} or higher.
5544 @item -fsched-stalled-insns=@var{n}
5545 @opindex fsched-stalled-insns
5546 Define how many insns (if any) can be moved prematurely from the queue
5547 of stalled insns into the ready list, during the second scheduling pass.
5549 @item -fsched-stalled-insns-dep=@var{n}
5550 @opindex fsched-stalled-insns-dep
5551 Define how many insn groups (cycles) will be examined for a dependency
5552 on a stalled insn that is candidate for premature removal from the queue
5553 of stalled insns. Has an effect only during the second scheduling pass,
5554 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5556 @item -fsched2-use-superblocks
5557 @opindex fsched2-use-superblocks
5558 When scheduling after register allocation, do use superblock scheduling
5559 algorithm. Superblock scheduling allows motion across basic block boundaries
5560 resulting on faster schedules. This option is experimental, as not all machine
5561 descriptions used by GCC model the CPU closely enough to avoid unreliable
5562 results from the algorithm.
5564 This only makes sense when scheduling after register allocation, i.e.@: with
5565 @option{-fschedule-insns2} or at @option{-O2} or higher.
5567 @item -fsched2-use-traces
5568 @opindex fsched2-use-traces
5569 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5570 allocation and additionally perform code duplication in order to increase the
5571 size of superblocks using tracer pass. See @option{-ftracer} for details on
5574 This mode should produce faster but significantly longer programs. Also
5575 without @option{-fbranch-probabilities} the traces constructed may not
5576 match the reality and hurt the performance. This only makes
5577 sense when scheduling after register allocation, i.e.@: with
5578 @option{-fschedule-insns2} or at @option{-O2} or higher.
5582 Eliminates redundant extension instructions and move the non redundant
5583 ones to optimal placement using LCM.
5585 @item -freschedule-modulo-scheduled-loops
5586 @opindex fscheduling-in-modulo-scheduled-loops
5587 The modulo scheduling comes before the traditional scheduling, if a loop
5588 was modulo scheduled we may want to prevent the later scheduling passes
5589 from changing its schedule, we use this option to control that.
5591 @item -fcaller-saves
5592 @opindex fcaller-saves
5593 Enable values to be allocated in registers that will be clobbered by
5594 function calls, by emitting extra instructions to save and restore the
5595 registers around such calls. Such allocation is done only when it
5596 seems to result in better code than would otherwise be produced.
5598 This option is always enabled by default on certain machines, usually
5599 those which have no call-preserved registers to use instead.
5601 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5603 @item -ftree-reassoc
5604 Perform Reassociation on trees This flag is enabled by default
5605 at @option{-O} and higher.
5608 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5609 enabled by default at @option{-O2} and @option{-O3}.
5612 Perform Full Redundancy Elimination (FRE) on trees. The difference
5613 between FRE and PRE is that FRE only considers expressions
5614 that are computed on all paths leading to the redundant computation.
5615 This analysis is faster than PRE, though it exposes fewer redundancies.
5616 This flag is enabled by default at @option{-O} and higher.
5618 @item -ftree-copy-prop
5619 Perform copy propagation on trees. This pass eliminates unnecessary
5620 copy operations. This flag is enabled by default at @option{-O} and
5623 @item -ftree-store-copy-prop
5624 Perform copy propagation of memory loads and stores. This pass
5625 eliminates unnecessary copy operations in memory references
5626 (structures, global variables, arrays, etc). This flag is enabled by
5627 default at @option{-O2} and higher.
5630 Perform structural alias analysis on trees. This flag
5631 is enabled by default at @option{-O} and higher.
5634 Perform interprocedural pointer analysis.
5637 Perform forward store motion on trees. This flag is
5638 enabled by default at @option{-O} and higher.
5641 Perform sparse conditional constant propagation (CCP) on trees. This
5642 pass only operates on local scalar variables and is enabled by default
5643 at @option{-O} and higher.
5645 @item -ftree-store-ccp
5646 Perform sparse conditional constant propagation (CCP) on trees. This
5647 pass operates on both local scalar variables and memory stores and
5648 loads (global variables, structures, arrays, etc). This flag is
5649 enabled by default at @option{-O2} and higher.
5652 Perform dead code elimination (DCE) on trees. This flag is enabled by
5653 default at @option{-O} and higher.
5655 @item -ftree-dominator-opts
5656 Perform a variety of simple scalar cleanups (constant/copy
5657 propagation, redundancy elimination, range propagation and expression
5658 simplification) based on a dominator tree traversal. This also
5659 performs jump threading (to reduce jumps to jumps). This flag is
5660 enabled by default at @option{-O} and higher.
5663 Perform loop header copying on trees. This is beneficial since it increases
5664 effectiveness of code motion optimizations. It also saves one jump. This flag
5665 is enabled by default at @option{-O} and higher. It is not enabled
5666 for @option{-Os}, since it usually increases code size.
5668 @item -ftree-loop-optimize
5669 Perform loop optimizations on trees. This flag is enabled by default
5670 at @option{-O} and higher.
5672 @item -ftree-loop-linear
5673 Perform linear loop transformations on tree. This flag can improve cache
5674 performance and allow further loop optimizations to take place.
5676 @item -fcheck-data-deps
5677 Compare the results of several data dependence analyzers. This option
5678 is used for debugging the data dependence analyzers.
5680 @item -ftree-loop-im
5681 Perform loop invariant motion on trees. This pass moves only invariants that
5682 would be hard to handle at RTL level (function calls, operations that expand to
5683 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5684 operands of conditions that are invariant out of the loop, so that we can use
5685 just trivial invariantness analysis in loop unswitching. The pass also includes
5688 @item -ftree-loop-ivcanon
5689 Create a canonical counter for number of iterations in the loop for that
5690 determining number of iterations requires complicated analysis. Later
5691 optimizations then may determine the number easily. Useful especially
5692 in connection with unrolling.
5695 Perform induction variable optimizations (strength reduction, induction
5696 variable merging and induction variable elimination) on trees.
5699 Perform scalar replacement of aggregates. This pass replaces structure
5700 references with scalars to prevent committing structures to memory too
5701 early. This flag is enabled by default at @option{-O} and higher.
5703 @item -ftree-copyrename
5704 Perform copy renaming on trees. This pass attempts to rename compiler
5705 temporaries to other variables at copy locations, usually resulting in
5706 variable names which more closely resemble the original variables. This flag
5707 is enabled by default at @option{-O} and higher.
5710 Perform temporary expression replacement during the SSA->normal phase. Single
5711 use/single def temporaries are replaced at their use location with their
5712 defining expression. This results in non-GIMPLE code, but gives the expanders
5713 much more complex trees to work on resulting in better RTL generation. This is
5714 enabled by default at @option{-O} and higher.
5716 @item -ftree-vectorize
5717 Perform loop vectorization on trees.
5719 @item -ftree-vect-loop-version
5720 @opindex ftree-vect-loop-version
5721 Perform loop versioning when doing loop vectorization on trees. When a loop
5722 appears to be vectorizable except that data alignment or data dependence cannot
5723 be determined at compile time then vectorized and non-vectorized versions of
5724 the loop are generated along with runtime checks for alignment or dependence
5725 to control which version is executed. This option is enabled by default
5726 except at level @option{-Os} where it is disabled.
5728 @item -fvect-cost-model
5729 Enable cost model for vectorization.
5732 Perform Value Range Propagation on trees. This is similar to the
5733 constant propagation pass, but instead of values, ranges of values are
5734 propagated. This allows the optimizers to remove unnecessary range
5735 checks like array bound checks and null pointer checks. This is
5736 enabled by default at @option{-O2} and higher. Null pointer check
5737 elimination is only done if @option{-fdelete-null-pointer-checks} is
5742 Perform tail duplication to enlarge superblock size. This transformation
5743 simplifies the control flow of the function allowing other optimizations to do
5746 @item -funroll-loops
5747 @opindex funroll-loops
5748 Unroll loops whose number of iterations can be determined at compile
5749 time or upon entry to the loop. @option{-funroll-loops} implies
5750 @option{-frerun-cse-after-loop}. This option makes code larger,
5751 and may or may not make it run faster.
5753 @item -funroll-all-loops
5754 @opindex funroll-all-loops
5755 Unroll all loops, even if their number of iterations is uncertain when
5756 the loop is entered. This usually makes programs run more slowly.
5757 @option{-funroll-all-loops} implies the same options as
5758 @option{-funroll-loops},
5760 @item -fsplit-ivs-in-unroller
5761 @opindex fsplit-ivs-in-unroller
5762 Enables expressing of values of induction variables in later iterations
5763 of the unrolled loop using the value in the first iteration. This breaks
5764 long dependency chains, thus improving efficiency of the scheduling passes.
5766 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5767 same effect. However in cases the loop body is more complicated than
5768 a single basic block, this is not reliable. It also does not work at all
5769 on some of the architectures due to restrictions in the CSE pass.
5771 This optimization is enabled by default.
5773 @item -fvariable-expansion-in-unroller
5774 @opindex fvariable-expansion-in-unroller
5775 With this option, the compiler will create multiple copies of some
5776 local variables when unrolling a loop which can result in superior code.
5778 @item -fpredictive-commoning
5779 @opindex fpredictive-commoning
5780 Perform predictive commoning optimization, i.e., reusing computations
5781 (especially memory loads and stores) performed in previous
5782 iterations of loops.
5784 This option is enabled at level @option{-O3}.
5786 @item -fprefetch-loop-arrays
5787 @opindex fprefetch-loop-arrays
5788 If supported by the target machine, generate instructions to prefetch
5789 memory to improve the performance of loops that access large arrays.
5791 This option may generate better or worse code; results are highly
5792 dependent on the structure of loops within the source code.
5794 Disabled at level @option{-Os}.
5797 @itemx -fno-peephole2
5798 @opindex fno-peephole
5799 @opindex fno-peephole2
5800 Disable any machine-specific peephole optimizations. The difference
5801 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5802 are implemented in the compiler; some targets use one, some use the
5803 other, a few use both.
5805 @option{-fpeephole} is enabled by default.
5806 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5808 @item -fno-guess-branch-probability
5809 @opindex fno-guess-branch-probability
5810 Do not guess branch probabilities using heuristics.
5812 GCC will use heuristics to guess branch probabilities if they are
5813 not provided by profiling feedback (@option{-fprofile-arcs}). These
5814 heuristics are based on the control flow graph. If some branch probabilities
5815 are specified by @samp{__builtin_expect}, then the heuristics will be
5816 used to guess branch probabilities for the rest of the control flow graph,
5817 taking the @samp{__builtin_expect} info into account. The interactions
5818 between the heuristics and @samp{__builtin_expect} can be complex, and in
5819 some cases, it may be useful to disable the heuristics so that the effects
5820 of @samp{__builtin_expect} are easier to understand.
5822 The default is @option{-fguess-branch-probability} at levels
5823 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5825 @item -freorder-blocks
5826 @opindex freorder-blocks
5827 Reorder basic blocks in the compiled function in order to reduce number of
5828 taken branches and improve code locality.
5830 Enabled at levels @option{-O2}, @option{-O3}.
5832 @item -freorder-blocks-and-partition
5833 @opindex freorder-blocks-and-partition
5834 In addition to reordering basic blocks in the compiled function, in order
5835 to reduce number of taken branches, partitions hot and cold basic blocks
5836 into separate sections of the assembly and .o files, to improve
5837 paging and cache locality performance.
5839 This optimization is automatically turned off in the presence of
5840 exception handling, for linkonce sections, for functions with a user-defined
5841 section attribute and on any architecture that does not support named
5844 @item -freorder-functions
5845 @opindex freorder-functions
5846 Reorder functions in the object file in order to
5847 improve code locality. This is implemented by using special
5848 subsections @code{.text.hot} for most frequently executed functions and
5849 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5850 the linker so object file format must support named sections and linker must
5851 place them in a reasonable way.
5853 Also profile feedback must be available in to make this option effective. See
5854 @option{-fprofile-arcs} for details.
5856 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5858 @item -fstrict-aliasing
5859 @opindex fstrict-aliasing
5860 Allows the compiler to assume the strictest aliasing rules applicable to
5861 the language being compiled. For C (and C++), this activates
5862 optimizations based on the type of expressions. In particular, an
5863 object of one type is assumed never to reside at the same address as an
5864 object of a different type, unless the types are almost the same. For
5865 example, an @code{unsigned int} can alias an @code{int}, but not a
5866 @code{void*} or a @code{double}. A character type may alias any other
5869 Pay special attention to code like this:
5882 The practice of reading from a different union member than the one most
5883 recently written to (called ``type-punning'') is common. Even with
5884 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5885 is accessed through the union type. So, the code above will work as
5886 expected. However, this code might not:
5897 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5899 @item -fstrict-overflow
5900 @opindex fstrict-overflow
5901 Allow the compiler to assume strict signed overflow rules, depending
5902 on the language being compiled. For C (and C++) this means that
5903 overflow when doing arithmetic with signed numbers is undefined, which
5904 means that the compiler may assume that it will not happen. This
5905 permits various optimizations. For example, the compiler will assume
5906 that an expression like @code{i + 10 > i} will always be true for
5907 signed @code{i}. This assumption is only valid if signed overflow is
5908 undefined, as the expression is false if @code{i + 10} overflows when
5909 using twos complement arithmetic. When this option is in effect any
5910 attempt to determine whether an operation on signed numbers will
5911 overflow must be written carefully to not actually involve overflow.
5913 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
5914 that signed overflow is fully defined: it wraps. When
5915 @option{-fwrapv} is used, there is no difference between
5916 @option{-fstrict-overflow} and @option{-fno-strict-overflow}. With
5917 @option{-fwrapv} certain types of overflow are permitted. For
5918 example, if the compiler gets an overflow when doing arithmetic on
5919 constants, the overflowed value can still be used with
5920 @option{-fwrapv}, but not otherwise.
5922 The @option{-fstrict-overflow} option is enabled at levels
5923 @option{-O2}, @option{-O3}, @option{-Os}.
5925 @item -falign-functions
5926 @itemx -falign-functions=@var{n}
5927 @opindex falign-functions
5928 Align the start of functions to the next power-of-two greater than
5929 @var{n}, skipping up to @var{n} bytes. For instance,
5930 @option{-falign-functions=32} aligns functions to the next 32-byte
5931 boundary, but @option{-falign-functions=24} would align to the next
5932 32-byte boundary only if this can be done by skipping 23 bytes or less.
5934 @option{-fno-align-functions} and @option{-falign-functions=1} are
5935 equivalent and mean that functions will not be aligned.
5937 Some assemblers only support this flag when @var{n} is a power of two;
5938 in that case, it is rounded up.
5940 If @var{n} is not specified or is zero, use a machine-dependent default.
5942 Enabled at levels @option{-O2}, @option{-O3}.
5944 @item -falign-labels
5945 @itemx -falign-labels=@var{n}
5946 @opindex falign-labels
5947 Align all branch targets to a power-of-two boundary, skipping up to
5948 @var{n} bytes like @option{-falign-functions}. This option can easily
5949 make code slower, because it must insert dummy operations for when the
5950 branch target is reached in the usual flow of the code.
5952 @option{-fno-align-labels} and @option{-falign-labels=1} are
5953 equivalent and mean that labels will not be aligned.
5955 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5956 are greater than this value, then their values are used instead.
5958 If @var{n} is not specified or is zero, use a machine-dependent default
5959 which is very likely to be @samp{1}, meaning no alignment.
5961 Enabled at levels @option{-O2}, @option{-O3}.
5964 @itemx -falign-loops=@var{n}
5965 @opindex falign-loops
5966 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5967 like @option{-falign-functions}. The hope is that the loop will be
5968 executed many times, which will make up for any execution of the dummy
5971 @option{-fno-align-loops} and @option{-falign-loops=1} are
5972 equivalent and mean that loops will not be aligned.
5974 If @var{n} is not specified or is zero, use a machine-dependent default.
5976 Enabled at levels @option{-O2}, @option{-O3}.
5979 @itemx -falign-jumps=@var{n}
5980 @opindex falign-jumps
5981 Align branch targets to a power-of-two boundary, for branch targets
5982 where the targets can only be reached by jumping, skipping up to @var{n}
5983 bytes like @option{-falign-functions}. In this case, no dummy operations
5986 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5987 equivalent and mean that loops will not be aligned.
5989 If @var{n} is not specified or is zero, use a machine-dependent default.
5991 Enabled at levels @option{-O2}, @option{-O3}.
5993 @item -funit-at-a-time
5994 @opindex funit-at-a-time
5995 Parse the whole compilation unit before starting to produce code.
5996 This allows some extra optimizations to take place but consumes
5997 more memory (in general). There are some compatibility issues
5998 with @emph{unit-at-a-time} mode:
6001 enabling @emph{unit-at-a-time} mode may change the order
6002 in which functions, variables, and top-level @code{asm} statements
6003 are emitted, and will likely break code relying on some particular
6004 ordering. The majority of such top-level @code{asm} statements,
6005 though, can be replaced by @code{section} attributes. The
6006 @option{fno-toplevel-reorder} option may be used to keep the ordering
6007 used in the input file, at the cost of some optimizations.
6010 @emph{unit-at-a-time} mode removes unreferenced static variables
6011 and functions. This may result in undefined references
6012 when an @code{asm} statement refers directly to variables or functions
6013 that are otherwise unused. In that case either the variable/function
6014 shall be listed as an operand of the @code{asm} statement operand or,
6015 in the case of top-level @code{asm} statements the attribute @code{used}
6016 shall be used on the declaration.
6019 Static functions now can use non-standard passing conventions that
6020 may break @code{asm} statements calling functions directly. Again,
6021 attribute @code{used} will prevent this behavior.
6024 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
6025 but this scheme may not be supported by future releases of GCC@.
6027 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6029 @item -fno-toplevel-reorder
6030 Do not reorder top-level functions, variables, and @code{asm}
6031 statements. Output them in the same order that they appear in the
6032 input file. When this option is used, unreferenced static variables
6033 will not be removed. This option is intended to support existing code
6034 which relies on a particular ordering. For new code, it is better to
6039 Constructs webs as commonly used for register allocation purposes and assign
6040 each web individual pseudo register. This allows the register allocation pass
6041 to operate on pseudos directly, but also strengthens several other optimization
6042 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
6043 however, make debugging impossible, since variables will no longer stay in a
6046 Enabled by default with @option{-funroll-loops}.
6048 @item -fwhole-program
6049 @opindex fwhole-program
6050 Assume that the current compilation unit represents whole program being
6051 compiled. All public functions and variables with the exception of @code{main}
6052 and those merged by attribute @code{externally_visible} become static functions
6053 and in a affect gets more aggressively optimized by interprocedural optimizers.
6054 While this option is equivalent to proper use of @code{static} keyword for
6055 programs consisting of single file, in combination with option
6056 @option{--combine} this flag can be used to compile most of smaller scale C
6057 programs since the functions and variables become local for the whole combined
6058 compilation unit, not for the single source file itself.
6061 @item -fno-cprop-registers
6062 @opindex fno-cprop-registers
6063 After register allocation and post-register allocation instruction splitting,
6064 we perform a copy-propagation pass to try to reduce scheduling dependencies
6065 and occasionally eliminate the copy.
6067 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6069 @item -fprofile-generate
6070 @opindex fprofile-generate
6072 Enable options usually used for instrumenting application to produce
6073 profile useful for later recompilation with profile feedback based
6074 optimization. You must use @option{-fprofile-generate} both when
6075 compiling and when linking your program.
6077 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
6080 @opindex fprofile-use
6081 Enable profile feedback directed optimizations, and optimizations
6082 generally profitable only with profile feedback available.
6084 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
6085 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
6087 By default, GCC emits an error message if the feedback profiles do not
6088 match the source code. This error can be turned into a warning by using
6089 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
6093 The following options control compiler behavior regarding floating
6094 point arithmetic. These options trade off between speed and
6095 correctness. All must be specifically enabled.
6099 @opindex ffloat-store
6100 Do not store floating point variables in registers, and inhibit other
6101 options that might change whether a floating point value is taken from a
6104 @cindex floating point precision
6105 This option prevents undesirable excess precision on machines such as
6106 the 68000 where the floating registers (of the 68881) keep more
6107 precision than a @code{double} is supposed to have. Similarly for the
6108 x86 architecture. For most programs, the excess precision does only
6109 good, but a few programs rely on the precise definition of IEEE floating
6110 point. Use @option{-ffloat-store} for such programs, after modifying
6111 them to store all pertinent intermediate computations into variables.
6115 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
6116 @option{-fno-trapping-math}, @option{-ffinite-math-only},
6117 @option{-fno-rounding-math}, @option{-fno-signaling-nans},
6118 @option{-fno-signed-zeros} and @option{fcx-limited-range}.
6120 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
6122 This option is not turned on by any @option{-O} option since
6123 it can result in incorrect output for programs which depend on
6124 an exact implementation of IEEE or ISO rules/specifications for
6125 math functions. It may, however, yield faster code for programs
6126 that do not require the guarantees of these specifications.
6128 @item -fno-math-errno
6129 @opindex fno-math-errno
6130 Do not set ERRNO after calling math functions that are executed
6131 with a single instruction, e.g., sqrt. A program that relies on
6132 IEEE exceptions for math error handling may want to use this flag
6133 for speed while maintaining IEEE arithmetic compatibility.
6135 This option is not turned on by any @option{-O} option since
6136 it can result in incorrect output for programs which depend on
6137 an exact implementation of IEEE or ISO rules/specifications for
6138 math functions. It may, however, yield faster code for programs
6139 that do not require the guarantees of these specifications.
6141 The default is @option{-fmath-errno}.
6143 On Darwin systems, the math library never sets @code{errno}. There is
6144 therefore no reason for the compiler to consider the possibility that
6145 it might, and @option{-fno-math-errno} is the default.
6147 @item -funsafe-math-optimizations
6148 @opindex funsafe-math-optimizations
6149 Allow optimizations for floating-point arithmetic that (a) assume
6150 that arguments and results are valid and (b) may violate IEEE or
6151 ANSI standards. When used at link-time, it may include libraries
6152 or startup files that change the default FPU control word or other
6153 similar optimizations.
6155 This option is not turned on by any @option{-O} option since
6156 it can result in incorrect output for programs which depend on
6157 an exact implementation of IEEE or ISO rules/specifications for
6158 math functions. It may, however, yield faster code for programs
6159 that do not require the guarantees of these specifications.
6161 The default is @option{-fno-unsafe-math-optimizations}.
6163 @item -ffinite-math-only
6164 @opindex ffinite-math-only
6165 Allow optimizations for floating-point arithmetic that assume
6166 that arguments and results are not NaNs or +-Infs.
6168 This option is not turned on by any @option{-O} option since
6169 it can result in incorrect output for programs which depend on
6170 an exact implementation of IEEE or ISO rules/specifications for
6171 math functions. It may, however, yield faster code for programs
6172 that do not require the guarantees of these specifications.
6174 The default is @option{-fno-finite-math-only}.
6176 @item -fno-signed-zeros
6177 @opindex fno-signed-zeros
6178 Allow optimizations for floating point arithmetic that ignore the
6179 signedness of zero. IEEE arithmetic specifies the behavior of
6180 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
6181 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
6182 This option implies that the sign of a zero result isn't significant.
6184 The default is @option{-fsigned-zeros}.
6186 @item -fno-trapping-math
6187 @opindex fno-trapping-math
6188 Compile code assuming that floating-point operations cannot generate
6189 user-visible traps. These traps include division by zero, overflow,
6190 underflow, inexact result and invalid operation. This option implies
6191 @option{-fno-signaling-nans}. Setting this option may allow faster
6192 code if one relies on ``non-stop'' IEEE arithmetic, for example.
6194 This option should never be turned on by any @option{-O} option since
6195 it can result in incorrect output for programs which depend on
6196 an exact implementation of IEEE or ISO rules/specifications for
6199 The default is @option{-ftrapping-math}.
6201 @item -frounding-math
6202 @opindex frounding-math
6203 Disable transformations and optimizations that assume default floating
6204 point rounding behavior. This is round-to-zero for all floating point
6205 to integer conversions, and round-to-nearest for all other arithmetic
6206 truncations. This option should be specified for programs that change
6207 the FP rounding mode dynamically, or that may be executed with a
6208 non-default rounding mode. This option disables constant folding of
6209 floating point expressions at compile-time (which may be affected by
6210 rounding mode) and arithmetic transformations that are unsafe in the
6211 presence of sign-dependent rounding modes.
6213 The default is @option{-fno-rounding-math}.
6215 This option is experimental and does not currently guarantee to
6216 disable all GCC optimizations that are affected by rounding mode.
6217 Future versions of GCC may provide finer control of this setting
6218 using C99's @code{FENV_ACCESS} pragma. This command line option
6219 will be used to specify the default state for @code{FENV_ACCESS}.
6221 @item -frtl-abstract-sequences
6222 @opindex frtl-abstract-sequences
6223 It is a size optimization method. This option is to find identical
6224 sequences of code, which can be turned into pseudo-procedures and
6225 then replace all occurrences with calls to the newly created
6226 subroutine. It is kind of an opposite of @option{-finline-functions}.
6227 This optimization runs at RTL level.
6229 @item -fsignaling-nans
6230 @opindex fsignaling-nans
6231 Compile code assuming that IEEE signaling NaNs may generate user-visible
6232 traps during floating-point operations. Setting this option disables
6233 optimizations that may change the number of exceptions visible with
6234 signaling NaNs. This option implies @option{-ftrapping-math}.
6236 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
6239 The default is @option{-fno-signaling-nans}.
6241 This option is experimental and does not currently guarantee to
6242 disable all GCC optimizations that affect signaling NaN behavior.
6244 @item -fsingle-precision-constant
6245 @opindex fsingle-precision-constant
6246 Treat floating point constant as single precision constant instead of
6247 implicitly converting it to double precision constant.
6249 @item -fcx-limited-range
6250 @itemx -fno-cx-limited-range
6251 @opindex fcx-limited-range
6252 @opindex fno-cx-limited-range
6253 When enabled, this option states that a range reduction step is not
6254 needed when performing complex division. The default is
6255 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
6257 This option controls the default setting of the ISO C99
6258 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
6263 The following options control optimizations that may improve
6264 performance, but are not enabled by any @option{-O} options. This
6265 section includes experimental options that may produce broken code.
6268 @item -fbranch-probabilities
6269 @opindex fbranch-probabilities
6270 After running a program compiled with @option{-fprofile-arcs}
6271 (@pxref{Debugging Options,, Options for Debugging Your Program or
6272 @command{gcc}}), you can compile it a second time using
6273 @option{-fbranch-probabilities}, to improve optimizations based on
6274 the number of times each branch was taken. When the program
6275 compiled with @option{-fprofile-arcs} exits it saves arc execution
6276 counts to a file called @file{@var{sourcename}.gcda} for each source
6277 file. The information in this data file is very dependent on the
6278 structure of the generated code, so you must use the same source code
6279 and the same optimization options for both compilations.
6281 With @option{-fbranch-probabilities}, GCC puts a
6282 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
6283 These can be used to improve optimization. Currently, they are only
6284 used in one place: in @file{reorg.c}, instead of guessing which path a
6285 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
6286 exactly determine which path is taken more often.
6288 @item -fprofile-values
6289 @opindex fprofile-values
6290 If combined with @option{-fprofile-arcs}, it adds code so that some
6291 data about values of expressions in the program is gathered.
6293 With @option{-fbranch-probabilities}, it reads back the data gathered
6294 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
6295 notes to instructions for their later usage in optimizations.
6297 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
6301 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
6302 a code to gather information about values of expressions.
6304 With @option{-fbranch-probabilities}, it reads back the data gathered
6305 and actually performs the optimizations based on them.
6306 Currently the optimizations include specialization of division operation
6307 using the knowledge about the value of the denominator.
6309 @item -frename-registers
6310 @opindex frename-registers
6311 Attempt to avoid false dependencies in scheduled code by making use
6312 of registers left over after register allocation. This optimization
6313 will most benefit processors with lots of registers. Depending on the
6314 debug information format adopted by the target, however, it can
6315 make debugging impossible, since variables will no longer stay in
6316 a ``home register''.
6318 Enabled by default with @option{-funroll-loops}.
6322 Perform tail duplication to enlarge superblock size. This transformation
6323 simplifies the control flow of the function allowing other optimizations to do
6326 Enabled with @option{-fprofile-use}.
6328 @item -funroll-loops
6329 @opindex funroll-loops
6330 Unroll loops whose number of iterations can be determined at compile time or
6331 upon entry to the loop. @option{-funroll-loops} implies
6332 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
6333 It also turns on complete loop peeling (i.e.@: complete removal of loops with
6334 small constant number of iterations). This option makes code larger, and may
6335 or may not make it run faster.
6337 Enabled with @option{-fprofile-use}.
6339 @item -funroll-all-loops
6340 @opindex funroll-all-loops
6341 Unroll all loops, even if their number of iterations is uncertain when
6342 the loop is entered. This usually makes programs run more slowly.
6343 @option{-funroll-all-loops} implies the same options as
6344 @option{-funroll-loops}.
6347 @opindex fpeel-loops
6348 Peels the loops for that there is enough information that they do not
6349 roll much (from profile feedback). It also turns on complete loop peeling
6350 (i.e.@: complete removal of loops with small constant number of iterations).
6352 Enabled with @option{-fprofile-use}.
6354 @item -fmove-loop-invariants
6355 @opindex fmove-loop-invariants
6356 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
6357 at level @option{-O1}
6359 @item -funswitch-loops
6360 @opindex funswitch-loops
6361 Move branches with loop invariant conditions out of the loop, with duplicates
6362 of the loop on both branches (modified according to result of the condition).
6364 @item -ffunction-sections
6365 @itemx -fdata-sections
6366 @opindex ffunction-sections
6367 @opindex fdata-sections
6368 Place each function or data item into its own section in the output
6369 file if the target supports arbitrary sections. The name of the
6370 function or the name of the data item determines the section's name
6373 Use these options on systems where the linker can perform optimizations
6374 to improve locality of reference in the instruction space. Most systems
6375 using the ELF object format and SPARC processors running Solaris 2 have
6376 linkers with such optimizations. AIX may have these optimizations in
6379 Only use these options when there are significant benefits from doing
6380 so. When you specify these options, the assembler and linker will
6381 create larger object and executable files and will also be slower.
6382 You will not be able to use @code{gprof} on all systems if you
6383 specify this option and you may have problems with debugging if
6384 you specify both this option and @option{-g}.
6386 @item -fbranch-target-load-optimize
6387 @opindex fbranch-target-load-optimize
6388 Perform branch target register load optimization before prologue / epilogue
6390 The use of target registers can typically be exposed only during reload,
6391 thus hoisting loads out of loops and doing inter-block scheduling needs
6392 a separate optimization pass.
6394 @item -fbranch-target-load-optimize2
6395 @opindex fbranch-target-load-optimize2
6396 Perform branch target register load optimization after prologue / epilogue
6399 @item -fbtr-bb-exclusive
6400 @opindex fbtr-bb-exclusive
6401 When performing branch target register load optimization, don't reuse
6402 branch target registers in within any basic block.
6404 @item -fstack-protector
6405 Emit extra code to check for buffer overflows, such as stack smashing
6406 attacks. This is done by adding a guard variable to functions with
6407 vulnerable objects. This includes functions that call alloca, and
6408 functions with buffers larger than 8 bytes. The guards are initialized
6409 when a function is entered and then checked when the function exits.
6410 If a guard check fails, an error message is printed and the program exits.
6412 @item -fstack-protector-all
6413 Like @option{-fstack-protector} except that all functions are protected.
6415 @item -fsection-anchors
6416 @opindex fsection-anchors
6417 Try to reduce the number of symbolic address calculations by using
6418 shared ``anchor'' symbols to address nearby objects. This transformation
6419 can help to reduce the number of GOT entries and GOT accesses on some
6422 For example, the implementation of the following function @code{foo}:
6426 int foo (void) @{ return a + b + c; @}
6429 would usually calculate the addresses of all three variables, but if you
6430 compile it with @option{-fsection-anchors}, it will access the variables
6431 from a common anchor point instead. The effect is similar to the
6432 following pseudocode (which isn't valid C):
6437 register int *xr = &x;
6438 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6442 Not all targets support this option.
6444 @item --param @var{name}=@var{value}
6446 In some places, GCC uses various constants to control the amount of
6447 optimization that is done. For example, GCC will not inline functions
6448 that contain more that a certain number of instructions. You can
6449 control some of these constants on the command-line using the
6450 @option{--param} option.
6452 The names of specific parameters, and the meaning of the values, are
6453 tied to the internals of the compiler, and are subject to change
6454 without notice in future releases.
6456 In each case, the @var{value} is an integer. The allowable choices for
6457 @var{name} are given in the following table:
6460 @item salias-max-implicit-fields
6461 The maximum number of fields in a variable without direct
6462 structure accesses for which structure aliasing will consider trying
6463 to track each field. The default is 5
6465 @item salias-max-array-elements
6466 The maximum number of elements an array can have and its elements
6467 still be tracked individually by structure aliasing. The default is 4
6469 @item sra-max-structure-size
6470 The maximum structure size, in bytes, at which the scalar replacement
6471 of aggregates (SRA) optimization will perform block copies. The
6472 default value, 0, implies that GCC will select the most appropriate
6475 @item sra-field-structure-ratio
6476 The threshold ratio (as a percentage) between instantiated fields and
6477 the complete structure size. We say that if the ratio of the number
6478 of bytes in instantiated fields to the number of bytes in the complete
6479 structure exceeds this parameter, then block copies are not used. The
6482 @item max-crossjump-edges
6483 The maximum number of incoming edges to consider for crossjumping.
6484 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
6485 the number of edges incoming to each block. Increasing values mean
6486 more aggressive optimization, making the compile time increase with
6487 probably small improvement in executable size.
6489 @item min-crossjump-insns
6490 The minimum number of instructions which must be matched at the end
6491 of two blocks before crossjumping will be performed on them. This
6492 value is ignored in the case where all instructions in the block being
6493 crossjumped from are matched. The default value is 5.
6495 @item max-grow-copy-bb-insns
6496 The maximum code size expansion factor when copying basic blocks
6497 instead of jumping. The expansion is relative to a jump instruction.
6498 The default value is 8.
6500 @item max-goto-duplication-insns
6501 The maximum number of instructions to duplicate to a block that jumps
6502 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
6503 passes, GCC factors computed gotos early in the compilation process,
6504 and unfactors them as late as possible. Only computed jumps at the
6505 end of a basic blocks with no more than max-goto-duplication-insns are
6506 unfactored. The default value is 8.
6508 @item max-delay-slot-insn-search
6509 The maximum number of instructions to consider when looking for an
6510 instruction to fill a delay slot. If more than this arbitrary number of
6511 instructions is searched, the time savings from filling the delay slot
6512 will be minimal so stop searching. Increasing values mean more
6513 aggressive optimization, making the compile time increase with probably
6514 small improvement in executable run time.
6516 @item max-delay-slot-live-search
6517 When trying to fill delay slots, the maximum number of instructions to
6518 consider when searching for a block with valid live register
6519 information. Increasing this arbitrarily chosen value means more
6520 aggressive optimization, increasing the compile time. This parameter
6521 should be removed when the delay slot code is rewritten to maintain the
6524 @item max-gcse-memory
6525 The approximate maximum amount of memory that will be allocated in
6526 order to perform the global common subexpression elimination
6527 optimization. If more memory than specified is required, the
6528 optimization will not be done.
6530 @item max-gcse-passes
6531 The maximum number of passes of GCSE to run. The default is 1.
6533 @item max-pending-list-length
6534 The maximum number of pending dependencies scheduling will allow
6535 before flushing the current state and starting over. Large functions
6536 with few branches or calls can create excessively large lists which
6537 needlessly consume memory and resources.
6539 @item max-inline-insns-single
6540 Several parameters control the tree inliner used in gcc.
6541 This number sets the maximum number of instructions (counted in GCC's
6542 internal representation) in a single function that the tree inliner
6543 will consider for inlining. This only affects functions declared
6544 inline and methods implemented in a class declaration (C++).
6545 The default value is 450.
6547 @item max-inline-insns-auto
6548 When you use @option{-finline-functions} (included in @option{-O3}),
6549 a lot of functions that would otherwise not be considered for inlining
6550 by the compiler will be investigated. To those functions, a different
6551 (more restrictive) limit compared to functions declared inline can
6553 The default value is 90.
6555 @item large-function-insns
6556 The limit specifying really large functions. For functions larger than this
6557 limit after inlining inlining is constrained by
6558 @option{--param large-function-growth}. This parameter is useful primarily
6559 to avoid extreme compilation time caused by non-linear algorithms used by the
6561 This parameter is ignored when @option{-funit-at-a-time} is not used.
6562 The default value is 2700.
6564 @item large-function-growth
6565 Specifies maximal growth of large function caused by inlining in percents.
6566 This parameter is ignored when @option{-funit-at-a-time} is not used.
6567 The default value is 100 which limits large function growth to 2.0 times
6570 @item large-unit-insns
6571 The limit specifying large translation unit. Growth caused by inlining of
6572 units larger than this limit is limited by @option{--param inline-unit-growth}.
6573 For small units this might be too tight (consider unit consisting of function A
6574 that is inline and B that just calls A three time. If B is small relative to
6575 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6576 large units consisting of small inlininable functions however the overall unit
6577 growth limit is needed to avoid exponential explosion of code size. Thus for
6578 smaller units, the size is increased to @option{--param large-unit-insns}
6579 before applying @option{--param inline-unit-growth}. The default is 10000
6581 @item inline-unit-growth
6582 Specifies maximal overall growth of the compilation unit caused by inlining.
6583 This parameter is ignored when @option{-funit-at-a-time} is not used.
6584 The default value is 30 which limits unit growth to 1.3 times the original
6587 @item large-stack-frame
6588 The limit specifying large stack frames. While inlining the algorithm is trying
6589 to not grow past this limit too much. Default value is 256 bytes.
6591 @item large-stack-frame-growth
6592 Specifies maximal growth of large stack frames caused by inlining in percents.
6593 The default value is 1000 which limits large stack frame growth to 11 times
6596 @item max-inline-insns-recursive
6597 @itemx max-inline-insns-recursive-auto
6598 Specifies maximum number of instructions out-of-line copy of self recursive inline
6599 function can grow into by performing recursive inlining.
6601 For functions declared inline @option{--param max-inline-insns-recursive} is
6602 taken into account. For function not declared inline, recursive inlining
6603 happens only when @option{-finline-functions} (included in @option{-O3}) is
6604 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6605 default value is 450.
6607 @item max-inline-recursive-depth
6608 @itemx max-inline-recursive-depth-auto
6609 Specifies maximum recursion depth used by the recursive inlining.
6611 For functions declared inline @option{--param max-inline-recursive-depth} is
6612 taken into account. For function not declared inline, recursive inlining
6613 happens only when @option{-finline-functions} (included in @option{-O3}) is
6614 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6615 default value is 450.
6617 @item min-inline-recursive-probability
6618 Recursive inlining is profitable only for function having deep recursion
6619 in average and can hurt for function having little recursion depth by
6620 increasing the prologue size or complexity of function body to other
6623 When profile feedback is available (see @option{-fprofile-generate}) the actual
6624 recursion depth can be guessed from probability that function will recurse via
6625 given call expression. This parameter limits inlining only to call expression
6626 whose probability exceeds given threshold (in percents). The default value is
6629 @item inline-call-cost
6630 Specify cost of call instruction relative to simple arithmetics operations
6631 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6632 functions and at the same time increases size of leaf function that is believed to
6633 reduce function size by being inlined. In effect it increases amount of
6634 inlining for code having large abstraction penalty (many functions that just
6635 pass the arguments to other functions) and decrease inlining for code with low
6636 abstraction penalty. The default value is 16.
6638 @item min-vect-loop-bound
6639 The minimum number of iterations under which a loop will not get vectorized
6640 when @option{-ftree-vectorize} is used. The number of iterations after
6641 vectorization needs to be greater than the value specified by this option
6642 to allow vectorization. The default value is 0.
6644 @item max-unrolled-insns
6645 The maximum number of instructions that a loop should have if that loop
6646 is unrolled, and if the loop is unrolled, it determines how many times
6647 the loop code is unrolled.
6649 @item max-average-unrolled-insns
6650 The maximum number of instructions biased by probabilities of their execution
6651 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6652 it determines how many times the loop code is unrolled.
6654 @item max-unroll-times
6655 The maximum number of unrollings of a single loop.
6657 @item max-peeled-insns
6658 The maximum number of instructions that a loop should have if that loop
6659 is peeled, and if the loop is peeled, it determines how many times
6660 the loop code is peeled.
6662 @item max-peel-times
6663 The maximum number of peelings of a single loop.
6665 @item max-completely-peeled-insns
6666 The maximum number of insns of a completely peeled loop.
6668 @item max-completely-peel-times
6669 The maximum number of iterations of a loop to be suitable for complete peeling.
6671 @item max-unswitch-insns
6672 The maximum number of insns of an unswitched loop.
6674 @item max-unswitch-level
6675 The maximum number of branches unswitched in a single loop.
6678 The minimum cost of an expensive expression in the loop invariant motion.
6680 @item iv-consider-all-candidates-bound
6681 Bound on number of candidates for induction variables below that
6682 all candidates are considered for each use in induction variable
6683 optimizations. Only the most relevant candidates are considered
6684 if there are more candidates, to avoid quadratic time complexity.
6686 @item iv-max-considered-uses
6687 The induction variable optimizations give up on loops that contain more
6688 induction variable uses.
6690 @item iv-always-prune-cand-set-bound
6691 If number of candidates in the set is smaller than this value,
6692 we always try to remove unnecessary ivs from the set during its
6693 optimization when a new iv is added to the set.
6695 @item scev-max-expr-size
6696 Bound on size of expressions used in the scalar evolutions analyzer.
6697 Large expressions slow the analyzer.
6699 @item omega-max-vars
6700 The maximum number of variables in an Omega constraint system.
6701 The default value is 128.
6703 @item omega-max-geqs
6704 The maximum number of inequalities in an Omega constraint system.
6705 The default value is 256.
6708 The maximum number of equalities in an Omega constraint system.
6709 The default value is 128.
6711 @item omega-max-wild-cards
6712 The maximum number of wildcard variables that the Omega solver will
6713 be able to insert. The default value is 18.
6715 @item omega-hash-table-size
6716 The size of the hash table in the Omega solver. The default value is
6719 @item omega-max-keys
6720 The maximal number of keys used by the Omega solver. The default
6723 @item omega-eliminate-redundant-constraints
6724 When set to 1, use expensive methods to eliminate all redundant
6725 constraints. The default value is 0.
6727 @item vect-max-version-checks
6728 The maximum number of runtime checks that can be performed when doing
6729 loop versioning in the vectorizer. See option ftree-vect-loop-version
6730 for more information.
6732 @item max-iterations-to-track
6734 The maximum number of iterations of a loop the brute force algorithm
6735 for analysis of # of iterations of the loop tries to evaluate.
6737 @item hot-bb-count-fraction
6738 Select fraction of the maximal count of repetitions of basic block in program
6739 given basic block needs to have to be considered hot.
6741 @item hot-bb-frequency-fraction
6742 Select fraction of the maximal frequency of executions of basic block in
6743 function given basic block needs to have to be considered hot
6745 @item max-predicted-iterations
6746 The maximum number of loop iterations we predict statically. This is useful
6747 in cases where function contain single loop with known bound and other loop
6748 with unknown. We predict the known number of iterations correctly, while
6749 the unknown number of iterations average to roughly 10. This means that the
6750 loop without bounds would appear artificially cold relative to the other one.
6752 @item tracer-dynamic-coverage
6753 @itemx tracer-dynamic-coverage-feedback
6755 This value is used to limit superblock formation once the given percentage of
6756 executed instructions is covered. This limits unnecessary code size
6759 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6760 feedback is available. The real profiles (as opposed to statically estimated
6761 ones) are much less balanced allowing the threshold to be larger value.
6763 @item tracer-max-code-growth
6764 Stop tail duplication once code growth has reached given percentage. This is
6765 rather hokey argument, as most of the duplicates will be eliminated later in
6766 cross jumping, so it may be set to much higher values than is the desired code
6769 @item tracer-min-branch-ratio
6771 Stop reverse growth when the reverse probability of best edge is less than this
6772 threshold (in percent).
6774 @item tracer-min-branch-ratio
6775 @itemx tracer-min-branch-ratio-feedback
6777 Stop forward growth if the best edge do have probability lower than this
6780 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6781 compilation for profile feedback and one for compilation without. The value
6782 for compilation with profile feedback needs to be more conservative (higher) in
6783 order to make tracer effective.
6785 @item max-cse-path-length
6787 Maximum number of basic blocks on path that cse considers. The default is 10.
6790 The maximum instructions CSE process before flushing. The default is 1000.
6792 @item max-aliased-vops
6794 Maximum number of virtual operands per function allowed to represent
6795 aliases before triggering the alias partitioning heuristic. Alias
6796 partitioning reduces compile times and memory consumption needed for
6797 aliasing at the expense of precision loss in alias information. The
6798 default value for this parameter is 100 for -O1, 500 for -O2 and 1000
6801 Notice that if a function contains more memory statements than the
6802 value of this parameter, it is not really possible to achieve this
6803 reduction. In this case, the compiler will use the number of memory
6804 statements as the value for @option{max-aliased-vops}.
6806 @item avg-aliased-vops
6808 Average number of virtual operands per statement allowed to represent
6809 aliases before triggering the alias partitioning heuristic. This
6810 works in conjunction with @option{max-aliased-vops}. If a function
6811 contains more than @option{max-aliased-vops} virtual operators, then
6812 memory symbols will be grouped into memory partitions until either the
6813 total number of virtual operators is below @option{max-aliased-vops}
6814 or the average number of virtual operators per memory statement is
6815 below @option{avg-aliased-vops}. The default value for this parameter
6816 is 1 for -O1 and -O2, and 3 for -O3.
6818 @item ggc-min-expand
6820 GCC uses a garbage collector to manage its own memory allocation. This
6821 parameter specifies the minimum percentage by which the garbage
6822 collector's heap should be allowed to expand between collections.
6823 Tuning this may improve compilation speed; it has no effect on code
6826 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6827 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6828 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6829 GCC is not able to calculate RAM on a particular platform, the lower
6830 bound of 30% is used. Setting this parameter and
6831 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6832 every opportunity. This is extremely slow, but can be useful for
6835 @item ggc-min-heapsize
6837 Minimum size of the garbage collector's heap before it begins bothering
6838 to collect garbage. The first collection occurs after the heap expands
6839 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6840 tuning this may improve compilation speed, and has no effect on code
6843 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6844 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6845 with a lower bound of 4096 (four megabytes) and an upper bound of
6846 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6847 particular platform, the lower bound is used. Setting this parameter
6848 very large effectively disables garbage collection. Setting this
6849 parameter and @option{ggc-min-expand} to zero causes a full collection
6850 to occur at every opportunity.
6852 @item max-reload-search-insns
6853 The maximum number of instruction reload should look backward for equivalent
6854 register. Increasing values mean more aggressive optimization, making the
6855 compile time increase with probably slightly better performance. The default
6858 @item max-cselib-memory-locations
6859 The maximum number of memory locations cselib should take into account.
6860 Increasing values mean more aggressive optimization, making the compile time
6861 increase with probably slightly better performance. The default value is 500.
6863 @item max-flow-memory-locations
6864 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6865 The default value is 100.
6867 @item reorder-blocks-duplicate
6868 @itemx reorder-blocks-duplicate-feedback
6870 Used by basic block reordering pass to decide whether to use unconditional
6871 branch or duplicate the code on its destination. Code is duplicated when its
6872 estimated size is smaller than this value multiplied by the estimated size of
6873 unconditional jump in the hot spots of the program.
6875 The @option{reorder-block-duplicate-feedback} is used only when profile
6876 feedback is available and may be set to higher values than
6877 @option{reorder-block-duplicate} since information about the hot spots is more
6880 @item max-sched-ready-insns
6881 The maximum number of instructions ready to be issued the scheduler should
6882 consider at any given time during the first scheduling pass. Increasing
6883 values mean more thorough searches, making the compilation time increase
6884 with probably little benefit. The default value is 100.
6886 @item max-sched-region-blocks
6887 The maximum number of blocks in a region to be considered for
6888 interblock scheduling. The default value is 10.
6890 @item max-sched-region-insns
6891 The maximum number of insns in a region to be considered for
6892 interblock scheduling. The default value is 100.
6895 The minimum probability (in percents) of reaching a source block
6896 for interblock speculative scheduling. The default value is 40.
6898 @item max-sched-extend-regions-iters
6899 The maximum number of iterations through CFG to extend regions.
6900 0 - disable region extension,
6901 N - do at most N iterations.
6902 The default value is 0.
6904 @item max-sched-insn-conflict-delay
6905 The maximum conflict delay for an insn to be considered for speculative motion.
6906 The default value is 3.
6908 @item sched-spec-prob-cutoff
6909 The minimal probability of speculation success (in percents), so that
6910 speculative insn will be scheduled.
6911 The default value is 40.
6913 @item max-last-value-rtl
6915 The maximum size measured as number of RTLs that can be recorded in an expression
6916 in combiner for a pseudo register as last known value of that register. The default
6919 @item integer-share-limit
6920 Small integer constants can use a shared data structure, reducing the
6921 compiler's memory usage and increasing its speed. This sets the maximum
6922 value of a shared integer constant's. The default value is 256.
6924 @item min-virtual-mappings
6925 Specifies the minimum number of virtual mappings in the incremental
6926 SSA updater that should be registered to trigger the virtual mappings
6927 heuristic defined by virtual-mappings-ratio. The default value is
6930 @item virtual-mappings-ratio
6931 If the number of virtual mappings is virtual-mappings-ratio bigger
6932 than the number of virtual symbols to be updated, then the incremental
6933 SSA updater switches to a full update for those symbols. The default
6936 @item ssp-buffer-size
6937 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6938 protection when @option{-fstack-protection} is used.
6940 @item max-jump-thread-duplication-stmts
6941 Maximum number of statements allowed in a block that needs to be
6942 duplicated when threading jumps.
6944 @item max-fields-for-field-sensitive
6945 Maximum number of fields in a structure we will treat in
6946 a field sensitive manner during pointer analysis.
6948 @item prefetch-latency
6949 Estimate on average number of instructions that are executed before
6950 prefetch finishes. The distance we prefetch ahead is proportional
6951 to this constant. Increasing this number may also lead to less
6952 streams being prefetched (see @option{simultaneous-prefetches}).
6954 @item simultaneous-prefetches
6955 Maximum number of prefetches that can run at the same time.
6957 @item l1-cache-line-size
6958 The size of cache line in L1 cache, in bytes.
6961 The size of L1 cache, in kilobytes.
6964 The size of L2 cache, in kilobytes.
6966 @item use-canonical-types
6967 Whether the compiler should use the ``canonical'' type system. By
6968 default, this should always be 1, which uses a more efficient internal
6969 mechanism for comparing types in C++ and Objective-C++. However, if
6970 bugs in the canonical type system are causing compilation failures,
6971 set this value to 0 to disable canonical types.
6976 @node Preprocessor Options
6977 @section Options Controlling the Preprocessor
6978 @cindex preprocessor options
6979 @cindex options, preprocessor
6981 These options control the C preprocessor, which is run on each C source
6982 file before actual compilation.
6984 If you use the @option{-E} option, nothing is done except preprocessing.
6985 Some of these options make sense only together with @option{-E} because
6986 they cause the preprocessor output to be unsuitable for actual
6991 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6992 and pass @var{option} directly through to the preprocessor. If
6993 @var{option} contains commas, it is split into multiple options at the
6994 commas. However, many options are modified, translated or interpreted
6995 by the compiler driver before being passed to the preprocessor, and
6996 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6997 interface is undocumented and subject to change, so whenever possible
6998 you should avoid using @option{-Wp} and let the driver handle the
7001 @item -Xpreprocessor @var{option}
7002 @opindex preprocessor
7003 Pass @var{option} as an option to the preprocessor. You can use this to
7004 supply system-specific preprocessor options which GCC does not know how to
7007 If you want to pass an option that takes an argument, you must use
7008 @option{-Xpreprocessor} twice, once for the option and once for the argument.
7011 @include cppopts.texi
7013 @node Assembler Options
7014 @section Passing Options to the Assembler
7016 @c prevent bad page break with this line
7017 You can pass options to the assembler.
7020 @item -Wa,@var{option}
7022 Pass @var{option} as an option to the assembler. If @var{option}
7023 contains commas, it is split into multiple options at the commas.
7025 @item -Xassembler @var{option}
7027 Pass @var{option} as an option to the assembler. You can use this to
7028 supply system-specific assembler options which GCC does not know how to
7031 If you want to pass an option that takes an argument, you must use
7032 @option{-Xassembler} twice, once for the option and once for the argument.
7037 @section Options for Linking
7038 @cindex link options
7039 @cindex options, linking
7041 These options come into play when the compiler links object files into
7042 an executable output file. They are meaningless if the compiler is
7043 not doing a link step.
7047 @item @var{object-file-name}
7048 A file name that does not end in a special recognized suffix is
7049 considered to name an object file or library. (Object files are
7050 distinguished from libraries by the linker according to the file
7051 contents.) If linking is done, these object files are used as input
7060 If any of these options is used, then the linker is not run, and
7061 object file names should not be used as arguments. @xref{Overall
7065 @item -l@var{library}
7066 @itemx -l @var{library}
7068 Search the library named @var{library} when linking. (The second
7069 alternative with the library as a separate argument is only for
7070 POSIX compliance and is not recommended.)
7072 It makes a difference where in the command you write this option; the
7073 linker searches and processes libraries and object files in the order they
7074 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
7075 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
7076 to functions in @samp{z}, those functions may not be loaded.
7078 The linker searches a standard list of directories for the library,
7079 which is actually a file named @file{lib@var{library}.a}. The linker
7080 then uses this file as if it had been specified precisely by name.
7082 The directories searched include several standard system directories
7083 plus any that you specify with @option{-L}.
7085 Normally the files found this way are library files---archive files
7086 whose members are object files. The linker handles an archive file by
7087 scanning through it for members which define symbols that have so far
7088 been referenced but not defined. But if the file that is found is an
7089 ordinary object file, it is linked in the usual fashion. The only
7090 difference between using an @option{-l} option and specifying a file name
7091 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
7092 and searches several directories.
7096 You need this special case of the @option{-l} option in order to
7097 link an Objective-C or Objective-C++ program.
7100 @opindex nostartfiles
7101 Do not use the standard system startup files when linking.
7102 The standard system libraries are used normally, unless @option{-nostdlib}
7103 or @option{-nodefaultlibs} is used.
7105 @item -nodefaultlibs
7106 @opindex nodefaultlibs
7107 Do not use the standard system libraries when linking.
7108 Only the libraries you specify will be passed to the linker.
7109 The standard startup files are used normally, unless @option{-nostartfiles}
7110 is used. The compiler may generate calls to @code{memcmp},
7111 @code{memset}, @code{memcpy} and @code{memmove}.
7112 These entries are usually resolved by entries in
7113 libc. These entry points should be supplied through some other
7114 mechanism when this option is specified.
7118 Do not use the standard system startup files or libraries when linking.
7119 No startup files and only the libraries you specify will be passed to
7120 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
7121 @code{memcpy} and @code{memmove}.
7122 These entries are usually resolved by entries in
7123 libc. These entry points should be supplied through some other
7124 mechanism when this option is specified.
7126 @cindex @option{-lgcc}, use with @option{-nostdlib}
7127 @cindex @option{-nostdlib} and unresolved references
7128 @cindex unresolved references and @option{-nostdlib}
7129 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
7130 @cindex @option{-nodefaultlibs} and unresolved references
7131 @cindex unresolved references and @option{-nodefaultlibs}
7132 One of the standard libraries bypassed by @option{-nostdlib} and
7133 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
7134 that GCC uses to overcome shortcomings of particular machines, or special
7135 needs for some languages.
7136 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
7137 Collection (GCC) Internals},
7138 for more discussion of @file{libgcc.a}.)
7139 In most cases, you need @file{libgcc.a} even when you want to avoid
7140 other standard libraries. In other words, when you specify @option{-nostdlib}
7141 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
7142 This ensures that you have no unresolved references to internal GCC
7143 library subroutines. (For example, @samp{__main}, used to ensure C++
7144 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
7145 GNU Compiler Collection (GCC) Internals}.)
7149 Produce a position independent executable on targets which support it.
7150 For predictable results, you must also specify the same set of options
7151 that were used to generate code (@option{-fpie}, @option{-fPIE},
7152 or model suboptions) when you specify this option.
7156 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
7157 that support it. This instructs the linker to add all symbols, not
7158 only used ones, to the dynamic symbol table. This option is needed
7159 for some uses of @code{dlopen} or to allow obtaining backtraces
7160 from within a program.
7164 Remove all symbol table and relocation information from the executable.
7168 On systems that support dynamic linking, this prevents linking with the shared
7169 libraries. On other systems, this option has no effect.
7173 Produce a shared object which can then be linked with other objects to
7174 form an executable. Not all systems support this option. For predictable
7175 results, you must also specify the same set of options that were used to
7176 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
7177 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
7178 needs to build supplementary stub code for constructors to work. On
7179 multi-libbed systems, @samp{gcc -shared} must select the correct support
7180 libraries to link against. Failing to supply the correct flags may lead
7181 to subtle defects. Supplying them in cases where they are not necessary
7184 @item -shared-libgcc
7185 @itemx -static-libgcc
7186 @opindex shared-libgcc
7187 @opindex static-libgcc
7188 On systems that provide @file{libgcc} as a shared library, these options
7189 force the use of either the shared or static version respectively.
7190 If no shared version of @file{libgcc} was built when the compiler was
7191 configured, these options have no effect.
7193 There are several situations in which an application should use the
7194 shared @file{libgcc} instead of the static version. The most common
7195 of these is when the application wishes to throw and catch exceptions
7196 across different shared libraries. In that case, each of the libraries
7197 as well as the application itself should use the shared @file{libgcc}.
7199 Therefore, the G++ and GCJ drivers automatically add
7200 @option{-shared-libgcc} whenever you build a shared library or a main
7201 executable, because C++ and Java programs typically use exceptions, so
7202 this is the right thing to do.
7204 If, instead, you use the GCC driver to create shared libraries, you may
7205 find that they will not always be linked with the shared @file{libgcc}.
7206 If GCC finds, at its configuration time, that you have a non-GNU linker
7207 or a GNU linker that does not support option @option{--eh-frame-hdr},
7208 it will link the shared version of @file{libgcc} into shared libraries
7209 by default. Otherwise, it will take advantage of the linker and optimize
7210 away the linking with the shared version of @file{libgcc}, linking with
7211 the static version of libgcc by default. This allows exceptions to
7212 propagate through such shared libraries, without incurring relocation
7213 costs at library load time.
7215 However, if a library or main executable is supposed to throw or catch
7216 exceptions, you must link it using the G++ or GCJ driver, as appropriate
7217 for the languages used in the program, or using the option
7218 @option{-shared-libgcc}, such that it is linked with the shared
7223 Bind references to global symbols when building a shared object. Warn
7224 about any unresolved references (unless overridden by the link editor
7225 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
7228 @item -Xlinker @var{option}
7230 Pass @var{option} as an option to the linker. You can use this to
7231 supply system-specific linker options which GCC does not know how to
7234 If you want to pass an option that takes an argument, you must use
7235 @option{-Xlinker} twice, once for the option and once for the argument.
7236 For example, to pass @option{-assert definitions}, you must write
7237 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
7238 @option{-Xlinker "-assert definitions"}, because this passes the entire
7239 string as a single argument, which is not what the linker expects.
7241 @item -Wl,@var{option}
7243 Pass @var{option} as an option to the linker. If @var{option} contains
7244 commas, it is split into multiple options at the commas.
7246 @item -u @var{symbol}
7248 Pretend the symbol @var{symbol} is undefined, to force linking of
7249 library modules to define it. You can use @option{-u} multiple times with
7250 different symbols to force loading of additional library modules.
7253 @node Directory Options
7254 @section Options for Directory Search
7255 @cindex directory options
7256 @cindex options, directory search
7259 These options specify directories to search for header files, for
7260 libraries and for parts of the compiler:
7265 Add the directory @var{dir} to the head of the list of directories to be
7266 searched for header files. This can be used to override a system header
7267 file, substituting your own version, since these directories are
7268 searched before the system header file directories. However, you should
7269 not use this option to add directories that contain vendor-supplied
7270 system header files (use @option{-isystem} for that). If you use more than
7271 one @option{-I} option, the directories are scanned in left-to-right
7272 order; the standard system directories come after.
7274 If a standard system include directory, or a directory specified with
7275 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
7276 option will be ignored. The directory will still be searched but as a
7277 system directory at its normal position in the system include chain.
7278 This is to ensure that GCC's procedure to fix buggy system headers and
7279 the ordering for the include_next directive are not inadvertently changed.
7280 If you really need to change the search order for system directories,
7281 use the @option{-nostdinc} and/or @option{-isystem} options.
7283 @item -iquote@var{dir}
7285 Add the directory @var{dir} to the head of the list of directories to
7286 be searched for header files only for the case of @samp{#include
7287 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
7288 otherwise just like @option{-I}.
7292 Add directory @var{dir} to the list of directories to be searched
7295 @item -B@var{prefix}
7297 This option specifies where to find the executables, libraries,
7298 include files, and data files of the compiler itself.
7300 The compiler driver program runs one or more of the subprograms
7301 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
7302 @var{prefix} as a prefix for each program it tries to run, both with and
7303 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
7305 For each subprogram to be run, the compiler driver first tries the
7306 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
7307 was not specified, the driver tries two standard prefixes, which are
7308 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
7309 those results in a file name that is found, the unmodified program
7310 name is searched for using the directories specified in your
7311 @env{PATH} environment variable.
7313 The compiler will check to see if the path provided by the @option{-B}
7314 refers to a directory, and if necessary it will add a directory
7315 separator character at the end of the path.
7317 @option{-B} prefixes that effectively specify directory names also apply
7318 to libraries in the linker, because the compiler translates these
7319 options into @option{-L} options for the linker. They also apply to
7320 includes files in the preprocessor, because the compiler translates these
7321 options into @option{-isystem} options for the preprocessor. In this case,
7322 the compiler appends @samp{include} to the prefix.
7324 The run-time support file @file{libgcc.a} can also be searched for using
7325 the @option{-B} prefix, if needed. If it is not found there, the two
7326 standard prefixes above are tried, and that is all. The file is left
7327 out of the link if it is not found by those means.
7329 Another way to specify a prefix much like the @option{-B} prefix is to use
7330 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
7333 As a special kludge, if the path provided by @option{-B} is
7334 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
7335 9, then it will be replaced by @file{[dir/]include}. This is to help
7336 with boot-strapping the compiler.
7338 @item -specs=@var{file}
7340 Process @var{file} after the compiler reads in the standard @file{specs}
7341 file, in order to override the defaults that the @file{gcc} driver
7342 program uses when determining what switches to pass to @file{cc1},
7343 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
7344 @option{-specs=@var{file}} can be specified on the command line, and they
7345 are processed in order, from left to right.
7347 @item --sysroot=@var{dir}
7349 Use @var{dir} as the logical root directory for headers and libraries.
7350 For example, if the compiler would normally search for headers in
7351 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
7352 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
7354 If you use both this option and the @option{-isysroot} option, then
7355 the @option{--sysroot} option will apply to libraries, but the
7356 @option{-isysroot} option will apply to header files.
7358 The GNU linker (beginning with version 2.16) has the necessary support
7359 for this option. If your linker does not support this option, the
7360 header file aspect of @option{--sysroot} will still work, but the
7361 library aspect will not.
7365 This option has been deprecated. Please use @option{-iquote} instead for
7366 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
7367 Any directories you specify with @option{-I} options before the @option{-I-}
7368 option are searched only for the case of @samp{#include "@var{file}"};
7369 they are not searched for @samp{#include <@var{file}>}.
7371 If additional directories are specified with @option{-I} options after
7372 the @option{-I-}, these directories are searched for all @samp{#include}
7373 directives. (Ordinarily @emph{all} @option{-I} directories are used
7376 In addition, the @option{-I-} option inhibits the use of the current
7377 directory (where the current input file came from) as the first search
7378 directory for @samp{#include "@var{file}"}. There is no way to
7379 override this effect of @option{-I-}. With @option{-I.} you can specify
7380 searching the directory which was current when the compiler was
7381 invoked. That is not exactly the same as what the preprocessor does
7382 by default, but it is often satisfactory.
7384 @option{-I-} does not inhibit the use of the standard system directories
7385 for header files. Thus, @option{-I-} and @option{-nostdinc} are
7392 @section Specifying subprocesses and the switches to pass to them
7395 @command{gcc} is a driver program. It performs its job by invoking a
7396 sequence of other programs to do the work of compiling, assembling and
7397 linking. GCC interprets its command-line parameters and uses these to
7398 deduce which programs it should invoke, and which command-line options
7399 it ought to place on their command lines. This behavior is controlled
7400 by @dfn{spec strings}. In most cases there is one spec string for each
7401 program that GCC can invoke, but a few programs have multiple spec
7402 strings to control their behavior. The spec strings built into GCC can
7403 be overridden by using the @option{-specs=} command-line switch to specify
7406 @dfn{Spec files} are plaintext files that are used to construct spec
7407 strings. They consist of a sequence of directives separated by blank
7408 lines. The type of directive is determined by the first non-whitespace
7409 character on the line and it can be one of the following:
7412 @item %@var{command}
7413 Issues a @var{command} to the spec file processor. The commands that can
7417 @item %include <@var{file}>
7419 Search for @var{file} and insert its text at the current point in the
7422 @item %include_noerr <@var{file}>
7423 @cindex %include_noerr
7424 Just like @samp{%include}, but do not generate an error message if the include
7425 file cannot be found.
7427 @item %rename @var{old_name} @var{new_name}
7429 Rename the spec string @var{old_name} to @var{new_name}.
7433 @item *[@var{spec_name}]:
7434 This tells the compiler to create, override or delete the named spec
7435 string. All lines after this directive up to the next directive or
7436 blank line are considered to be the text for the spec string. If this
7437 results in an empty string then the spec will be deleted. (Or, if the
7438 spec did not exist, then nothing will happened.) Otherwise, if the spec
7439 does not currently exist a new spec will be created. If the spec does
7440 exist then its contents will be overridden by the text of this
7441 directive, unless the first character of that text is the @samp{+}
7442 character, in which case the text will be appended to the spec.
7444 @item [@var{suffix}]:
7445 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
7446 and up to the next directive or blank line are considered to make up the
7447 spec string for the indicated suffix. When the compiler encounters an
7448 input file with the named suffix, it will processes the spec string in
7449 order to work out how to compile that file. For example:
7456 This says that any input file whose name ends in @samp{.ZZ} should be
7457 passed to the program @samp{z-compile}, which should be invoked with the
7458 command-line switch @option{-input} and with the result of performing the
7459 @samp{%i} substitution. (See below.)
7461 As an alternative to providing a spec string, the text that follows a
7462 suffix directive can be one of the following:
7465 @item @@@var{language}
7466 This says that the suffix is an alias for a known @var{language}. This is
7467 similar to using the @option{-x} command-line switch to GCC to specify a
7468 language explicitly. For example:
7475 Says that .ZZ files are, in fact, C++ source files.
7478 This causes an error messages saying:
7481 @var{name} compiler not installed on this system.
7485 GCC already has an extensive list of suffixes built into it.
7486 This directive will add an entry to the end of the list of suffixes, but
7487 since the list is searched from the end backwards, it is effectively
7488 possible to override earlier entries using this technique.
7492 GCC has the following spec strings built into it. Spec files can
7493 override these strings or create their own. Note that individual
7494 targets can also add their own spec strings to this list.
7497 asm Options to pass to the assembler
7498 asm_final Options to pass to the assembler post-processor
7499 cpp Options to pass to the C preprocessor
7500 cc1 Options to pass to the C compiler
7501 cc1plus Options to pass to the C++ compiler
7502 endfile Object files to include at the end of the link
7503 link Options to pass to the linker
7504 lib Libraries to include on the command line to the linker
7505 libgcc Decides which GCC support library to pass to the linker
7506 linker Sets the name of the linker
7507 predefines Defines to be passed to the C preprocessor
7508 signed_char Defines to pass to CPP to say whether @code{char} is signed
7510 startfile Object files to include at the start of the link
7513 Here is a small example of a spec file:
7519 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
7522 This example renames the spec called @samp{lib} to @samp{old_lib} and
7523 then overrides the previous definition of @samp{lib} with a new one.
7524 The new definition adds in some extra command-line options before
7525 including the text of the old definition.
7527 @dfn{Spec strings} are a list of command-line options to be passed to their
7528 corresponding program. In addition, the spec strings can contain
7529 @samp{%}-prefixed sequences to substitute variable text or to
7530 conditionally insert text into the command line. Using these constructs
7531 it is possible to generate quite complex command lines.
7533 Here is a table of all defined @samp{%}-sequences for spec
7534 strings. Note that spaces are not generated automatically around the
7535 results of expanding these sequences. Therefore you can concatenate them
7536 together or combine them with constant text in a single argument.
7540 Substitute one @samp{%} into the program name or argument.
7543 Substitute the name of the input file being processed.
7546 Substitute the basename of the input file being processed.
7547 This is the substring up to (and not including) the last period
7548 and not including the directory.
7551 This is the same as @samp{%b}, but include the file suffix (text after
7555 Marks the argument containing or following the @samp{%d} as a
7556 temporary file name, so that that file will be deleted if GCC exits
7557 successfully. Unlike @samp{%g}, this contributes no text to the
7560 @item %g@var{suffix}
7561 Substitute a file name that has suffix @var{suffix} and is chosen
7562 once per compilation, and mark the argument in the same way as
7563 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7564 name is now chosen in a way that is hard to predict even when previously
7565 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7566 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7567 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7568 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7569 was simply substituted with a file name chosen once per compilation,
7570 without regard to any appended suffix (which was therefore treated
7571 just like ordinary text), making such attacks more likely to succeed.
7573 @item %u@var{suffix}
7574 Like @samp{%g}, but generates a new temporary file name even if
7575 @samp{%u@var{suffix}} was already seen.
7577 @item %U@var{suffix}
7578 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7579 new one if there is no such last file name. In the absence of any
7580 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7581 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7582 would involve the generation of two distinct file names, one
7583 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7584 simply substituted with a file name chosen for the previous @samp{%u},
7585 without regard to any appended suffix.
7587 @item %j@var{suffix}
7588 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7589 writable, and if save-temps is off; otherwise, substitute the name
7590 of a temporary file, just like @samp{%u}. This temporary file is not
7591 meant for communication between processes, but rather as a junk
7594 @item %|@var{suffix}
7595 @itemx %m@var{suffix}
7596 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7597 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7598 all. These are the two most common ways to instruct a program that it
7599 should read from standard input or write to standard output. If you
7600 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7601 construct: see for example @file{f/lang-specs.h}.
7603 @item %.@var{SUFFIX}
7604 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7605 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7606 terminated by the next space or %.
7609 Marks the argument containing or following the @samp{%w} as the
7610 designated output file of this compilation. This puts the argument
7611 into the sequence of arguments that @samp{%o} will substitute later.
7614 Substitutes the names of all the output files, with spaces
7615 automatically placed around them. You should write spaces
7616 around the @samp{%o} as well or the results are undefined.
7617 @samp{%o} is for use in the specs for running the linker.
7618 Input files whose names have no recognized suffix are not compiled
7619 at all, but they are included among the output files, so they will
7623 Substitutes the suffix for object files. Note that this is
7624 handled specially when it immediately follows @samp{%g, %u, or %U},
7625 because of the need for those to form complete file names. The
7626 handling is such that @samp{%O} is treated exactly as if it had already
7627 been substituted, except that @samp{%g, %u, and %U} do not currently
7628 support additional @var{suffix} characters following @samp{%O} as they would
7629 following, for example, @samp{.o}.
7632 Substitutes the standard macro predefinitions for the
7633 current target machine. Use this when running @code{cpp}.
7636 Like @samp{%p}, but puts @samp{__} before and after the name of each
7637 predefined macro, except for macros that start with @samp{__} or with
7638 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7642 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7643 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7644 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7645 and @option{-imultilib} as necessary.
7648 Current argument is the name of a library or startup file of some sort.
7649 Search for that file in a standard list of directories and substitute
7650 the full name found.
7653 Print @var{str} as an error message. @var{str} is terminated by a newline.
7654 Use this when inconsistent options are detected.
7657 Substitute the contents of spec string @var{name} at this point.
7660 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7662 @item %x@{@var{option}@}
7663 Accumulate an option for @samp{%X}.
7666 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7670 Output the accumulated assembler options specified by @option{-Wa}.
7673 Output the accumulated preprocessor options specified by @option{-Wp}.
7676 Process the @code{asm} spec. This is used to compute the
7677 switches to be passed to the assembler.
7680 Process the @code{asm_final} spec. This is a spec string for
7681 passing switches to an assembler post-processor, if such a program is
7685 Process the @code{link} spec. This is the spec for computing the
7686 command line passed to the linker. Typically it will make use of the
7687 @samp{%L %G %S %D and %E} sequences.
7690 Dump out a @option{-L} option for each directory that GCC believes might
7691 contain startup files. If the target supports multilibs then the
7692 current multilib directory will be prepended to each of these paths.
7695 Process the @code{lib} spec. This is a spec string for deciding which
7696 libraries should be included on the command line to the linker.
7699 Process the @code{libgcc} spec. This is a spec string for deciding
7700 which GCC support library should be included on the command line to the linker.
7703 Process the @code{startfile} spec. This is a spec for deciding which
7704 object files should be the first ones passed to the linker. Typically
7705 this might be a file named @file{crt0.o}.
7708 Process the @code{endfile} spec. This is a spec string that specifies
7709 the last object files that will be passed to the linker.
7712 Process the @code{cpp} spec. This is used to construct the arguments
7713 to be passed to the C preprocessor.
7716 Process the @code{cc1} spec. This is used to construct the options to be
7717 passed to the actual C compiler (@samp{cc1}).
7720 Process the @code{cc1plus} spec. This is used to construct the options to be
7721 passed to the actual C++ compiler (@samp{cc1plus}).
7724 Substitute the variable part of a matched option. See below.
7725 Note that each comma in the substituted string is replaced by
7729 Remove all occurrences of @code{-S} from the command line. Note---this
7730 command is position dependent. @samp{%} commands in the spec string
7731 before this one will see @code{-S}, @samp{%} commands in the spec string
7732 after this one will not.
7734 @item %:@var{function}(@var{args})
7735 Call the named function @var{function}, passing it @var{args}.
7736 @var{args} is first processed as a nested spec string, then split
7737 into an argument vector in the usual fashion. The function returns
7738 a string which is processed as if it had appeared literally as part
7739 of the current spec.
7741 The following built-in spec functions are provided:
7745 The @code{getenv} spec function takes two arguments: an environment
7746 variable name and a string. If the environment variable is not
7747 defined, a fatal error is issued. Otherwise, the return value is the
7748 value of the environment variable concatenated with the string. For
7749 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
7752 %:getenv(TOPDIR /include)
7755 expands to @file{/path/to/top/include}.
7757 @item @code{if-exists}
7758 The @code{if-exists} spec function takes one argument, an absolute
7759 pathname to a file. If the file exists, @code{if-exists} returns the
7760 pathname. Here is a small example of its usage:
7764 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7767 @item @code{if-exists-else}
7768 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7769 spec function, except that it takes two arguments. The first argument is
7770 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7771 returns the pathname. If it does not exist, it returns the second argument.
7772 This way, @code{if-exists-else} can be used to select one file or another,
7773 based on the existence of the first. Here is a small example of its usage:
7777 crt0%O%s %:if-exists(crti%O%s) \
7778 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7781 @item @code{replace-outfile}
7782 The @code{replace-outfile} spec function takes two arguments. It looks for the
7783 first argument in the outfiles array and replaces it with the second argument. Here
7784 is a small example of its usage:
7787 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7790 @item @code{print-asm-header}
7791 The @code{print-asm-header} function takes no arguments and simply
7792 prints a banner like:
7798 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
7801 It is used to separate compiler options from assembler options
7802 in the @option{--target-help} output.
7806 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7807 If that switch was not specified, this substitutes nothing. Note that
7808 the leading dash is omitted when specifying this option, and it is
7809 automatically inserted if the substitution is performed. Thus the spec
7810 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7811 and would output the command line option @option{-foo}.
7813 @item %W@{@code{S}@}
7814 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7817 @item %@{@code{S}*@}
7818 Substitutes all the switches specified to GCC whose names start
7819 with @code{-S}, but which also take an argument. This is used for
7820 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7821 GCC considers @option{-o foo} as being
7822 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7823 text, including the space. Thus two arguments would be generated.
7825 @item %@{@code{S}*&@code{T}*@}
7826 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7827 (the order of @code{S} and @code{T} in the spec is not significant).
7828 There can be any number of ampersand-separated variables; for each the
7829 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7831 @item %@{@code{S}:@code{X}@}
7832 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7834 @item %@{!@code{S}:@code{X}@}
7835 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7837 @item %@{@code{S}*:@code{X}@}
7838 Substitutes @code{X} if one or more switches whose names start with
7839 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7840 once, no matter how many such switches appeared. However, if @code{%*}
7841 appears somewhere in @code{X}, then @code{X} will be substituted once
7842 for each matching switch, with the @code{%*} replaced by the part of
7843 that switch that matched the @code{*}.
7845 @item %@{.@code{S}:@code{X}@}
7846 Substitutes @code{X}, if processing a file with suffix @code{S}.
7848 @item %@{!.@code{S}:@code{X}@}
7849 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7851 @item %@{,@code{S}:@code{X}@}
7852 Substitutes @code{X}, if processing a file for language @code{S}.
7854 @item %@{!,@code{S}:@code{X}@}
7855 Substitutes @code{X}, if not processing a file for language @code{S}.
7857 @item %@{@code{S}|@code{P}:@code{X}@}
7858 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
7859 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
7860 @code{*} sequences as well, although they have a stronger binding than
7861 the @samp{|}. If @code{%*} appears in @code{X}, all of the
7862 alternatives must be starred, and only the first matching alternative
7865 For example, a spec string like this:
7868 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7871 will output the following command-line options from the following input
7872 command-line options:
7877 -d fred.c -foo -baz -boggle
7878 -d jim.d -bar -baz -boggle
7881 @item %@{S:X; T:Y; :D@}
7883 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7884 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7885 be as many clauses as you need. This may be combined with @code{.},
7886 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
7891 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7892 construct may contain other nested @samp{%} constructs or spaces, or
7893 even newlines. They are processed as usual, as described above.
7894 Trailing white space in @code{X} is ignored. White space may also
7895 appear anywhere on the left side of the colon in these constructs,
7896 except between @code{.} or @code{*} and the corresponding word.
7898 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7899 handled specifically in these constructs. If another value of
7900 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7901 @option{-W} switch is found later in the command line, the earlier
7902 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7903 just one letter, which passes all matching options.
7905 The character @samp{|} at the beginning of the predicate text is used to
7906 indicate that a command should be piped to the following command, but
7907 only if @option{-pipe} is specified.
7909 It is built into GCC which switches take arguments and which do not.
7910 (You might think it would be useful to generalize this to allow each
7911 compiler's spec to say which switches take arguments. But this cannot
7912 be done in a consistent fashion. GCC cannot even decide which input
7913 files have been specified without knowing which switches take arguments,
7914 and it must know which input files to compile in order to tell which
7917 GCC also knows implicitly that arguments starting in @option{-l} are to be
7918 treated as compiler output files, and passed to the linker in their
7919 proper position among the other output files.
7921 @c man begin OPTIONS
7923 @node Target Options
7924 @section Specifying Target Machine and Compiler Version
7925 @cindex target options
7926 @cindex cross compiling
7927 @cindex specifying machine version
7928 @cindex specifying compiler version and target machine
7929 @cindex compiler version, specifying
7930 @cindex target machine, specifying
7932 The usual way to run GCC is to run the executable called @file{gcc}, or
7933 @file{<machine>-gcc} when cross-compiling, or
7934 @file{<machine>-gcc-<version>} to run a version other than the one that
7935 was installed last. Sometimes this is inconvenient, so GCC provides
7936 options that will switch to another cross-compiler or version.
7939 @item -b @var{machine}
7941 The argument @var{machine} specifies the target machine for compilation.
7943 The value to use for @var{machine} is the same as was specified as the
7944 machine type when configuring GCC as a cross-compiler. For
7945 example, if a cross-compiler was configured with @samp{configure
7946 arm-elf}, meaning to compile for an arm processor with elf binaries,
7947 then you would specify @option{-b arm-elf} to run that cross compiler.
7948 Because there are other options beginning with @option{-b}, the
7949 configuration must contain a hyphen.
7951 @item -V @var{version}
7953 The argument @var{version} specifies which version of GCC to run.
7954 This is useful when multiple versions are installed. For example,
7955 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7958 The @option{-V} and @option{-b} options work by running the
7959 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7960 use them if you can just run that directly.
7962 @node Submodel Options
7963 @section Hardware Models and Configurations
7964 @cindex submodel options
7965 @cindex specifying hardware config
7966 @cindex hardware models and configurations, specifying
7967 @cindex machine dependent options
7969 Earlier we discussed the standard option @option{-b} which chooses among
7970 different installed compilers for completely different target
7971 machines, such as VAX vs.@: 68000 vs.@: 80386.
7973 In addition, each of these target machine types can have its own
7974 special options, starting with @samp{-m}, to choose among various
7975 hardware models or configurations---for example, 68010 vs 68020,
7976 floating coprocessor or none. A single installed version of the
7977 compiler can compile for any model or configuration, according to the
7980 Some configurations of the compiler also support additional special
7981 options, usually for compatibility with other compilers on the same
7984 @c This list is ordered alphanumerically by subsection name.
7985 @c It should be the same order and spelling as these options are listed
7986 @c in Machine Dependent Options
7992 * Blackfin Options::
7996 * DEC Alpha Options::
7997 * DEC Alpha/VMS Options::
7999 * GNU/Linux Options::
8002 * i386 and x86-64 Options::
8015 * RS/6000 and PowerPC Options::
8016 * S/390 and zSeries Options::
8021 * System V Options::
8022 * TMS320C3x/C4x Options::
8027 * Xstormy16 Options::
8033 @subsection ARC Options
8036 These options are defined for ARC implementations:
8041 Compile code for little endian mode. This is the default.
8045 Compile code for big endian mode.
8048 @opindex mmangle-cpu
8049 Prepend the name of the cpu to all public symbol names.
8050 In multiple-processor systems, there are many ARC variants with different
8051 instruction and register set characteristics. This flag prevents code
8052 compiled for one cpu to be linked with code compiled for another.
8053 No facility exists for handling variants that are ``almost identical''.
8054 This is an all or nothing option.
8056 @item -mcpu=@var{cpu}
8058 Compile code for ARC variant @var{cpu}.
8059 Which variants are supported depend on the configuration.
8060 All variants support @option{-mcpu=base}, this is the default.
8062 @item -mtext=@var{text-section}
8063 @itemx -mdata=@var{data-section}
8064 @itemx -mrodata=@var{readonly-data-section}
8068 Put functions, data, and readonly data in @var{text-section},
8069 @var{data-section}, and @var{readonly-data-section} respectively
8070 by default. This can be overridden with the @code{section} attribute.
8071 @xref{Variable Attributes}.
8076 @subsection ARM Options
8079 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
8083 @item -mabi=@var{name}
8085 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
8086 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
8089 @opindex mapcs-frame
8090 Generate a stack frame that is compliant with the ARM Procedure Call
8091 Standard for all functions, even if this is not strictly necessary for
8092 correct execution of the code. Specifying @option{-fomit-frame-pointer}
8093 with this option will cause the stack frames not to be generated for
8094 leaf functions. The default is @option{-mno-apcs-frame}.
8098 This is a synonym for @option{-mapcs-frame}.
8101 @c not currently implemented
8102 @item -mapcs-stack-check
8103 @opindex mapcs-stack-check
8104 Generate code to check the amount of stack space available upon entry to
8105 every function (that actually uses some stack space). If there is
8106 insufficient space available then either the function
8107 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
8108 called, depending upon the amount of stack space required. The run time
8109 system is required to provide these functions. The default is
8110 @option{-mno-apcs-stack-check}, since this produces smaller code.
8112 @c not currently implemented
8114 @opindex mapcs-float
8115 Pass floating point arguments using the float point registers. This is
8116 one of the variants of the APCS@. This option is recommended if the
8117 target hardware has a floating point unit or if a lot of floating point
8118 arithmetic is going to be performed by the code. The default is
8119 @option{-mno-apcs-float}, since integer only code is slightly increased in
8120 size if @option{-mapcs-float} is used.
8122 @c not currently implemented
8123 @item -mapcs-reentrant
8124 @opindex mapcs-reentrant
8125 Generate reentrant, position independent code. The default is
8126 @option{-mno-apcs-reentrant}.
8129 @item -mthumb-interwork
8130 @opindex mthumb-interwork
8131 Generate code which supports calling between the ARM and Thumb
8132 instruction sets. Without this option the two instruction sets cannot
8133 be reliably used inside one program. The default is
8134 @option{-mno-thumb-interwork}, since slightly larger code is generated
8135 when @option{-mthumb-interwork} is specified.
8137 @item -mno-sched-prolog
8138 @opindex mno-sched-prolog
8139 Prevent the reordering of instructions in the function prolog, or the
8140 merging of those instruction with the instructions in the function's
8141 body. This means that all functions will start with a recognizable set
8142 of instructions (or in fact one of a choice from a small set of
8143 different function prologues), and this information can be used to
8144 locate the start if functions inside an executable piece of code. The
8145 default is @option{-msched-prolog}.
8148 @opindex mhard-float
8149 Generate output containing floating point instructions. This is the
8153 @opindex msoft-float
8154 Generate output containing library calls for floating point.
8155 @strong{Warning:} the requisite libraries are not available for all ARM
8156 targets. Normally the facilities of the machine's usual C compiler are
8157 used, but this cannot be done directly in cross-compilation. You must make
8158 your own arrangements to provide suitable library functions for
8161 @option{-msoft-float} changes the calling convention in the output file;
8162 therefore, it is only useful if you compile @emph{all} of a program with
8163 this option. In particular, you need to compile @file{libgcc.a}, the
8164 library that comes with GCC, with @option{-msoft-float} in order for
8167 @item -mfloat-abi=@var{name}
8169 Specifies which ABI to use for floating point values. Permissible values
8170 are: @samp{soft}, @samp{softfp} and @samp{hard}.
8172 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
8173 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
8174 of floating point instructions, but still uses the soft-float calling
8177 @item -mlittle-endian
8178 @opindex mlittle-endian
8179 Generate code for a processor running in little-endian mode. This is
8180 the default for all standard configurations.
8183 @opindex mbig-endian
8184 Generate code for a processor running in big-endian mode; the default is
8185 to compile code for a little-endian processor.
8187 @item -mwords-little-endian
8188 @opindex mwords-little-endian
8189 This option only applies when generating code for big-endian processors.
8190 Generate code for a little-endian word order but a big-endian byte
8191 order. That is, a byte order of the form @samp{32107654}. Note: this
8192 option should only be used if you require compatibility with code for
8193 big-endian ARM processors generated by versions of the compiler prior to
8196 @item -mcpu=@var{name}
8198 This specifies the name of the target ARM processor. GCC uses this name
8199 to determine what kind of instructions it can emit when generating
8200 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
8201 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
8202 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
8203 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
8204 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
8205 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
8206 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
8207 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
8208 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
8209 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
8210 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
8211 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
8212 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
8213 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
8214 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
8215 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
8217 @itemx -mtune=@var{name}
8219 This option is very similar to the @option{-mcpu=} option, except that
8220 instead of specifying the actual target processor type, and hence
8221 restricting which instructions can be used, it specifies that GCC should
8222 tune the performance of the code as if the target were of the type
8223 specified in this option, but still choosing the instructions that it
8224 will generate based on the cpu specified by a @option{-mcpu=} option.
8225 For some ARM implementations better performance can be obtained by using
8228 @item -march=@var{name}
8230 This specifies the name of the target ARM architecture. GCC uses this
8231 name to determine what kind of instructions it can emit when generating
8232 assembly code. This option can be used in conjunction with or instead
8233 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
8234 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
8235 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
8236 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
8237 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
8239 @item -mfpu=@var{name}
8240 @itemx -mfpe=@var{number}
8241 @itemx -mfp=@var{number}
8245 This specifies what floating point hardware (or hardware emulation) is
8246 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
8247 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
8248 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
8249 with older versions of GCC@.
8251 If @option{-msoft-float} is specified this specifies the format of
8252 floating point values.
8254 @item -mstructure-size-boundary=@var{n}
8255 @opindex mstructure-size-boundary
8256 The size of all structures and unions will be rounded up to a multiple
8257 of the number of bits set by this option. Permissible values are 8, 32
8258 and 64. The default value varies for different toolchains. For the COFF
8259 targeted toolchain the default value is 8. A value of 64 is only allowed
8260 if the underlying ABI supports it.
8262 Specifying the larger number can produce faster, more efficient code, but
8263 can also increase the size of the program. Different values are potentially
8264 incompatible. Code compiled with one value cannot necessarily expect to
8265 work with code or libraries compiled with another value, if they exchange
8266 information using structures or unions.
8268 @item -mabort-on-noreturn
8269 @opindex mabort-on-noreturn
8270 Generate a call to the function @code{abort} at the end of a
8271 @code{noreturn} function. It will be executed if the function tries to
8275 @itemx -mno-long-calls
8276 @opindex mlong-calls
8277 @opindex mno-long-calls
8278 Tells the compiler to perform function calls by first loading the
8279 address of the function into a register and then performing a subroutine
8280 call on this register. This switch is needed if the target function
8281 will lie outside of the 64 megabyte addressing range of the offset based
8282 version of subroutine call instruction.
8284 Even if this switch is enabled, not all function calls will be turned
8285 into long calls. The heuristic is that static functions, functions
8286 which have the @samp{short-call} attribute, functions that are inside
8287 the scope of a @samp{#pragma no_long_calls} directive and functions whose
8288 definitions have already been compiled within the current compilation
8289 unit, will not be turned into long calls. The exception to this rule is
8290 that weak function definitions, functions with the @samp{long-call}
8291 attribute or the @samp{section} attribute, and functions that are within
8292 the scope of a @samp{#pragma long_calls} directive, will always be
8293 turned into long calls.
8295 This feature is not enabled by default. Specifying
8296 @option{-mno-long-calls} will restore the default behavior, as will
8297 placing the function calls within the scope of a @samp{#pragma
8298 long_calls_off} directive. Note these switches have no effect on how
8299 the compiler generates code to handle function calls via function
8302 @item -mnop-fun-dllimport
8303 @opindex mnop-fun-dllimport
8304 Disable support for the @code{dllimport} attribute.
8306 @item -msingle-pic-base
8307 @opindex msingle-pic-base
8308 Treat the register used for PIC addressing as read-only, rather than
8309 loading it in the prologue for each function. The run-time system is
8310 responsible for initializing this register with an appropriate value
8311 before execution begins.
8313 @item -mpic-register=@var{reg}
8314 @opindex mpic-register
8315 Specify the register to be used for PIC addressing. The default is R10
8316 unless stack-checking is enabled, when R9 is used.
8318 @item -mcirrus-fix-invalid-insns
8319 @opindex mcirrus-fix-invalid-insns
8320 @opindex mno-cirrus-fix-invalid-insns
8321 Insert NOPs into the instruction stream to in order to work around
8322 problems with invalid Maverick instruction combinations. This option
8323 is only valid if the @option{-mcpu=ep9312} option has been used to
8324 enable generation of instructions for the Cirrus Maverick floating
8325 point co-processor. This option is not enabled by default, since the
8326 problem is only present in older Maverick implementations. The default
8327 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
8330 @item -mpoke-function-name
8331 @opindex mpoke-function-name
8332 Write the name of each function into the text section, directly
8333 preceding the function prologue. The generated code is similar to this:
8337 .ascii "arm_poke_function_name", 0
8340 .word 0xff000000 + (t1 - t0)
8341 arm_poke_function_name
8343 stmfd sp!, @{fp, ip, lr, pc@}
8347 When performing a stack backtrace, code can inspect the value of
8348 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
8349 location @code{pc - 12} and the top 8 bits are set, then we know that
8350 there is a function name embedded immediately preceding this location
8351 and has length @code{((pc[-3]) & 0xff000000)}.
8355 Generate code for the Thumb instruction set. The default is to
8356 use the 32-bit ARM instruction set.
8357 This option automatically enables either 16-bit Thumb-1 or
8358 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
8359 and @option{-march=@var{name}} options.
8362 @opindex mtpcs-frame
8363 Generate a stack frame that is compliant with the Thumb Procedure Call
8364 Standard for all non-leaf functions. (A leaf function is one that does
8365 not call any other functions.) The default is @option{-mno-tpcs-frame}.
8367 @item -mtpcs-leaf-frame
8368 @opindex mtpcs-leaf-frame
8369 Generate a stack frame that is compliant with the Thumb Procedure Call
8370 Standard for all leaf functions. (A leaf function is one that does
8371 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
8373 @item -mcallee-super-interworking
8374 @opindex mcallee-super-interworking
8375 Gives all externally visible functions in the file being compiled an ARM
8376 instruction set header which switches to Thumb mode before executing the
8377 rest of the function. This allows these functions to be called from
8378 non-interworking code.
8380 @item -mcaller-super-interworking
8381 @opindex mcaller-super-interworking
8382 Allows calls via function pointers (including virtual functions) to
8383 execute correctly regardless of whether the target code has been
8384 compiled for interworking or not. There is a small overhead in the cost
8385 of executing a function pointer if this option is enabled.
8387 @item -mtp=@var{name}
8389 Specify the access model for the thread local storage pointer. The valid
8390 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
8391 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
8392 (supported in the arm6k architecture), and @option{auto}, which uses the
8393 best available method for the selected processor. The default setting is
8399 @subsection AVR Options
8402 These options are defined for AVR implementations:
8405 @item -mmcu=@var{mcu}
8407 Specify ATMEL AVR instruction set or MCU type.
8409 Instruction set avr1 is for the minimal AVR core, not supported by the C
8410 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
8411 attiny11, attiny12, attiny15, attiny28).
8413 Instruction set avr2 (default) is for the classic AVR core with up to
8414 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
8415 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
8416 at90c8534, at90s8535).
8418 Instruction set avr3 is for the classic AVR core with up to 128K program
8419 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
8421 Instruction set avr4 is for the enhanced AVR core with up to 8K program
8422 memory space (MCU types: atmega8, atmega83, atmega85).
8424 Instruction set avr5 is for the enhanced AVR core with up to 128K program
8425 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
8426 atmega64, atmega128, at43usb355, at94k).
8430 Output instruction sizes to the asm file.
8432 @item -minit-stack=@var{N}
8433 @opindex minit-stack
8434 Specify the initial stack address, which may be a symbol or numeric value,
8435 @samp{__stack} is the default.
8437 @item -mno-interrupts
8438 @opindex mno-interrupts
8439 Generated code is not compatible with hardware interrupts.
8440 Code size will be smaller.
8442 @item -mcall-prologues
8443 @opindex mcall-prologues
8444 Functions prologues/epilogues expanded as call to appropriate
8445 subroutines. Code size will be smaller.
8447 @item -mno-tablejump
8448 @opindex mno-tablejump
8449 Do not generate tablejump insns which sometimes increase code size.
8452 @opindex mtiny-stack
8453 Change only the low 8 bits of the stack pointer.
8457 Assume int to be 8 bit integer. This affects the sizes of all types: A
8458 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
8459 and long long will be 4 bytes. Please note that this option does not
8460 comply to the C standards, but it will provide you with smaller code
8464 @node Blackfin Options
8465 @subsection Blackfin Options
8466 @cindex Blackfin Options
8469 @item -mcpu=@var{cpu}
8471 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
8472 can be one of @samp{bf531}, @samp{bf532}, @samp{bf533},
8473 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf561}.
8474 Without this option, @samp{bf532} is used as the processor by default.
8475 The corresponding predefined processor macros for @var{cpu} is to
8476 be defined. For the @samp{bfin-elf} toolchain, this causes the hardware
8477 BSP provided by libgloss to be linked in if @samp{-msim} is not given.
8478 Support for @samp{bf561} is incomplete; only the processor macro is defined.
8482 Specifies that the program will be run on the simulator. This causes
8483 the simulator BSP provided by libgloss to be linked in. This option
8484 has effect only for @samp{bfin-elf} toolchain.
8486 @item -momit-leaf-frame-pointer
8487 @opindex momit-leaf-frame-pointer
8488 Don't keep the frame pointer in a register for leaf functions. This
8489 avoids the instructions to save, set up and restore frame pointers and
8490 makes an extra register available in leaf functions. The option
8491 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8492 which might make debugging harder.
8494 @item -mspecld-anomaly
8495 @opindex mspecld-anomaly
8496 When enabled, the compiler will ensure that the generated code does not
8497 contain speculative loads after jump instructions. This option is enabled
8500 @item -mno-specld-anomaly
8501 @opindex mno-specld-anomaly
8502 Don't generate extra code to prevent speculative loads from occurring.
8504 @item -mcsync-anomaly
8505 @opindex mcsync-anomaly
8506 When enabled, the compiler will ensure that the generated code does not
8507 contain CSYNC or SSYNC instructions too soon after conditional branches.
8508 This option is enabled by default.
8510 @item -mno-csync-anomaly
8511 @opindex mno-csync-anomaly
8512 Don't generate extra code to prevent CSYNC or SSYNC instructions from
8513 occurring too soon after a conditional branch.
8517 When enabled, the compiler is free to take advantage of the knowledge that
8518 the entire program fits into the low 64k of memory.
8521 @opindex mno-low-64k
8522 Assume that the program is arbitrarily large. This is the default.
8524 @item -mstack-check-l1
8525 @opindex mstack-check-l1
8526 Do stack checking using information placed into L1 scratchpad memory by the
8529 @item -mid-shared-library
8530 @opindex mid-shared-library
8531 Generate code that supports shared libraries via the library ID method.
8532 This allows for execute in place and shared libraries in an environment
8533 without virtual memory management. This option implies @option{-fPIC}.
8535 @item -mno-id-shared-library
8536 @opindex mno-id-shared-library
8537 Generate code that doesn't assume ID based shared libraries are being used.
8538 This is the default.
8540 @item -mleaf-id-shared-library
8541 @opindex mleaf-id-shared-library
8542 Generate code that supports shared libraries via the library ID method,
8543 but assumes that this library or executable won't link against any other
8544 ID shared libraries. That allows the compiler to use faster code for jumps
8547 @item -mno-leaf-id-shared-library
8548 @opindex mno-leaf-id-shared-library
8549 Do not assume that the code being compiled won't link against any ID shared
8550 libraries. Slower code will be generated for jump and call insns.
8552 @item -mshared-library-id=n
8553 @opindex mshared-library-id
8554 Specified the identification number of the ID based shared library being
8555 compiled. Specifying a value of 0 will generate more compact code, specifying
8556 other values will force the allocation of that number to the current
8557 library but is no more space or time efficient than omitting this option.
8561 Generate code that allows the data segment to be located in a different
8562 area of memory from the text segment. This allows for execute in place in
8563 an environment without virtual memory management by eliminating relocations
8564 against the text section.
8567 @opindex mno-sep-data
8568 Generate code that assumes that the data segment follows the text segment.
8569 This is the default.
8572 @itemx -mno-long-calls
8573 @opindex mlong-calls
8574 @opindex mno-long-calls
8575 Tells the compiler to perform function calls by first loading the
8576 address of the function into a register and then performing a subroutine
8577 call on this register. This switch is needed if the target function
8578 will lie outside of the 24 bit addressing range of the offset based
8579 version of subroutine call instruction.
8581 This feature is not enabled by default. Specifying
8582 @option{-mno-long-calls} will restore the default behavior. Note these
8583 switches have no effect on how the compiler generates code to handle
8584 function calls via function pointers.
8588 @subsection CRIS Options
8589 @cindex CRIS Options
8591 These options are defined specifically for the CRIS ports.
8594 @item -march=@var{architecture-type}
8595 @itemx -mcpu=@var{architecture-type}
8598 Generate code for the specified architecture. The choices for
8599 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
8600 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
8601 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
8604 @item -mtune=@var{architecture-type}
8606 Tune to @var{architecture-type} everything applicable about the generated
8607 code, except for the ABI and the set of available instructions. The
8608 choices for @var{architecture-type} are the same as for
8609 @option{-march=@var{architecture-type}}.
8611 @item -mmax-stack-frame=@var{n}
8612 @opindex mmax-stack-frame
8613 Warn when the stack frame of a function exceeds @var{n} bytes.
8615 @item -melinux-stacksize=@var{n}
8616 @opindex melinux-stacksize
8617 Only available with the @samp{cris-axis-aout} target. Arranges for
8618 indications in the program to the kernel loader that the stack of the
8619 program should be set to @var{n} bytes.
8625 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8626 @option{-march=v3} and @option{-march=v8} respectively.
8628 @item -mmul-bug-workaround
8629 @itemx -mno-mul-bug-workaround
8630 @opindex mmul-bug-workaround
8631 @opindex mno-mul-bug-workaround
8632 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8633 models where it applies. This option is active by default.
8637 Enable CRIS-specific verbose debug-related information in the assembly
8638 code. This option also has the effect to turn off the @samp{#NO_APP}
8639 formatted-code indicator to the assembler at the beginning of the
8644 Do not use condition-code results from previous instruction; always emit
8645 compare and test instructions before use of condition codes.
8647 @item -mno-side-effects
8648 @opindex mno-side-effects
8649 Do not emit instructions with side-effects in addressing modes other than
8653 @itemx -mno-stack-align
8655 @itemx -mno-data-align
8656 @itemx -mconst-align
8657 @itemx -mno-const-align
8658 @opindex mstack-align
8659 @opindex mno-stack-align
8660 @opindex mdata-align
8661 @opindex mno-data-align
8662 @opindex mconst-align
8663 @opindex mno-const-align
8664 These options (no-options) arranges (eliminate arrangements) for the
8665 stack-frame, individual data and constants to be aligned for the maximum
8666 single data access size for the chosen CPU model. The default is to
8667 arrange for 32-bit alignment. ABI details such as structure layout are
8668 not affected by these options.
8676 Similar to the stack- data- and const-align options above, these options
8677 arrange for stack-frame, writable data and constants to all be 32-bit,
8678 16-bit or 8-bit aligned. The default is 32-bit alignment.
8680 @item -mno-prologue-epilogue
8681 @itemx -mprologue-epilogue
8682 @opindex mno-prologue-epilogue
8683 @opindex mprologue-epilogue
8684 With @option{-mno-prologue-epilogue}, the normal function prologue and
8685 epilogue that sets up the stack-frame are omitted and no return
8686 instructions or return sequences are generated in the code. Use this
8687 option only together with visual inspection of the compiled code: no
8688 warnings or errors are generated when call-saved registers must be saved,
8689 or storage for local variable needs to be allocated.
8695 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8696 instruction sequences that load addresses for functions from the PLT part
8697 of the GOT rather than (traditional on other architectures) calls to the
8698 PLT@. The default is @option{-mgotplt}.
8702 Legacy no-op option only recognized with the cris-axis-aout target.
8706 Legacy no-op option only recognized with the cris-axis-elf and
8707 cris-axis-linux-gnu targets.
8711 Only recognized with the cris-axis-aout target, where it selects a
8712 GNU/linux-like multilib, include files and instruction set for
8717 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8721 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8722 to link with input-output functions from a simulator library. Code,
8723 initialized data and zero-initialized data are allocated consecutively.
8727 Like @option{-sim}, but pass linker options to locate initialized data at
8728 0x40000000 and zero-initialized data at 0x80000000.
8732 @subsection CRX Options
8735 These options are defined specifically for the CRX ports.
8741 Enable the use of multiply-accumulate instructions. Disabled by default.
8745 Push instructions will be used to pass outgoing arguments when functions
8746 are called. Enabled by default.
8749 @node Darwin Options
8750 @subsection Darwin Options
8751 @cindex Darwin options
8753 These options are defined for all architectures running the Darwin operating
8756 FSF GCC on Darwin does not create ``fat'' object files; it will create
8757 an object file for the single architecture that it was built to
8758 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8759 @option{-arch} options are used; it does so by running the compiler or
8760 linker multiple times and joining the results together with
8763 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8764 @samp{i686}) is determined by the flags that specify the ISA
8765 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8766 @option{-force_cpusubtype_ALL} option can be used to override this.
8768 The Darwin tools vary in their behavior when presented with an ISA
8769 mismatch. The assembler, @file{as}, will only permit instructions to
8770 be used that are valid for the subtype of the file it is generating,
8771 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8772 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8773 and print an error if asked to create a shared library with a less
8774 restrictive subtype than its input files (for instance, trying to put
8775 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8776 for executables, @file{ld}, will quietly give the executable the most
8777 restrictive subtype of any of its input files.
8782 Add the framework directory @var{dir} to the head of the list of
8783 directories to be searched for header files. These directories are
8784 interleaved with those specified by @option{-I} options and are
8785 scanned in a left-to-right order.
8787 A framework directory is a directory with frameworks in it. A
8788 framework is a directory with a @samp{"Headers"} and/or
8789 @samp{"PrivateHeaders"} directory contained directly in it that ends
8790 in @samp{".framework"}. The name of a framework is the name of this
8791 directory excluding the @samp{".framework"}. Headers associated with
8792 the framework are found in one of those two directories, with
8793 @samp{"Headers"} being searched first. A subframework is a framework
8794 directory that is in a framework's @samp{"Frameworks"} directory.
8795 Includes of subframework headers can only appear in a header of a
8796 framework that contains the subframework, or in a sibling subframework
8797 header. Two subframeworks are siblings if they occur in the same
8798 framework. A subframework should not have the same name as a
8799 framework, a warning will be issued if this is violated. Currently a
8800 subframework cannot have subframeworks, in the future, the mechanism
8801 may be extended to support this. The standard frameworks can be found
8802 in @samp{"/System/Library/Frameworks"} and
8803 @samp{"/Library/Frameworks"}. An example include looks like
8804 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8805 the name of the framework and header.h is found in the
8806 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8808 @item -iframework@var{dir}
8810 Like @option{-F} except the directory is a treated as a system
8811 directory. The main difference between this @option{-iframework} and
8812 @option{-F} is that with @option{-iframework} the compiler does not
8813 warn about constructs contained within header files found via
8814 @var{dir}. This option is valid only for the C family of languages.
8818 Emit debugging information for symbols that are used. For STABS
8819 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8820 This is by default ON@.
8824 Emit debugging information for all symbols and types.
8826 @item -mmacosx-version-min=@var{version}
8827 The earliest version of MacOS X that this executable will run on
8828 is @var{version}. Typical values of @var{version} include @code{10.1},
8829 @code{10.2}, and @code{10.3.9}.
8831 If the compiler was built to use the system's headers by default,
8832 then the default for this option is the system version on which the
8833 compiler is running, otherwise the default is to make choices which
8834 are compatible with as many systems and code bases as possible.
8838 Enable kernel development mode. The @option{-mkernel} option sets
8839 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8840 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8841 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8842 applicable. This mode also sets @option{-mno-altivec},
8843 @option{-msoft-float}, @option{-fno-builtin} and
8844 @option{-mlong-branch} for PowerPC targets.
8846 @item -mone-byte-bool
8847 @opindex mone-byte-bool
8848 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8849 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8850 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8851 option has no effect on x86.
8853 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8854 to generate code that is not binary compatible with code generated
8855 without that switch. Using this switch may require recompiling all
8856 other modules in a program, including system libraries. Use this
8857 switch to conform to a non-default data model.
8859 @item -mfix-and-continue
8860 @itemx -ffix-and-continue
8861 @itemx -findirect-data
8862 @opindex mfix-and-continue
8863 @opindex ffix-and-continue
8864 @opindex findirect-data
8865 Generate code suitable for fast turn around development. Needed to
8866 enable gdb to dynamically load @code{.o} files into already running
8867 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8868 are provided for backwards compatibility.
8872 Loads all members of static archive libraries.
8873 See man ld(1) for more information.
8875 @item -arch_errors_fatal
8876 @opindex arch_errors_fatal
8877 Cause the errors having to do with files that have the wrong architecture
8881 @opindex bind_at_load
8882 Causes the output file to be marked such that the dynamic linker will
8883 bind all undefined references when the file is loaded or launched.
8887 Produce a Mach-o bundle format file.
8888 See man ld(1) for more information.
8890 @item -bundle_loader @var{executable}
8891 @opindex bundle_loader
8892 This option specifies the @var{executable} that will be loading the build
8893 output file being linked. See man ld(1) for more information.
8897 When passed this option, GCC will produce a dynamic library instead of
8898 an executable when linking, using the Darwin @file{libtool} command.
8900 @item -force_cpusubtype_ALL
8901 @opindex force_cpusubtype_ALL
8902 This causes GCC's output file to have the @var{ALL} subtype, instead of
8903 one controlled by the @option{-mcpu} or @option{-march} option.
8905 @item -allowable_client @var{client_name}
8907 @itemx -compatibility_version
8908 @itemx -current_version
8910 @itemx -dependency-file
8912 @itemx -dylinker_install_name
8914 @itemx -exported_symbols_list
8916 @itemx -flat_namespace
8917 @itemx -force_flat_namespace
8918 @itemx -headerpad_max_install_names
8921 @itemx -install_name
8922 @itemx -keep_private_externs
8923 @itemx -multi_module
8924 @itemx -multiply_defined
8925 @itemx -multiply_defined_unused
8927 @itemx -no_dead_strip_inits_and_terms
8928 @itemx -nofixprebinding
8931 @itemx -noseglinkedit
8932 @itemx -pagezero_size
8934 @itemx -prebind_all_twolevel_modules
8935 @itemx -private_bundle
8936 @itemx -read_only_relocs
8938 @itemx -sectobjectsymbols
8942 @itemx -sectobjectsymbols
8945 @itemx -segs_read_only_addr
8946 @itemx -segs_read_write_addr
8947 @itemx -seg_addr_table
8948 @itemx -seg_addr_table_filename
8951 @itemx -segs_read_only_addr
8952 @itemx -segs_read_write_addr
8953 @itemx -single_module
8956 @itemx -sub_umbrella
8957 @itemx -twolevel_namespace
8960 @itemx -unexported_symbols_list
8961 @itemx -weak_reference_mismatches
8964 @opindex allowable_client
8965 @opindex client_name
8966 @opindex compatibility_version
8967 @opindex current_version
8969 @opindex dependency-file
8971 @opindex dylinker_install_name
8973 @opindex exported_symbols_list
8975 @opindex flat_namespace
8976 @opindex force_flat_namespace
8977 @opindex headerpad_max_install_names
8980 @opindex install_name
8981 @opindex keep_private_externs
8982 @opindex multi_module
8983 @opindex multiply_defined
8984 @opindex multiply_defined_unused
8986 @opindex no_dead_strip_inits_and_terms
8987 @opindex nofixprebinding
8988 @opindex nomultidefs
8990 @opindex noseglinkedit
8991 @opindex pagezero_size
8993 @opindex prebind_all_twolevel_modules
8994 @opindex private_bundle
8995 @opindex read_only_relocs
8997 @opindex sectobjectsymbols
9001 @opindex sectobjectsymbols
9004 @opindex segs_read_only_addr
9005 @opindex segs_read_write_addr
9006 @opindex seg_addr_table
9007 @opindex seg_addr_table_filename
9008 @opindex seglinkedit
9010 @opindex segs_read_only_addr
9011 @opindex segs_read_write_addr
9012 @opindex single_module
9014 @opindex sub_library
9015 @opindex sub_umbrella
9016 @opindex twolevel_namespace
9019 @opindex unexported_symbols_list
9020 @opindex weak_reference_mismatches
9021 @opindex whatsloaded
9023 These options are passed to the Darwin linker. The Darwin linker man page
9024 describes them in detail.
9027 @node DEC Alpha Options
9028 @subsection DEC Alpha Options
9030 These @samp{-m} options are defined for the DEC Alpha implementations:
9033 @item -mno-soft-float
9035 @opindex mno-soft-float
9036 @opindex msoft-float
9037 Use (do not use) the hardware floating-point instructions for
9038 floating-point operations. When @option{-msoft-float} is specified,
9039 functions in @file{libgcc.a} will be used to perform floating-point
9040 operations. Unless they are replaced by routines that emulate the
9041 floating-point operations, or compiled in such a way as to call such
9042 emulations routines, these routines will issue floating-point
9043 operations. If you are compiling for an Alpha without floating-point
9044 operations, you must ensure that the library is built so as not to call
9047 Note that Alpha implementations without floating-point operations are
9048 required to have floating-point registers.
9053 @opindex mno-fp-regs
9054 Generate code that uses (does not use) the floating-point register set.
9055 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
9056 register set is not used, floating point operands are passed in integer
9057 registers as if they were integers and floating-point results are passed
9058 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
9059 so any function with a floating-point argument or return value called by code
9060 compiled with @option{-mno-fp-regs} must also be compiled with that
9063 A typical use of this option is building a kernel that does not use,
9064 and hence need not save and restore, any floating-point registers.
9068 The Alpha architecture implements floating-point hardware optimized for
9069 maximum performance. It is mostly compliant with the IEEE floating
9070 point standard. However, for full compliance, software assistance is
9071 required. This option generates code fully IEEE compliant code
9072 @emph{except} that the @var{inexact-flag} is not maintained (see below).
9073 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
9074 defined during compilation. The resulting code is less efficient but is
9075 able to correctly support denormalized numbers and exceptional IEEE
9076 values such as not-a-number and plus/minus infinity. Other Alpha
9077 compilers call this option @option{-ieee_with_no_inexact}.
9079 @item -mieee-with-inexact
9080 @opindex mieee-with-inexact
9081 This is like @option{-mieee} except the generated code also maintains
9082 the IEEE @var{inexact-flag}. Turning on this option causes the
9083 generated code to implement fully-compliant IEEE math. In addition to
9084 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
9085 macro. On some Alpha implementations the resulting code may execute
9086 significantly slower than the code generated by default. Since there is
9087 very little code that depends on the @var{inexact-flag}, you should
9088 normally not specify this option. Other Alpha compilers call this
9089 option @option{-ieee_with_inexact}.
9091 @item -mfp-trap-mode=@var{trap-mode}
9092 @opindex mfp-trap-mode
9093 This option controls what floating-point related traps are enabled.
9094 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
9095 The trap mode can be set to one of four values:
9099 This is the default (normal) setting. The only traps that are enabled
9100 are the ones that cannot be disabled in software (e.g., division by zero
9104 In addition to the traps enabled by @samp{n}, underflow traps are enabled
9108 Like @samp{u}, but the instructions are marked to be safe for software
9109 completion (see Alpha architecture manual for details).
9112 Like @samp{su}, but inexact traps are enabled as well.
9115 @item -mfp-rounding-mode=@var{rounding-mode}
9116 @opindex mfp-rounding-mode
9117 Selects the IEEE rounding mode. Other Alpha compilers call this option
9118 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
9123 Normal IEEE rounding mode. Floating point numbers are rounded towards
9124 the nearest machine number or towards the even machine number in case
9128 Round towards minus infinity.
9131 Chopped rounding mode. Floating point numbers are rounded towards zero.
9134 Dynamic rounding mode. A field in the floating point control register
9135 (@var{fpcr}, see Alpha architecture reference manual) controls the
9136 rounding mode in effect. The C library initializes this register for
9137 rounding towards plus infinity. Thus, unless your program modifies the
9138 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
9141 @item -mtrap-precision=@var{trap-precision}
9142 @opindex mtrap-precision
9143 In the Alpha architecture, floating point traps are imprecise. This
9144 means without software assistance it is impossible to recover from a
9145 floating trap and program execution normally needs to be terminated.
9146 GCC can generate code that can assist operating system trap handlers
9147 in determining the exact location that caused a floating point trap.
9148 Depending on the requirements of an application, different levels of
9149 precisions can be selected:
9153 Program precision. This option is the default and means a trap handler
9154 can only identify which program caused a floating point exception.
9157 Function precision. The trap handler can determine the function that
9158 caused a floating point exception.
9161 Instruction precision. The trap handler can determine the exact
9162 instruction that caused a floating point exception.
9165 Other Alpha compilers provide the equivalent options called
9166 @option{-scope_safe} and @option{-resumption_safe}.
9168 @item -mieee-conformant
9169 @opindex mieee-conformant
9170 This option marks the generated code as IEEE conformant. You must not
9171 use this option unless you also specify @option{-mtrap-precision=i} and either
9172 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
9173 is to emit the line @samp{.eflag 48} in the function prologue of the
9174 generated assembly file. Under DEC Unix, this has the effect that
9175 IEEE-conformant math library routines will be linked in.
9177 @item -mbuild-constants
9178 @opindex mbuild-constants
9179 Normally GCC examines a 32- or 64-bit integer constant to
9180 see if it can construct it from smaller constants in two or three
9181 instructions. If it cannot, it will output the constant as a literal and
9182 generate code to load it from the data segment at runtime.
9184 Use this option to require GCC to construct @emph{all} integer constants
9185 using code, even if it takes more instructions (the maximum is six).
9187 You would typically use this option to build a shared library dynamic
9188 loader. Itself a shared library, it must relocate itself in memory
9189 before it can find the variables and constants in its own data segment.
9195 Select whether to generate code to be assembled by the vendor-supplied
9196 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
9214 Indicate whether GCC should generate code to use the optional BWX,
9215 CIX, FIX and MAX instruction sets. The default is to use the instruction
9216 sets supported by the CPU type specified via @option{-mcpu=} option or that
9217 of the CPU on which GCC was built if none was specified.
9222 @opindex mfloat-ieee
9223 Generate code that uses (does not use) VAX F and G floating point
9224 arithmetic instead of IEEE single and double precision.
9226 @item -mexplicit-relocs
9227 @itemx -mno-explicit-relocs
9228 @opindex mexplicit-relocs
9229 @opindex mno-explicit-relocs
9230 Older Alpha assemblers provided no way to generate symbol relocations
9231 except via assembler macros. Use of these macros does not allow
9232 optimal instruction scheduling. GNU binutils as of version 2.12
9233 supports a new syntax that allows the compiler to explicitly mark
9234 which relocations should apply to which instructions. This option
9235 is mostly useful for debugging, as GCC detects the capabilities of
9236 the assembler when it is built and sets the default accordingly.
9240 @opindex msmall-data
9241 @opindex mlarge-data
9242 When @option{-mexplicit-relocs} is in effect, static data is
9243 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
9244 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
9245 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
9246 16-bit relocations off of the @code{$gp} register. This limits the
9247 size of the small data area to 64KB, but allows the variables to be
9248 directly accessed via a single instruction.
9250 The default is @option{-mlarge-data}. With this option the data area
9251 is limited to just below 2GB@. Programs that require more than 2GB of
9252 data must use @code{malloc} or @code{mmap} to allocate the data in the
9253 heap instead of in the program's data segment.
9255 When generating code for shared libraries, @option{-fpic} implies
9256 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
9260 @opindex msmall-text
9261 @opindex mlarge-text
9262 When @option{-msmall-text} is used, the compiler assumes that the
9263 code of the entire program (or shared library) fits in 4MB, and is
9264 thus reachable with a branch instruction. When @option{-msmall-data}
9265 is used, the compiler can assume that all local symbols share the
9266 same @code{$gp} value, and thus reduce the number of instructions
9267 required for a function call from 4 to 1.
9269 The default is @option{-mlarge-text}.
9271 @item -mcpu=@var{cpu_type}
9273 Set the instruction set and instruction scheduling parameters for
9274 machine type @var{cpu_type}. You can specify either the @samp{EV}
9275 style name or the corresponding chip number. GCC supports scheduling
9276 parameters for the EV4, EV5 and EV6 family of processors and will
9277 choose the default values for the instruction set from the processor
9278 you specify. If you do not specify a processor type, GCC will default
9279 to the processor on which the compiler was built.
9281 Supported values for @var{cpu_type} are
9287 Schedules as an EV4 and has no instruction set extensions.
9291 Schedules as an EV5 and has no instruction set extensions.
9295 Schedules as an EV5 and supports the BWX extension.
9300 Schedules as an EV5 and supports the BWX and MAX extensions.
9304 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
9308 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
9311 @item -mtune=@var{cpu_type}
9313 Set only the instruction scheduling parameters for machine type
9314 @var{cpu_type}. The instruction set is not changed.
9316 @item -mmemory-latency=@var{time}
9317 @opindex mmemory-latency
9318 Sets the latency the scheduler should assume for typical memory
9319 references as seen by the application. This number is highly
9320 dependent on the memory access patterns used by the application
9321 and the size of the external cache on the machine.
9323 Valid options for @var{time} are
9327 A decimal number representing clock cycles.
9333 The compiler contains estimates of the number of clock cycles for
9334 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
9335 (also called Dcache, Scache, and Bcache), as well as to main memory.
9336 Note that L3 is only valid for EV5.
9341 @node DEC Alpha/VMS Options
9342 @subsection DEC Alpha/VMS Options
9344 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
9347 @item -mvms-return-codes
9348 @opindex mvms-return-codes
9349 Return VMS condition codes from main. The default is to return POSIX
9350 style condition (e.g.@: error) codes.
9354 @subsection FRV Options
9361 Only use the first 32 general purpose registers.
9366 Use all 64 general purpose registers.
9371 Use only the first 32 floating point registers.
9376 Use all 64 floating point registers
9379 @opindex mhard-float
9381 Use hardware instructions for floating point operations.
9384 @opindex msoft-float
9386 Use library routines for floating point operations.
9391 Dynamically allocate condition code registers.
9396 Do not try to dynamically allocate condition code registers, only
9397 use @code{icc0} and @code{fcc0}.
9402 Change ABI to use double word insns.
9407 Do not use double word instructions.
9412 Use floating point double instructions.
9417 Do not use floating point double instructions.
9422 Use media instructions.
9427 Do not use media instructions.
9432 Use multiply and add/subtract instructions.
9437 Do not use multiply and add/subtract instructions.
9442 Select the FDPIC ABI, that uses function descriptors to represent
9443 pointers to functions. Without any PIC/PIE-related options, it
9444 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
9445 assumes GOT entries and small data are within a 12-bit range from the
9446 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
9447 are computed with 32 bits.
9450 @opindex minline-plt
9452 Enable inlining of PLT entries in function calls to functions that are
9453 not known to bind locally. It has no effect without @option{-mfdpic}.
9454 It's enabled by default if optimizing for speed and compiling for
9455 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
9456 optimization option such as @option{-O3} or above is present in the
9462 Assume a large TLS segment when generating thread-local code.
9467 Do not assume a large TLS segment when generating thread-local code.
9472 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
9473 that is known to be in read-only sections. It's enabled by default,
9474 except for @option{-fpic} or @option{-fpie}: even though it may help
9475 make the global offset table smaller, it trades 1 instruction for 4.
9476 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
9477 one of which may be shared by multiple symbols, and it avoids the need
9478 for a GOT entry for the referenced symbol, so it's more likely to be a
9479 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
9481 @item -multilib-library-pic
9482 @opindex multilib-library-pic
9484 Link with the (library, not FD) pic libraries. It's implied by
9485 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
9486 @option{-fpic} without @option{-mfdpic}. You should never have to use
9492 Follow the EABI requirement of always creating a frame pointer whenever
9493 a stack frame is allocated. This option is enabled by default and can
9494 be disabled with @option{-mno-linked-fp}.
9497 @opindex mlong-calls
9499 Use indirect addressing to call functions outside the current
9500 compilation unit. This allows the functions to be placed anywhere
9501 within the 32-bit address space.
9503 @item -malign-labels
9504 @opindex malign-labels
9506 Try to align labels to an 8-byte boundary by inserting nops into the
9507 previous packet. This option only has an effect when VLIW packing
9508 is enabled. It doesn't create new packets; it merely adds nops to
9512 @opindex mlibrary-pic
9514 Generate position-independent EABI code.
9519 Use only the first four media accumulator registers.
9524 Use all eight media accumulator registers.
9529 Pack VLIW instructions.
9534 Do not pack VLIW instructions.
9539 Do not mark ABI switches in e_flags.
9544 Enable the use of conditional-move instructions (default).
9546 This switch is mainly for debugging the compiler and will likely be removed
9547 in a future version.
9549 @item -mno-cond-move
9550 @opindex mno-cond-move
9552 Disable the use of conditional-move instructions.
9554 This switch is mainly for debugging the compiler and will likely be removed
9555 in a future version.
9560 Enable the use of conditional set instructions (default).
9562 This switch is mainly for debugging the compiler and will likely be removed
9563 in a future version.
9568 Disable the use of conditional set instructions.
9570 This switch is mainly for debugging the compiler and will likely be removed
9571 in a future version.
9576 Enable the use of conditional execution (default).
9578 This switch is mainly for debugging the compiler and will likely be removed
9579 in a future version.
9581 @item -mno-cond-exec
9582 @opindex mno-cond-exec
9584 Disable the use of conditional execution.
9586 This switch is mainly for debugging the compiler and will likely be removed
9587 in a future version.
9590 @opindex mvliw-branch
9592 Run a pass to pack branches into VLIW instructions (default).
9594 This switch is mainly for debugging the compiler and will likely be removed
9595 in a future version.
9597 @item -mno-vliw-branch
9598 @opindex mno-vliw-branch
9600 Do not run a pass to pack branches into VLIW instructions.
9602 This switch is mainly for debugging the compiler and will likely be removed
9603 in a future version.
9605 @item -mmulti-cond-exec
9606 @opindex mmulti-cond-exec
9608 Enable optimization of @code{&&} and @code{||} in conditional execution
9611 This switch is mainly for debugging the compiler and will likely be removed
9612 in a future version.
9614 @item -mno-multi-cond-exec
9615 @opindex mno-multi-cond-exec
9617 Disable optimization of @code{&&} and @code{||} in conditional execution.
9619 This switch is mainly for debugging the compiler and will likely be removed
9620 in a future version.
9622 @item -mnested-cond-exec
9623 @opindex mnested-cond-exec
9625 Enable nested conditional execution optimizations (default).
9627 This switch is mainly for debugging the compiler and will likely be removed
9628 in a future version.
9630 @item -mno-nested-cond-exec
9631 @opindex mno-nested-cond-exec
9633 Disable nested conditional execution optimizations.
9635 This switch is mainly for debugging the compiler and will likely be removed
9636 in a future version.
9638 @item -moptimize-membar
9639 @opindex moptimize-membar
9641 This switch removes redundant @code{membar} instructions from the
9642 compiler generated code. It is enabled by default.
9644 @item -mno-optimize-membar
9645 @opindex mno-optimize-membar
9647 This switch disables the automatic removal of redundant @code{membar}
9648 instructions from the generated code.
9650 @item -mtomcat-stats
9651 @opindex mtomcat-stats
9653 Cause gas to print out tomcat statistics.
9655 @item -mcpu=@var{cpu}
9658 Select the processor type for which to generate code. Possible values are
9659 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9660 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9664 @node GNU/Linux Options
9665 @subsection GNU/Linux Options
9667 These @samp{-m} options are defined for GNU/Linux targets:
9672 Use the GNU C library instead of uClibc. This is the default except
9673 on @samp{*-*-linux-*uclibc*} targets.
9677 Use uClibc instead of the GNU C library. This is the default on
9678 @samp{*-*-linux-*uclibc*} targets.
9681 @node H8/300 Options
9682 @subsection H8/300 Options
9684 These @samp{-m} options are defined for the H8/300 implementations:
9689 Shorten some address references at link time, when possible; uses the
9690 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9691 ld, Using ld}, for a fuller description.
9695 Generate code for the H8/300H@.
9699 Generate code for the H8S@.
9703 Generate code for the H8S and H8/300H in the normal mode. This switch
9704 must be used either with @option{-mh} or @option{-ms}.
9708 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9712 Make @code{int} data 32 bits by default.
9716 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9717 The default for the H8/300H and H8S is to align longs and floats on 4
9719 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9720 This option has no effect on the H8/300.
9724 @subsection HPPA Options
9725 @cindex HPPA Options
9727 These @samp{-m} options are defined for the HPPA family of computers:
9730 @item -march=@var{architecture-type}
9732 Generate code for the specified architecture. The choices for
9733 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9734 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9735 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9736 architecture option for your machine. Code compiled for lower numbered
9737 architectures will run on higher numbered architectures, but not the
9741 @itemx -mpa-risc-1-1
9742 @itemx -mpa-risc-2-0
9743 @opindex mpa-risc-1-0
9744 @opindex mpa-risc-1-1
9745 @opindex mpa-risc-2-0
9746 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9749 @opindex mbig-switch
9750 Generate code suitable for big switch tables. Use this option only if
9751 the assembler/linker complain about out of range branches within a switch
9754 @item -mjump-in-delay
9755 @opindex mjump-in-delay
9756 Fill delay slots of function calls with unconditional jump instructions
9757 by modifying the return pointer for the function call to be the target
9758 of the conditional jump.
9760 @item -mdisable-fpregs
9761 @opindex mdisable-fpregs
9762 Prevent floating point registers from being used in any manner. This is
9763 necessary for compiling kernels which perform lazy context switching of
9764 floating point registers. If you use this option and attempt to perform
9765 floating point operations, the compiler will abort.
9767 @item -mdisable-indexing
9768 @opindex mdisable-indexing
9769 Prevent the compiler from using indexing address modes. This avoids some
9770 rather obscure problems when compiling MIG generated code under MACH@.
9772 @item -mno-space-regs
9773 @opindex mno-space-regs
9774 Generate code that assumes the target has no space registers. This allows
9775 GCC to generate faster indirect calls and use unscaled index address modes.
9777 Such code is suitable for level 0 PA systems and kernels.
9779 @item -mfast-indirect-calls
9780 @opindex mfast-indirect-calls
9781 Generate code that assumes calls never cross space boundaries. This
9782 allows GCC to emit code which performs faster indirect calls.
9784 This option will not work in the presence of shared libraries or nested
9787 @item -mfixed-range=@var{register-range}
9788 @opindex mfixed-range
9789 Generate code treating the given register range as fixed registers.
9790 A fixed register is one that the register allocator can not use. This is
9791 useful when compiling kernel code. A register range is specified as
9792 two registers separated by a dash. Multiple register ranges can be
9793 specified separated by a comma.
9795 @item -mlong-load-store
9796 @opindex mlong-load-store
9797 Generate 3-instruction load and store sequences as sometimes required by
9798 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9801 @item -mportable-runtime
9802 @opindex mportable-runtime
9803 Use the portable calling conventions proposed by HP for ELF systems.
9807 Enable the use of assembler directives only GAS understands.
9809 @item -mschedule=@var{cpu-type}
9811 Schedule code according to the constraints for the machine type
9812 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9813 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9814 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9815 proper scheduling option for your machine. The default scheduling is
9819 @opindex mlinker-opt
9820 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9821 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9822 linkers in which they give bogus error messages when linking some programs.
9825 @opindex msoft-float
9826 Generate output containing library calls for floating point.
9827 @strong{Warning:} the requisite libraries are not available for all HPPA
9828 targets. Normally the facilities of the machine's usual C compiler are
9829 used, but this cannot be done directly in cross-compilation. You must make
9830 your own arrangements to provide suitable library functions for
9831 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9832 does provide software floating point support.
9834 @option{-msoft-float} changes the calling convention in the output file;
9835 therefore, it is only useful if you compile @emph{all} of a program with
9836 this option. In particular, you need to compile @file{libgcc.a}, the
9837 library that comes with GCC, with @option{-msoft-float} in order for
9842 Generate the predefine, @code{_SIO}, for server IO@. The default is
9843 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9844 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9845 options are available under HP-UX and HI-UX@.
9849 Use GNU ld specific options. This passes @option{-shared} to ld when
9850 building a shared library. It is the default when GCC is configured,
9851 explicitly or implicitly, with the GNU linker. This option does not
9852 have any affect on which ld is called, it only changes what parameters
9853 are passed to that ld. The ld that is called is determined by the
9854 @option{--with-ld} configure option, GCC's program search path, and
9855 finally by the user's @env{PATH}. The linker used by GCC can be printed
9856 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9857 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9861 Use HP ld specific options. This passes @option{-b} to ld when building
9862 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9863 links. It is the default when GCC is configured, explicitly or
9864 implicitly, with the HP linker. This option does not have any affect on
9865 which ld is called, it only changes what parameters are passed to that
9866 ld. The ld that is called is determined by the @option{--with-ld}
9867 configure option, GCC's program search path, and finally by the user's
9868 @env{PATH}. The linker used by GCC can be printed using @samp{which
9869 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9870 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9873 @opindex mno-long-calls
9874 Generate code that uses long call sequences. This ensures that a call
9875 is always able to reach linker generated stubs. The default is to generate
9876 long calls only when the distance from the call site to the beginning
9877 of the function or translation unit, as the case may be, exceeds a
9878 predefined limit set by the branch type being used. The limits for
9879 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9880 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9883 Distances are measured from the beginning of functions when using the
9884 @option{-ffunction-sections} option, or when using the @option{-mgas}
9885 and @option{-mno-portable-runtime} options together under HP-UX with
9888 It is normally not desirable to use this option as it will degrade
9889 performance. However, it may be useful in large applications,
9890 particularly when partial linking is used to build the application.
9892 The types of long calls used depends on the capabilities of the
9893 assembler and linker, and the type of code being generated. The
9894 impact on systems that support long absolute calls, and long pic
9895 symbol-difference or pc-relative calls should be relatively small.
9896 However, an indirect call is used on 32-bit ELF systems in pic code
9897 and it is quite long.
9899 @item -munix=@var{unix-std}
9901 Generate compiler predefines and select a startfile for the specified
9902 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9903 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9904 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9905 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9906 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9909 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9910 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9911 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9912 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9913 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9914 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9916 It is @emph{important} to note that this option changes the interfaces
9917 for various library routines. It also affects the operational behavior
9918 of the C library. Thus, @emph{extreme} care is needed in using this
9921 Library code that is intended to operate with more than one UNIX
9922 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9923 as appropriate. Most GNU software doesn't provide this capability.
9927 Suppress the generation of link options to search libdld.sl when the
9928 @option{-static} option is specified on HP-UX 10 and later.
9932 The HP-UX implementation of setlocale in libc has a dependency on
9933 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9934 when the @option{-static} option is specified, special link options
9935 are needed to resolve this dependency.
9937 On HP-UX 10 and later, the GCC driver adds the necessary options to
9938 link with libdld.sl when the @option{-static} option is specified.
9939 This causes the resulting binary to be dynamic. On the 64-bit port,
9940 the linkers generate dynamic binaries by default in any case. The
9941 @option{-nolibdld} option can be used to prevent the GCC driver from
9942 adding these link options.
9946 Add support for multithreading with the @dfn{dce thread} library
9947 under HP-UX@. This option sets flags for both the preprocessor and
9951 @node i386 and x86-64 Options
9952 @subsection Intel 386 and AMD x86-64 Options
9953 @cindex i386 Options
9954 @cindex x86-64 Options
9955 @cindex Intel 386 Options
9956 @cindex AMD x86-64 Options
9958 These @samp{-m} options are defined for the i386 and x86-64 family of
9962 @item -mtune=@var{cpu-type}
9964 Tune to @var{cpu-type} everything applicable about the generated code, except
9965 for the ABI and the set of available instructions. The choices for
9969 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9970 If you know the CPU on which your code will run, then you should use
9971 the corresponding @option{-mtune} option instead of
9972 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9973 of your application will have, then you should use this option.
9975 As new processors are deployed in the marketplace, the behavior of this
9976 option will change. Therefore, if you upgrade to a newer version of
9977 GCC, the code generated option will change to reflect the processors
9978 that were most common when that version of GCC was released.
9980 There is no @option{-march=generic} option because @option{-march}
9981 indicates the instruction set the compiler can use, and there is no
9982 generic instruction set applicable to all processors. In contrast,
9983 @option{-mtune} indicates the processor (or, in this case, collection of
9984 processors) for which the code is optimized.
9986 This selects the CPU to tune for at compilation time by determining
9987 the processor type of the compiling machine. Using @option{-mtune=native}
9988 will produce code optimized for the local machine under the constraints
9989 of the selected instruction set. Using @option{-march=native} will
9990 enable all instruction subsets supported by the local machine (hence
9991 the result might not run on different machines).
9993 Original Intel's i386 CPU@.
9995 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9997 Intel Pentium CPU with no MMX support.
9999 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
10001 Intel PentiumPro CPU@.
10003 Same as @code{generic}, but when used as @code{march} option, PentiumPro
10004 instruction set will be used, so the code will run on all i686 family chips.
10006 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
10007 @item pentium3, pentium3m
10008 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
10011 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
10012 support. Used by Centrino notebooks.
10013 @item pentium4, pentium4m
10014 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
10016 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
10019 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
10020 SSE2 and SSE3 instruction set support.
10022 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
10023 instruction set support.
10025 AMD K6 CPU with MMX instruction set support.
10027 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
10028 @item athlon, athlon-tbird
10029 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
10031 @item athlon-4, athlon-xp, athlon-mp
10032 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
10033 instruction set support.
10034 @item k8, opteron, athlon64, athlon-fx
10035 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
10036 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
10037 @item k8-sse3, opteron-sse3, athlon64-sse3
10038 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
10039 @item amdfam10, barcelona
10040 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
10041 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3dNOW!, enhanced 3dNOW!, ABM and 64-bit
10042 instruction set extensions.)
10044 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
10047 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
10048 instruction set support.
10050 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
10051 implemented for this chip.)
10053 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
10054 implemented for this chip.)
10056 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
10059 While picking a specific @var{cpu-type} will schedule things appropriately
10060 for that particular chip, the compiler will not generate any code that
10061 does not run on the i386 without the @option{-march=@var{cpu-type}} option
10064 @item -march=@var{cpu-type}
10066 Generate instructions for the machine type @var{cpu-type}. The choices
10067 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
10068 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
10070 @item -mcpu=@var{cpu-type}
10072 A deprecated synonym for @option{-mtune}.
10074 @item -mfpmath=@var{unit}
10076 Generate floating point arithmetics for selected unit @var{unit}. The choices
10077 for @var{unit} are:
10081 Use the standard 387 floating point coprocessor present majority of chips and
10082 emulated otherwise. Code compiled with this option will run almost everywhere.
10083 The temporary results are computed in 80bit precision instead of precision
10084 specified by the type resulting in slightly different results compared to most
10085 of other chips. See @option{-ffloat-store} for more detailed description.
10087 This is the default choice for i386 compiler.
10090 Use scalar floating point instructions present in the SSE instruction set.
10091 This instruction set is supported by Pentium3 and newer chips, in the AMD line
10092 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
10093 instruction set supports only single precision arithmetics, thus the double and
10094 extended precision arithmetics is still done using 387. Later version, present
10095 only in Pentium4 and the future AMD x86-64 chips supports double precision
10098 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
10099 or @option{-msse2} switches to enable SSE extensions and make this option
10100 effective. For the x86-64 compiler, these extensions are enabled by default.
10102 The resulting code should be considerably faster in the majority of cases and avoid
10103 the numerical instability problems of 387 code, but may break some existing
10104 code that expects temporaries to be 80bit.
10106 This is the default choice for the x86-64 compiler.
10109 Attempt to utilize both instruction sets at once. This effectively double the
10110 amount of available registers and on chips with separate execution units for
10111 387 and SSE the execution resources too. Use this option with care, as it is
10112 still experimental, because the GCC register allocator does not model separate
10113 functional units well resulting in instable performance.
10116 @item -masm=@var{dialect}
10117 @opindex masm=@var{dialect}
10118 Output asm instructions using selected @var{dialect}. Supported
10119 choices are @samp{intel} or @samp{att} (the default one). Darwin does
10120 not support @samp{intel}.
10123 @itemx -mno-ieee-fp
10125 @opindex mno-ieee-fp
10126 Control whether or not the compiler uses IEEE floating point
10127 comparisons. These handle correctly the case where the result of a
10128 comparison is unordered.
10131 @opindex msoft-float
10132 Generate output containing library calls for floating point.
10133 @strong{Warning:} the requisite libraries are not part of GCC@.
10134 Normally the facilities of the machine's usual C compiler are used, but
10135 this can't be done directly in cross-compilation. You must make your
10136 own arrangements to provide suitable library functions for
10139 On machines where a function returns floating point results in the 80387
10140 register stack, some floating point opcodes may be emitted even if
10141 @option{-msoft-float} is used.
10143 @item -mno-fp-ret-in-387
10144 @opindex mno-fp-ret-in-387
10145 Do not use the FPU registers for return values of functions.
10147 The usual calling convention has functions return values of types
10148 @code{float} and @code{double} in an FPU register, even if there
10149 is no FPU@. The idea is that the operating system should emulate
10152 The option @option{-mno-fp-ret-in-387} causes such values to be returned
10153 in ordinary CPU registers instead.
10155 @item -mno-fancy-math-387
10156 @opindex mno-fancy-math-387
10157 Some 387 emulators do not support the @code{sin}, @code{cos} and
10158 @code{sqrt} instructions for the 387. Specify this option to avoid
10159 generating those instructions. This option is the default on FreeBSD,
10160 OpenBSD and NetBSD@. This option is overridden when @option{-march}
10161 indicates that the target cpu will always have an FPU and so the
10162 instruction will not need emulation. As of revision 2.6.1, these
10163 instructions are not generated unless you also use the
10164 @option{-funsafe-math-optimizations} switch.
10166 @item -malign-double
10167 @itemx -mno-align-double
10168 @opindex malign-double
10169 @opindex mno-align-double
10170 Control whether GCC aligns @code{double}, @code{long double}, and
10171 @code{long long} variables on a two word boundary or a one word
10172 boundary. Aligning @code{double} variables on a two word boundary will
10173 produce code that runs somewhat faster on a @samp{Pentium} at the
10174 expense of more memory.
10176 On x86-64, @option{-malign-double} is enabled by default.
10178 @strong{Warning:} if you use the @option{-malign-double} switch,
10179 structures containing the above types will be aligned differently than
10180 the published application binary interface specifications for the 386
10181 and will not be binary compatible with structures in code compiled
10182 without that switch.
10184 @item -m96bit-long-double
10185 @itemx -m128bit-long-double
10186 @opindex m96bit-long-double
10187 @opindex m128bit-long-double
10188 These switches control the size of @code{long double} type. The i386
10189 application binary interface specifies the size to be 96 bits,
10190 so @option{-m96bit-long-double} is the default in 32 bit mode.
10192 Modern architectures (Pentium and newer) would prefer @code{long double}
10193 to be aligned to an 8 or 16 byte boundary. In arrays or structures
10194 conforming to the ABI, this would not be possible. So specifying a
10195 @option{-m128bit-long-double} will align @code{long double}
10196 to a 16 byte boundary by padding the @code{long double} with an additional
10199 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
10200 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
10202 Notice that neither of these options enable any extra precision over the x87
10203 standard of 80 bits for a @code{long double}.
10205 @strong{Warning:} if you override the default value for your target ABI, the
10206 structures and arrays containing @code{long double} variables will change
10207 their size as well as function calling convention for function taking
10208 @code{long double} will be modified. Hence they will not be binary
10209 compatible with arrays or structures in code compiled without that switch.
10211 @item -mmlarge-data-threshold=@var{number}
10212 @opindex mlarge-data-threshold=@var{number}
10213 When @option{-mcmodel=medium} is specified, the data greater than
10214 @var{threshold} are placed in large data section. This value must be the
10215 same across all object linked into the binary and defaults to 65535.
10219 Use a different function-calling convention, in which functions that
10220 take a fixed number of arguments return with the @code{ret} @var{num}
10221 instruction, which pops their arguments while returning. This saves one
10222 instruction in the caller since there is no need to pop the arguments
10225 You can specify that an individual function is called with this calling
10226 sequence with the function attribute @samp{stdcall}. You can also
10227 override the @option{-mrtd} option by using the function attribute
10228 @samp{cdecl}. @xref{Function Attributes}.
10230 @strong{Warning:} this calling convention is incompatible with the one
10231 normally used on Unix, so you cannot use it if you need to call
10232 libraries compiled with the Unix compiler.
10234 Also, you must provide function prototypes for all functions that
10235 take variable numbers of arguments (including @code{printf});
10236 otherwise incorrect code will be generated for calls to those
10239 In addition, seriously incorrect code will result if you call a
10240 function with too many arguments. (Normally, extra arguments are
10241 harmlessly ignored.)
10243 @item -mregparm=@var{num}
10245 Control how many registers are used to pass integer arguments. By
10246 default, no registers are used to pass arguments, and at most 3
10247 registers can be used. You can control this behavior for a specific
10248 function by using the function attribute @samp{regparm}.
10249 @xref{Function Attributes}.
10251 @strong{Warning:} if you use this switch, and
10252 @var{num} is nonzero, then you must build all modules with the same
10253 value, including any libraries. This includes the system libraries and
10257 @opindex msseregparm
10258 Use SSE register passing conventions for float and double arguments
10259 and return values. You can control this behavior for a specific
10260 function by using the function attribute @samp{sseregparm}.
10261 @xref{Function Attributes}.
10263 @strong{Warning:} if you use this switch then you must build all
10264 modules with the same value, including any libraries. This includes
10265 the system libraries and startup modules.
10274 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
10275 is specified, the significands of results of floating-point operations are
10276 rounded to 24 bits (single precision); @option{-mpc64} rounds the the
10277 significands of results of floating-point operations to 53 bits (double
10278 precision) and @option{-mpc80} rounds the significands of results of
10279 floating-point operations to 64 bits (extended double precision), which is
10280 the default. When this option is used, floating-point operations in higher
10281 precisions are not available to the programmer without setting the FPU
10282 control word explicitly.
10284 Setting the rounding of floating-point operations to less than the default
10285 80 bits can speed some programs by 2% or more. Note that some mathematical
10286 libraries assume that extended precision (80 bit) floating-point operations
10287 are enabled by default; routines in such libraries could suffer significant
10288 loss of accuracy, typically through so-called "catastrophic cancellation",
10289 when this option is used to set the precision to less than extended precision.
10291 @item -mstackrealign
10292 @opindex mstackrealign
10293 Realign the stack at entry. On the Intel x86, the
10294 @option{-mstackrealign} option will generate an alternate prologue and
10295 epilogue that realigns the runtime stack. This supports mixing legacy
10296 codes that keep a 4-byte aligned stack with modern codes that keep a
10297 16-byte stack for SSE compatibility. The alternate prologue and
10298 epilogue are slower and bigger than the regular ones, and the
10299 alternate prologue requires an extra scratch register; this lowers the
10300 number of registers available if used in conjunction with the
10301 @code{regparm} attribute. The @option{-mstackrealign} option is
10302 incompatible with the nested function prologue; this is considered a
10303 hard error. See also the attribute @code{force_align_arg_pointer},
10304 applicable to individual functions.
10306 @item -mpreferred-stack-boundary=@var{num}
10307 @opindex mpreferred-stack-boundary
10308 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
10309 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
10310 the default is 4 (16 bytes or 128 bits).
10312 On Pentium and PentiumPro, @code{double} and @code{long double} values
10313 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
10314 suffer significant run time performance penalties. On Pentium III, the
10315 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
10316 properly if it is not 16 byte aligned.
10318 To ensure proper alignment of this values on the stack, the stack boundary
10319 must be as aligned as that required by any value stored on the stack.
10320 Further, every function must be generated such that it keeps the stack
10321 aligned. Thus calling a function compiled with a higher preferred
10322 stack boundary from a function compiled with a lower preferred stack
10323 boundary will most likely misalign the stack. It is recommended that
10324 libraries that use callbacks always use the default setting.
10326 This extra alignment does consume extra stack space, and generally
10327 increases code size. Code that is sensitive to stack space usage, such
10328 as embedded systems and operating system kernels, may want to reduce the
10329 preferred alignment to @option{-mpreferred-stack-boundary=2}.
10361 These switches enable or disable the use of instructions in the MMX,
10362 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4A, ABM or 3DNow! extended
10364 These extensions are also available as built-in functions: see
10365 @ref{X86 Built-in Functions}, for details of the functions enabled and
10366 disabled by these switches.
10368 To have SSE/SSE2 instructions generated automatically from floating-point
10369 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
10371 These options will enable GCC to use these extended instructions in
10372 generated code, even without @option{-mfpmath=sse}. Applications which
10373 perform runtime CPU detection must compile separate files for each
10374 supported architecture, using the appropriate flags. In particular,
10375 the file containing the CPU detection code should be compiled without
10380 This option will enable GCC to use CMPXCHG16B instruction in generated code.
10381 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
10382 data types. This is useful for high resolution counters that could be updated
10383 by multiple processors (or cores). This instruction is generated as part of
10384 atomic built-in functions: see @ref{Atomic Builtins} for details.
10388 This option will enable GCC to use SAHF instruction in generated 64-bit code.
10389 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
10390 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
10391 SAHF are load and store instructions, respectively, for certain status flags.
10392 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
10393 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
10397 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
10398 vectorized variants RCPPS and RSQRTPS) instead of DIVSS and SQRTSS (and their
10399 vectorized variants). These instructions will be generated only when
10400 @option{-funsafe-math-optimizations} is enabled.
10403 @itemx -mno-push-args
10404 @opindex mpush-args
10405 @opindex mno-push-args
10406 Use PUSH operations to store outgoing parameters. This method is shorter
10407 and usually equally fast as method using SUB/MOV operations and is enabled
10408 by default. In some cases disabling it may improve performance because of
10409 improved scheduling and reduced dependencies.
10411 @item -maccumulate-outgoing-args
10412 @opindex maccumulate-outgoing-args
10413 If enabled, the maximum amount of space required for outgoing arguments will be
10414 computed in the function prologue. This is faster on most modern CPUs
10415 because of reduced dependencies, improved scheduling and reduced stack usage
10416 when preferred stack boundary is not equal to 2. The drawback is a notable
10417 increase in code size. This switch implies @option{-mno-push-args}.
10421 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
10422 on thread-safe exception handling must compile and link all code with the
10423 @option{-mthreads} option. When compiling, @option{-mthreads} defines
10424 @option{-D_MT}; when linking, it links in a special thread helper library
10425 @option{-lmingwthrd} which cleans up per thread exception handling data.
10427 @item -mno-align-stringops
10428 @opindex mno-align-stringops
10429 Do not align destination of inlined string operations. This switch reduces
10430 code size and improves performance in case the destination is already aligned,
10431 but GCC doesn't know about it.
10433 @item -minline-all-stringops
10434 @opindex minline-all-stringops
10435 By default GCC inlines string operations only when destination is known to be
10436 aligned at least to 4 byte boundary. This enables more inlining, increase code
10437 size, but may improve performance of code that depends on fast memcpy, strlen
10438 and memset for short lengths.
10440 @item -minline-stringops-dynamically
10441 @opindex minline-stringops-dynamically
10442 For string operation of unknown size, inline runtime checks so for small
10443 blocks inline code is used, while for large blocks library call is used.
10445 @item -mstringop-strategy=@var{alg}
10446 @opindex mstringop-strategy=@var{alg}
10447 Overwrite internal decision heuristic about particular algorithm to inline
10448 string operation with. The allowed values are @code{rep_byte},
10449 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
10450 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
10451 expanding inline loop, @code{libcall} for always expanding library call.
10453 @item -momit-leaf-frame-pointer
10454 @opindex momit-leaf-frame-pointer
10455 Don't keep the frame pointer in a register for leaf functions. This
10456 avoids the instructions to save, set up and restore frame pointers and
10457 makes an extra register available in leaf functions. The option
10458 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10459 which might make debugging harder.
10461 @item -mtls-direct-seg-refs
10462 @itemx -mno-tls-direct-seg-refs
10463 @opindex mtls-direct-seg-refs
10464 Controls whether TLS variables may be accessed with offsets from the
10465 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
10466 or whether the thread base pointer must be added. Whether or not this
10467 is legal depends on the operating system, and whether it maps the
10468 segment to cover the entire TLS area.
10470 For systems that use GNU libc, the default is on.
10473 These @samp{-m} switches are supported in addition to the above
10474 on AMD x86-64 processors in 64-bit environments.
10481 Generate code for a 32-bit or 64-bit environment.
10482 The 32-bit environment sets int, long and pointer to 32 bits and
10483 generates code that runs on any i386 system.
10484 The 64-bit environment sets int to 32 bits and long and pointer
10485 to 64 bits and generates code for AMD's x86-64 architecture. For
10486 darwin only the -m64 option turns off the @option{-fno-pic} and
10487 @option{-mdynamic-no-pic} options.
10489 @item -mno-red-zone
10490 @opindex no-red-zone
10491 Do not use a so called red zone for x86-64 code. The red zone is mandated
10492 by the x86-64 ABI, it is a 128-byte area beyond the location of the
10493 stack pointer that will not be modified by signal or interrupt handlers
10494 and therefore can be used for temporary data without adjusting the stack
10495 pointer. The flag @option{-mno-red-zone} disables this red zone.
10497 @item -mcmodel=small
10498 @opindex mcmodel=small
10499 Generate code for the small code model: the program and its symbols must
10500 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
10501 Programs can be statically or dynamically linked. This is the default
10504 @item -mcmodel=kernel
10505 @opindex mcmodel=kernel
10506 Generate code for the kernel code model. The kernel runs in the
10507 negative 2 GB of the address space.
10508 This model has to be used for Linux kernel code.
10510 @item -mcmodel=medium
10511 @opindex mcmodel=medium
10512 Generate code for the medium model: The program is linked in the lower 2
10513 GB of the address space but symbols can be located anywhere in the
10514 address space. Programs can be statically or dynamically linked, but
10515 building of shared libraries are not supported with the medium model.
10517 @item -mcmodel=large
10518 @opindex mcmodel=large
10519 Generate code for the large model: This model makes no assumptions
10520 about addresses and sizes of sections.
10523 @node IA-64 Options
10524 @subsection IA-64 Options
10525 @cindex IA-64 Options
10527 These are the @samp{-m} options defined for the Intel IA-64 architecture.
10531 @opindex mbig-endian
10532 Generate code for a big endian target. This is the default for HP-UX@.
10534 @item -mlittle-endian
10535 @opindex mlittle-endian
10536 Generate code for a little endian target. This is the default for AIX5
10542 @opindex mno-gnu-as
10543 Generate (or don't) code for the GNU assembler. This is the default.
10544 @c Also, this is the default if the configure option @option{--with-gnu-as}
10550 @opindex mno-gnu-ld
10551 Generate (or don't) code for the GNU linker. This is the default.
10552 @c Also, this is the default if the configure option @option{--with-gnu-ld}
10557 Generate code that does not use a global pointer register. The result
10558 is not position independent code, and violates the IA-64 ABI@.
10560 @item -mvolatile-asm-stop
10561 @itemx -mno-volatile-asm-stop
10562 @opindex mvolatile-asm-stop
10563 @opindex mno-volatile-asm-stop
10564 Generate (or don't) a stop bit immediately before and after volatile asm
10567 @item -mregister-names
10568 @itemx -mno-register-names
10569 @opindex mregister-names
10570 @opindex mno-register-names
10571 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
10572 the stacked registers. This may make assembler output more readable.
10578 Disable (or enable) optimizations that use the small data section. This may
10579 be useful for working around optimizer bugs.
10581 @item -mconstant-gp
10582 @opindex mconstant-gp
10583 Generate code that uses a single constant global pointer value. This is
10584 useful when compiling kernel code.
10588 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
10589 This is useful when compiling firmware code.
10591 @item -minline-float-divide-min-latency
10592 @opindex minline-float-divide-min-latency
10593 Generate code for inline divides of floating point values
10594 using the minimum latency algorithm.
10596 @item -minline-float-divide-max-throughput
10597 @opindex minline-float-divide-max-throughput
10598 Generate code for inline divides of floating point values
10599 using the maximum throughput algorithm.
10601 @item -minline-int-divide-min-latency
10602 @opindex minline-int-divide-min-latency
10603 Generate code for inline divides of integer values
10604 using the minimum latency algorithm.
10606 @item -minline-int-divide-max-throughput
10607 @opindex minline-int-divide-max-throughput
10608 Generate code for inline divides of integer values
10609 using the maximum throughput algorithm.
10611 @item -minline-sqrt-min-latency
10612 @opindex minline-sqrt-min-latency
10613 Generate code for inline square roots
10614 using the minimum latency algorithm.
10616 @item -minline-sqrt-max-throughput
10617 @opindex minline-sqrt-max-throughput
10618 Generate code for inline square roots
10619 using the maximum throughput algorithm.
10621 @item -mno-dwarf2-asm
10622 @itemx -mdwarf2-asm
10623 @opindex mno-dwarf2-asm
10624 @opindex mdwarf2-asm
10625 Don't (or do) generate assembler code for the DWARF2 line number debugging
10626 info. This may be useful when not using the GNU assembler.
10628 @item -mearly-stop-bits
10629 @itemx -mno-early-stop-bits
10630 @opindex mearly-stop-bits
10631 @opindex mno-early-stop-bits
10632 Allow stop bits to be placed earlier than immediately preceding the
10633 instruction that triggered the stop bit. This can improve instruction
10634 scheduling, but does not always do so.
10636 @item -mfixed-range=@var{register-range}
10637 @opindex mfixed-range
10638 Generate code treating the given register range as fixed registers.
10639 A fixed register is one that the register allocator can not use. This is
10640 useful when compiling kernel code. A register range is specified as
10641 two registers separated by a dash. Multiple register ranges can be
10642 specified separated by a comma.
10644 @item -mtls-size=@var{tls-size}
10646 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
10649 @item -mtune=@var{cpu-type}
10651 Tune the instruction scheduling for a particular CPU, Valid values are
10652 itanium, itanium1, merced, itanium2, and mckinley.
10658 Add support for multithreading using the POSIX threads library. This
10659 option sets flags for both the preprocessor and linker. It does
10660 not affect the thread safety of object code produced by the compiler or
10661 that of libraries supplied with it. These are HP-UX specific flags.
10667 Generate code for a 32-bit or 64-bit environment.
10668 The 32-bit environment sets int, long and pointer to 32 bits.
10669 The 64-bit environment sets int to 32 bits and long and pointer
10670 to 64 bits. These are HP-UX specific flags.
10672 @item -mno-sched-br-data-spec
10673 @itemx -msched-br-data-spec
10674 @opindex mno-sched-br-data-spec
10675 @opindex msched-br-data-spec
10676 (Dis/En)able data speculative scheduling before reload.
10677 This will result in generation of the ld.a instructions and
10678 the corresponding check instructions (ld.c / chk.a).
10679 The default is 'disable'.
10681 @item -msched-ar-data-spec
10682 @itemx -mno-sched-ar-data-spec
10683 @opindex msched-ar-data-spec
10684 @opindex mno-sched-ar-data-spec
10685 (En/Dis)able data speculative scheduling after reload.
10686 This will result in generation of the ld.a instructions and
10687 the corresponding check instructions (ld.c / chk.a).
10688 The default is 'enable'.
10690 @item -mno-sched-control-spec
10691 @itemx -msched-control-spec
10692 @opindex mno-sched-control-spec
10693 @opindex msched-control-spec
10694 (Dis/En)able control speculative scheduling. This feature is
10695 available only during region scheduling (i.e. before reload).
10696 This will result in generation of the ld.s instructions and
10697 the corresponding check instructions chk.s .
10698 The default is 'disable'.
10700 @item -msched-br-in-data-spec
10701 @itemx -mno-sched-br-in-data-spec
10702 @opindex msched-br-in-data-spec
10703 @opindex mno-sched-br-in-data-spec
10704 (En/Dis)able speculative scheduling of the instructions that
10705 are dependent on the data speculative loads before reload.
10706 This is effective only with @option{-msched-br-data-spec} enabled.
10707 The default is 'enable'.
10709 @item -msched-ar-in-data-spec
10710 @itemx -mno-sched-ar-in-data-spec
10711 @opindex msched-ar-in-data-spec
10712 @opindex mno-sched-ar-in-data-spec
10713 (En/Dis)able speculative scheduling of the instructions that
10714 are dependent on the data speculative loads after reload.
10715 This is effective only with @option{-msched-ar-data-spec} enabled.
10716 The default is 'enable'.
10718 @item -msched-in-control-spec
10719 @itemx -mno-sched-in-control-spec
10720 @opindex msched-in-control-spec
10721 @opindex mno-sched-in-control-spec
10722 (En/Dis)able speculative scheduling of the instructions that
10723 are dependent on the control speculative loads.
10724 This is effective only with @option{-msched-control-spec} enabled.
10725 The default is 'enable'.
10728 @itemx -mno-sched-ldc
10729 @opindex msched-ldc
10730 @opindex mno-sched-ldc
10731 (En/Dis)able use of simple data speculation checks ld.c .
10732 If disabled, only chk.a instructions will be emitted to check
10733 data speculative loads.
10734 The default is 'enable'.
10736 @item -mno-sched-control-ldc
10737 @itemx -msched-control-ldc
10738 @opindex mno-sched-control-ldc
10739 @opindex msched-control-ldc
10740 (Dis/En)able use of ld.c instructions to check control speculative loads.
10741 If enabled, in case of control speculative load with no speculatively
10742 scheduled dependent instructions this load will be emitted as ld.sa and
10743 ld.c will be used to check it.
10744 The default is 'disable'.
10746 @item -mno-sched-spec-verbose
10747 @itemx -msched-spec-verbose
10748 @opindex mno-sched-spec-verbose
10749 @opindex msched-spec-verbose
10750 (Dis/En)able printing of the information about speculative motions.
10752 @item -mno-sched-prefer-non-data-spec-insns
10753 @itemx -msched-prefer-non-data-spec-insns
10754 @opindex mno-sched-prefer-non-data-spec-insns
10755 @opindex msched-prefer-non-data-spec-insns
10756 If enabled, data speculative instructions will be chosen for schedule
10757 only if there are no other choices at the moment. This will make
10758 the use of the data speculation much more conservative.
10759 The default is 'disable'.
10761 @item -mno-sched-prefer-non-control-spec-insns
10762 @itemx -msched-prefer-non-control-spec-insns
10763 @opindex mno-sched-prefer-non-control-spec-insns
10764 @opindex msched-prefer-non-control-spec-insns
10765 If enabled, control speculative instructions will be chosen for schedule
10766 only if there are no other choices at the moment. This will make
10767 the use of the control speculation much more conservative.
10768 The default is 'disable'.
10770 @item -mno-sched-count-spec-in-critical-path
10771 @itemx -msched-count-spec-in-critical-path
10772 @opindex mno-sched-count-spec-in-critical-path
10773 @opindex msched-count-spec-in-critical-path
10774 If enabled, speculative dependencies will be considered during
10775 computation of the instructions priorities. This will make the use of the
10776 speculation a bit more conservative.
10777 The default is 'disable'.
10782 @subsection M32C Options
10783 @cindex M32C options
10786 @item -mcpu=@var{name}
10788 Select the CPU for which code is generated. @var{name} may be one of
10789 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10790 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10791 the M32C/80 series.
10795 Specifies that the program will be run on the simulator. This causes
10796 an alternate runtime library to be linked in which supports, for
10797 example, file I/O. You must not use this option when generating
10798 programs that will run on real hardware; you must provide your own
10799 runtime library for whatever I/O functions are needed.
10801 @item -memregs=@var{number}
10803 Specifies the number of memory-based pseudo-registers GCC will use
10804 during code generation. These pseudo-registers will be used like real
10805 registers, so there is a tradeoff between GCC's ability to fit the
10806 code into available registers, and the performance penalty of using
10807 memory instead of registers. Note that all modules in a program must
10808 be compiled with the same value for this option. Because of that, you
10809 must not use this option with the default runtime libraries gcc
10814 @node M32R/D Options
10815 @subsection M32R/D Options
10816 @cindex M32R/D options
10818 These @option{-m} options are defined for Renesas M32R/D architectures:
10823 Generate code for the M32R/2@.
10827 Generate code for the M32R/X@.
10831 Generate code for the M32R@. This is the default.
10833 @item -mmodel=small
10834 @opindex mmodel=small
10835 Assume all objects live in the lower 16MB of memory (so that their addresses
10836 can be loaded with the @code{ld24} instruction), and assume all subroutines
10837 are reachable with the @code{bl} instruction.
10838 This is the default.
10840 The addressability of a particular object can be set with the
10841 @code{model} attribute.
10843 @item -mmodel=medium
10844 @opindex mmodel=medium
10845 Assume objects may be anywhere in the 32-bit address space (the compiler
10846 will generate @code{seth/add3} instructions to load their addresses), and
10847 assume all subroutines are reachable with the @code{bl} instruction.
10849 @item -mmodel=large
10850 @opindex mmodel=large
10851 Assume objects may be anywhere in the 32-bit address space (the compiler
10852 will generate @code{seth/add3} instructions to load their addresses), and
10853 assume subroutines may not be reachable with the @code{bl} instruction
10854 (the compiler will generate the much slower @code{seth/add3/jl}
10855 instruction sequence).
10858 @opindex msdata=none
10859 Disable use of the small data area. Variables will be put into
10860 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10861 @code{section} attribute has been specified).
10862 This is the default.
10864 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10865 Objects may be explicitly put in the small data area with the
10866 @code{section} attribute using one of these sections.
10868 @item -msdata=sdata
10869 @opindex msdata=sdata
10870 Put small global and static data in the small data area, but do not
10871 generate special code to reference them.
10874 @opindex msdata=use
10875 Put small global and static data in the small data area, and generate
10876 special instructions to reference them.
10880 @cindex smaller data references
10881 Put global and static objects less than or equal to @var{num} bytes
10882 into the small data or bss sections instead of the normal data or bss
10883 sections. The default value of @var{num} is 8.
10884 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10885 for this option to have any effect.
10887 All modules should be compiled with the same @option{-G @var{num}} value.
10888 Compiling with different values of @var{num} may or may not work; if it
10889 doesn't the linker will give an error message---incorrect code will not be
10894 Makes the M32R specific code in the compiler display some statistics
10895 that might help in debugging programs.
10897 @item -malign-loops
10898 @opindex malign-loops
10899 Align all loops to a 32-byte boundary.
10901 @item -mno-align-loops
10902 @opindex mno-align-loops
10903 Do not enforce a 32-byte alignment for loops. This is the default.
10905 @item -missue-rate=@var{number}
10906 @opindex missue-rate=@var{number}
10907 Issue @var{number} instructions per cycle. @var{number} can only be 1
10910 @item -mbranch-cost=@var{number}
10911 @opindex mbranch-cost=@var{number}
10912 @var{number} can only be 1 or 2. If it is 1 then branches will be
10913 preferred over conditional code, if it is 2, then the opposite will
10916 @item -mflush-trap=@var{number}
10917 @opindex mflush-trap=@var{number}
10918 Specifies the trap number to use to flush the cache. The default is
10919 12. Valid numbers are between 0 and 15 inclusive.
10921 @item -mno-flush-trap
10922 @opindex mno-flush-trap
10923 Specifies that the cache cannot be flushed by using a trap.
10925 @item -mflush-func=@var{name}
10926 @opindex mflush-func=@var{name}
10927 Specifies the name of the operating system function to call to flush
10928 the cache. The default is @emph{_flush_cache}, but a function call
10929 will only be used if a trap is not available.
10931 @item -mno-flush-func
10932 @opindex mno-flush-func
10933 Indicates that there is no OS function for flushing the cache.
10937 @node M680x0 Options
10938 @subsection M680x0 Options
10939 @cindex M680x0 options
10941 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
10942 The default settings depend on which architecture was selected when
10943 the compiler was configured; the defaults for the most common choices
10947 @item -march=@var{arch}
10949 Generate code for a specific M680x0 or ColdFire instruction set
10950 architecture. Permissible values of @var{arch} for M680x0
10951 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
10952 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
10953 architectures are selected according to Freescale's ISA classification
10954 and the permissible values are: @samp{isaa}, @samp{isaaplus},
10955 @samp{isab} and @samp{isac}.
10957 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
10958 code for a ColdFire target. The @var{arch} in this macro is one of the
10959 @option{-march} arguments given above.
10961 When used together, @option{-march} and @option{-mtune} select code
10962 that runs on a family of similar processors but that is optimized
10963 for a particular microarchitecture.
10965 @item -mcpu=@var{cpu}
10967 Generate code for a specific M680x0 or ColdFire processor.
10968 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
10969 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
10970 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
10971 below, which also classifies the CPUs into families:
10973 @multitable @columnfractions 0.20 0.80
10974 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
10975 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
10976 @item @samp{5206e} @tab @samp{5206e}
10977 @item @samp{5208} @tab @samp{5207} @samp{5208}
10978 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
10979 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
10980 @item @samp{5216} @tab @samp{5214} @samp{5216}
10981 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
10982 @item @samp{5225} @tab @samp{5224} @samp{5225}
10983 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
10984 @item @samp{5249} @tab @samp{5249}
10985 @item @samp{5250} @tab @samp{5250}
10986 @item @samp{5271} @tab @samp{5270} @samp{5271}
10987 @item @samp{5272} @tab @samp{5272}
10988 @item @samp{5275} @tab @samp{5274} @samp{5275}
10989 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
10990 @item @samp{5307} @tab @samp{5307}
10991 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
10992 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
10993 @item @samp{5407} @tab @samp{5407}
10994 @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}
10997 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
10998 @var{arch} is compatible with @var{cpu}. Other combinations of
10999 @option{-mcpu} and @option{-march} are rejected.
11001 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
11002 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
11003 where the value of @var{family} is given by the table above.
11005 @item -mtune=@var{tune}
11007 Tune the code for a particular microarchitecture, within the
11008 constraints set by @option{-march} and @option{-mcpu}.
11009 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
11010 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
11011 and @samp{cpu32}. The ColdFire microarchitectures
11012 are: @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
11014 You can also use @option{-mtune=68020-40} for code that needs
11015 to run relatively well on 68020, 68030 and 68040 targets.
11016 @option{-mtune=68020-60} is similar but includes 68060 targets
11017 as well. These two options select the same tuning decisions as
11018 @option{-m68020-40} and @option{-m68020-60} respectively.
11020 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
11021 when tuning for 680x0 architecture @var{arch}. It also defines
11022 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
11023 option is used. If gcc is tuning for a range of architectures,
11024 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
11025 it defines the macros for every architecture in the range.
11027 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
11028 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
11029 of the arguments given above.
11035 Generate output for a 68000. This is the default
11036 when the compiler is configured for 68000-based systems.
11037 It is equivalent to @option{-march=68000}.
11039 Use this option for microcontrollers with a 68000 or EC000 core,
11040 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
11044 Generate output for a 68010. This is the default
11045 when the compiler is configured for 68010-based systems.
11046 It is equivalent to @option{-march=68010}.
11052 Generate output for a 68020. This is the default
11053 when the compiler is configured for 68020-based systems.
11054 It is equivalent to @option{-march=68020}.
11058 Generate output for a 68030. This is the default when the compiler is
11059 configured for 68030-based systems. It is equivalent to
11060 @option{-march=68030}.
11064 Generate output for a 68040. This is the default when the compiler is
11065 configured for 68040-based systems. It is equivalent to
11066 @option{-march=68040}.
11068 This option inhibits the use of 68881/68882 instructions that have to be
11069 emulated by software on the 68040. Use this option if your 68040 does not
11070 have code to emulate those instructions.
11074 Generate output for a 68060. This is the default when the compiler is
11075 configured for 68060-based systems. It is equivalent to
11076 @option{-march=68060}.
11078 This option inhibits the use of 68020 and 68881/68882 instructions that
11079 have to be emulated by software on the 68060. Use this option if your 68060
11080 does not have code to emulate those instructions.
11084 Generate output for a CPU32. This is the default
11085 when the compiler is configured for CPU32-based systems.
11086 It is equivalent to @option{-march=cpu32}.
11088 Use this option for microcontrollers with a
11089 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
11090 68336, 68340, 68341, 68349 and 68360.
11094 Generate output for a 520X ColdFire CPU. This is the default
11095 when the compiler is configured for 520X-based systems.
11096 It is equivalent to @option{-mcpu=5206}, and is now deprecated
11097 in favor of that option.
11099 Use this option for microcontroller with a 5200 core, including
11100 the MCF5202, MCF5203, MCF5204 and MCF5206.
11104 Generate output for a 5206e ColdFire CPU. The option is now
11105 deprecated in favor of the equivalent @option{-mcpu=5206e}.
11109 Generate output for a member of the ColdFire 528X family.
11110 The option is now deprecated in favor of the equivalent
11111 @option{-mcpu=528x}.
11115 Generate output for a ColdFire 5307 CPU. The option is now deprecated
11116 in favor of the equivalent @option{-mcpu=5307}.
11120 Generate output for a ColdFire 5407 CPU. The option is now deprecated
11121 in favor of the equivalent @option{-mcpu=5407}.
11125 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
11126 This includes use of hardware floating point instructions.
11127 The option is equivalent to @option{-mcpu=547x}, and is now
11128 deprecated in favor of that option.
11132 Generate output for a 68040, without using any of the new instructions.
11133 This results in code which can run relatively efficiently on either a
11134 68020/68881 or a 68030 or a 68040. The generated code does use the
11135 68881 instructions that are emulated on the 68040.
11137 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
11141 Generate output for a 68060, without using any of the new instructions.
11142 This results in code which can run relatively efficiently on either a
11143 68020/68881 or a 68030 or a 68040. The generated code does use the
11144 68881 instructions that are emulated on the 68060.
11146 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
11150 @opindex mhard-float
11152 Generate floating-point instructions. This is the default for 68020
11153 and above, and for ColdFire devices that have an FPU. It defines the
11154 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
11155 on ColdFire targets.
11158 @opindex msoft-float
11159 Do not generate floating-point instructions; use library calls instead.
11160 This is the default for 68000, 68010, and 68832 targets. It is also
11161 the default for ColdFire devices that have no FPU.
11167 Generate (do not generate) ColdFire hardware divide and remainder
11168 instructions. If @option{-march} is used without @option{-mcpu},
11169 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
11170 architectures. Otherwise, the default is taken from the target CPU
11171 (either the default CPU, or the one specified by @option{-mcpu}). For
11172 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
11173 @option{-mcpu=5206e}.
11175 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
11179 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11180 Additionally, parameters passed on the stack are also aligned to a
11181 16-bit boundary even on targets whose API mandates promotion to 32-bit.
11185 Do not consider type @code{int} to be 16 bits wide. This is the default.
11188 @itemx -mno-bitfield
11189 @opindex mnobitfield
11190 @opindex mno-bitfield
11191 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
11192 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
11196 Do use the bit-field instructions. The @option{-m68020} option implies
11197 @option{-mbitfield}. This is the default if you use a configuration
11198 designed for a 68020.
11202 Use a different function-calling convention, in which functions
11203 that take a fixed number of arguments return with the @code{rtd}
11204 instruction, which pops their arguments while returning. This
11205 saves one instruction in the caller since there is no need to pop
11206 the arguments there.
11208 This calling convention is incompatible with the one normally
11209 used on Unix, so you cannot use it if you need to call libraries
11210 compiled with the Unix compiler.
11212 Also, you must provide function prototypes for all functions that
11213 take variable numbers of arguments (including @code{printf});
11214 otherwise incorrect code will be generated for calls to those
11217 In addition, seriously incorrect code will result if you call a
11218 function with too many arguments. (Normally, extra arguments are
11219 harmlessly ignored.)
11221 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
11222 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
11226 Do not use the calling conventions selected by @option{-mrtd}.
11227 This is the default.
11230 @itemx -mno-align-int
11231 @opindex malign-int
11232 @opindex mno-align-int
11233 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
11234 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
11235 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
11236 Aligning variables on 32-bit boundaries produces code that runs somewhat
11237 faster on processors with 32-bit busses at the expense of more memory.
11239 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
11240 align structures containing the above types differently than
11241 most published application binary interface specifications for the m68k.
11245 Use the pc-relative addressing mode of the 68000 directly, instead of
11246 using a global offset table. At present, this option implies @option{-fpic},
11247 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
11248 not presently supported with @option{-mpcrel}, though this could be supported for
11249 68020 and higher processors.
11251 @item -mno-strict-align
11252 @itemx -mstrict-align
11253 @opindex mno-strict-align
11254 @opindex mstrict-align
11255 Do not (do) assume that unaligned memory references will be handled by
11259 Generate code that allows the data segment to be located in a different
11260 area of memory from the text segment. This allows for execute in place in
11261 an environment without virtual memory management. This option implies
11264 @item -mno-sep-data
11265 Generate code that assumes that the data segment follows the text segment.
11266 This is the default.
11268 @item -mid-shared-library
11269 Generate code that supports shared libraries via the library ID method.
11270 This allows for execute in place and shared libraries in an environment
11271 without virtual memory management. This option implies @option{-fPIC}.
11273 @item -mno-id-shared-library
11274 Generate code that doesn't assume ID based shared libraries are being used.
11275 This is the default.
11277 @item -mshared-library-id=n
11278 Specified the identification number of the ID based shared library being
11279 compiled. Specifying a value of 0 will generate more compact code, specifying
11280 other values will force the allocation of that number to the current
11281 library but is no more space or time efficient than omitting this option.
11285 @node M68hc1x Options
11286 @subsection M68hc1x Options
11287 @cindex M68hc1x options
11289 These are the @samp{-m} options defined for the 68hc11 and 68hc12
11290 microcontrollers. The default values for these options depends on
11291 which style of microcontroller was selected when the compiler was configured;
11292 the defaults for the most common choices are given below.
11299 Generate output for a 68HC11. This is the default
11300 when the compiler is configured for 68HC11-based systems.
11306 Generate output for a 68HC12. This is the default
11307 when the compiler is configured for 68HC12-based systems.
11313 Generate output for a 68HCS12.
11315 @item -mauto-incdec
11316 @opindex mauto-incdec
11317 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
11324 Enable the use of 68HC12 min and max instructions.
11327 @itemx -mno-long-calls
11328 @opindex mlong-calls
11329 @opindex mno-long-calls
11330 Treat all calls as being far away (near). If calls are assumed to be
11331 far away, the compiler will use the @code{call} instruction to
11332 call a function and the @code{rtc} instruction for returning.
11336 Consider type @code{int} to be 16 bits wide, like @code{short int}.
11338 @item -msoft-reg-count=@var{count}
11339 @opindex msoft-reg-count
11340 Specify the number of pseudo-soft registers which are used for the
11341 code generation. The maximum number is 32. Using more pseudo-soft
11342 register may or may not result in better code depending on the program.
11343 The default is 4 for 68HC11 and 2 for 68HC12.
11347 @node MCore Options
11348 @subsection MCore Options
11349 @cindex MCore options
11351 These are the @samp{-m} options defined for the Motorola M*Core
11357 @itemx -mno-hardlit
11359 @opindex mno-hardlit
11360 Inline constants into the code stream if it can be done in two
11361 instructions or less.
11367 Use the divide instruction. (Enabled by default).
11369 @item -mrelax-immediate
11370 @itemx -mno-relax-immediate
11371 @opindex mrelax-immediate
11372 @opindex mno-relax-immediate
11373 Allow arbitrary sized immediates in bit operations.
11375 @item -mwide-bitfields
11376 @itemx -mno-wide-bitfields
11377 @opindex mwide-bitfields
11378 @opindex mno-wide-bitfields
11379 Always treat bit-fields as int-sized.
11381 @item -m4byte-functions
11382 @itemx -mno-4byte-functions
11383 @opindex m4byte-functions
11384 @opindex mno-4byte-functions
11385 Force all functions to be aligned to a four byte boundary.
11387 @item -mcallgraph-data
11388 @itemx -mno-callgraph-data
11389 @opindex mcallgraph-data
11390 @opindex mno-callgraph-data
11391 Emit callgraph information.
11394 @itemx -mno-slow-bytes
11395 @opindex mslow-bytes
11396 @opindex mno-slow-bytes
11397 Prefer word access when reading byte quantities.
11399 @item -mlittle-endian
11400 @itemx -mbig-endian
11401 @opindex mlittle-endian
11402 @opindex mbig-endian
11403 Generate code for a little endian target.
11409 Generate code for the 210 processor.
11413 @subsection MIPS Options
11414 @cindex MIPS options
11420 Generate big-endian code.
11424 Generate little-endian code. This is the default for @samp{mips*el-*-*}
11427 @item -march=@var{arch}
11429 Generate code that will run on @var{arch}, which can be the name of a
11430 generic MIPS ISA, or the name of a particular processor.
11432 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
11433 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
11434 The processor names are:
11435 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
11436 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
11437 @samp{5kc}, @samp{5kf},
11439 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
11440 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
11441 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
11442 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
11445 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
11446 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
11447 @samp{rm7000}, @samp{rm9000},
11450 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
11451 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
11452 The special value @samp{from-abi} selects the
11453 most compatible architecture for the selected ABI (that is,
11454 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
11456 In processor names, a final @samp{000} can be abbreviated as @samp{k}
11457 (for example, @samp{-march=r2k}). Prefixes are optional, and
11458 @samp{vr} may be written @samp{r}.
11460 Names of the form @samp{@var{n}f2_1} refer to processors with
11461 FPUs clocked at half the rate of the core, names of the form
11462 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
11463 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
11464 processors with FPUs clocked a ratio of 3:2 with respect to the core.
11465 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
11466 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
11467 accepted as synonyms for @samp{@var{n}f1_1}.
11469 GCC defines two macros based on the value of this option. The first
11470 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
11471 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
11472 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
11473 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
11474 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
11476 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
11477 above. In other words, it will have the full prefix and will not
11478 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
11479 the macro names the resolved architecture (either @samp{"mips1"} or
11480 @samp{"mips3"}). It names the default architecture when no
11481 @option{-march} option is given.
11483 @item -mtune=@var{arch}
11485 Optimize for @var{arch}. Among other things, this option controls
11486 the way instructions are scheduled, and the perceived cost of arithmetic
11487 operations. The list of @var{arch} values is the same as for
11490 When this option is not used, GCC will optimize for the processor
11491 specified by @option{-march}. By using @option{-march} and
11492 @option{-mtune} together, it is possible to generate code that will
11493 run on a family of processors, but optimize the code for one
11494 particular member of that family.
11496 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
11497 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
11498 @samp{-march} ones described above.
11502 Equivalent to @samp{-march=mips1}.
11506 Equivalent to @samp{-march=mips2}.
11510 Equivalent to @samp{-march=mips3}.
11514 Equivalent to @samp{-march=mips4}.
11518 Equivalent to @samp{-march=mips32}.
11522 Equivalent to @samp{-march=mips32r2}.
11526 Equivalent to @samp{-march=mips64}.
11531 @opindex mno-mips16
11532 Generate (do not generate) MIPS16 code. If GCC is targetting a
11533 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
11545 Generate code for the given ABI@.
11547 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
11548 generates 64-bit code when you select a 64-bit architecture, but you
11549 can use @option{-mgp32} to get 32-bit code instead.
11551 For information about the O64 ABI, see
11552 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
11554 GCC supports a variant of the o32 ABI in which floating-point registers
11555 are 64 rather than 32 bits wide. You can select this combination with
11556 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
11557 and @samp{mfhc1} instructions and is therefore only supported for
11558 MIPS32R2 processors.
11560 The register assignments for arguments and return values remain the
11561 same, but each scalar value is passed in a single 64-bit register
11562 rather than a pair of 32-bit registers. For example, scalar
11563 floating-point values are returned in @samp{$f0} only, not a
11564 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
11565 remains the same, but all 64 bits are saved.
11568 @itemx -mno-abicalls
11570 @opindex mno-abicalls
11571 Generate (do not generate) code that is suitable for SVR4-style
11572 dynamic objects. @option{-mabicalls} is the default for SVR4-based
11577 Generate (do not generate) code that is fully position-independent,
11578 and that can therefore be linked into shared libraries. This option
11579 only affects @option{-mabicalls}.
11581 All @option{-mabicalls} code has traditionally been position-independent,
11582 regardless of options like @option{-fPIC} and @option{-fpic}. However,
11583 as an extension, the GNU toolchain allows executables to use absolute
11584 accesses for locally-binding symbols. It can also use shorter GP
11585 initialization sequences and generate direct calls to locally-defined
11586 functions. This mode is selected by @option{-mno-shared}.
11588 @option{-mno-shared} depends on binutils 2.16 or higher and generates
11589 objects that can only be linked by the GNU linker. However, the option
11590 does not affect the ABI of the final executable; it only affects the ABI
11591 of relocatable objects. Using @option{-mno-shared} will generally make
11592 executables both smaller and quicker.
11594 @option{-mshared} is the default.
11600 Lift (do not lift) the usual restrictions on the size of the global
11603 GCC normally uses a single instruction to load values from the GOT@.
11604 While this is relatively efficient, it will only work if the GOT
11605 is smaller than about 64k. Anything larger will cause the linker
11606 to report an error such as:
11608 @cindex relocation truncated to fit (MIPS)
11610 relocation truncated to fit: R_MIPS_GOT16 foobar
11613 If this happens, you should recompile your code with @option{-mxgot}.
11614 It should then work with very large GOTs, although it will also be
11615 less efficient, since it will take three instructions to fetch the
11616 value of a global symbol.
11618 Note that some linkers can create multiple GOTs. If you have such a
11619 linker, you should only need to use @option{-mxgot} when a single object
11620 file accesses more than 64k's worth of GOT entries. Very few do.
11622 These options have no effect unless GCC is generating position
11627 Assume that general-purpose registers are 32 bits wide.
11631 Assume that general-purpose registers are 64 bits wide.
11635 Assume that floating-point registers are 32 bits wide.
11639 Assume that floating-point registers are 64 bits wide.
11642 @opindex mhard-float
11643 Use floating-point coprocessor instructions.
11646 @opindex msoft-float
11647 Do not use floating-point coprocessor instructions. Implement
11648 floating-point calculations using library calls instead.
11650 @item -msingle-float
11651 @opindex msingle-float
11652 Assume that the floating-point coprocessor only supports single-precision
11655 @item -mdouble-float
11656 @opindex mdouble-float
11657 Assume that the floating-point coprocessor supports double-precision
11658 operations. This is the default.
11664 Use (do not use) revision 1 of the MIPS DSP ASE.
11665 @xref{MIPS DSP Built-in Functions}. This option defines the
11666 preprocessor macro @samp{__mips_dsp}. It also defines
11667 @samp{__mips_dsp_rev} to 1.
11673 Use (do not use) revision 2 of the MIPS DSP ASE.
11674 @xref{MIPS DSP Built-in Functions}. This option defines the
11675 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
11676 It also defines @samp{__mips_dsp_rev} to 2.
11679 @itemx -mno-smartmips
11680 @opindex msmartmips
11681 @opindex mno-smartmips
11682 Use (do not use) the MIPS SmartMIPS ASE.
11684 @item -mpaired-single
11685 @itemx -mno-paired-single
11686 @opindex mpaired-single
11687 @opindex mno-paired-single
11688 Use (do not use) paired-single floating-point instructions.
11689 @xref{MIPS Paired-Single Support}. This option can only be used
11690 when generating 64-bit code and requires hardware floating-point
11691 support to be enabled.
11697 Use (do not use) MIPS Digital Media Extension instructions.
11698 This option can only be used when generating 64-bit code and requires
11699 hardware floating-point support to be enabled.
11704 @opindex mno-mips3d
11705 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
11706 The option @option{-mips3d} implies @option{-mpaired-single}.
11712 Use (do not use) MT Multithreading instructions.
11716 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
11717 an explanation of the default and the way that the pointer size is
11722 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
11724 The default size of @code{int}s, @code{long}s and pointers depends on
11725 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
11726 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
11727 32-bit @code{long}s. Pointers are the same size as @code{long}s,
11728 or the same size as integer registers, whichever is smaller.
11734 Assume (do not assume) that all symbols have 32-bit values, regardless
11735 of the selected ABI@. This option is useful in combination with
11736 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
11737 to generate shorter and faster references to symbolic addresses.
11741 @cindex smaller data references (MIPS)
11742 @cindex gp-relative references (MIPS)
11743 Put global and static items less than or equal to @var{num} bytes into
11744 the small data or bss section instead of the normal data or bss section.
11745 This allows the data to be accessed using a single instruction.
11747 All modules should be compiled with the same @option{-G @var{num}}
11750 @item -membedded-data
11751 @itemx -mno-embedded-data
11752 @opindex membedded-data
11753 @opindex mno-embedded-data
11754 Allocate variables to the read-only data section first if possible, then
11755 next in the small data section if possible, otherwise in data. This gives
11756 slightly slower code than the default, but reduces the amount of RAM required
11757 when executing, and thus may be preferred for some embedded systems.
11759 @item -muninit-const-in-rodata
11760 @itemx -mno-uninit-const-in-rodata
11761 @opindex muninit-const-in-rodata
11762 @opindex mno-uninit-const-in-rodata
11763 Put uninitialized @code{const} variables in the read-only data section.
11764 This option is only meaningful in conjunction with @option{-membedded-data}.
11766 @item -msplit-addresses
11767 @itemx -mno-split-addresses
11768 @opindex msplit-addresses
11769 @opindex mno-split-addresses
11770 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
11771 relocation operators. This option has been superseded by
11772 @option{-mexplicit-relocs} but is retained for backwards compatibility.
11774 @item -mexplicit-relocs
11775 @itemx -mno-explicit-relocs
11776 @opindex mexplicit-relocs
11777 @opindex mno-explicit-relocs
11778 Use (do not use) assembler relocation operators when dealing with symbolic
11779 addresses. The alternative, selected by @option{-mno-explicit-relocs},
11780 is to use assembler macros instead.
11782 @option{-mexplicit-relocs} is the default if GCC was configured
11783 to use an assembler that supports relocation operators.
11785 @item -mcheck-zero-division
11786 @itemx -mno-check-zero-division
11787 @opindex mcheck-zero-division
11788 @opindex mno-check-zero-division
11789 Trap (do not trap) on integer division by zero.
11791 The default is @option{-mcheck-zero-division}.
11793 @item -mdivide-traps
11794 @itemx -mdivide-breaks
11795 @opindex mdivide-traps
11796 @opindex mdivide-breaks
11797 MIPS systems check for division by zero by generating either a
11798 conditional trap or a break instruction. Using traps results in
11799 smaller code, but is only supported on MIPS II and later. Also, some
11800 versions of the Linux kernel have a bug that prevents trap from
11801 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
11802 allow conditional traps on architectures that support them and
11803 @option{-mdivide-breaks} to force the use of breaks.
11805 The default is usually @option{-mdivide-traps}, but this can be
11806 overridden at configure time using @option{--with-divide=breaks}.
11807 Divide-by-zero checks can be completely disabled using
11808 @option{-mno-check-zero-division}.
11813 @opindex mno-memcpy
11814 Force (do not force) the use of @code{memcpy()} for non-trivial block
11815 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
11816 most constant-sized copies.
11819 @itemx -mno-long-calls
11820 @opindex mlong-calls
11821 @opindex mno-long-calls
11822 Disable (do not disable) use of the @code{jal} instruction. Calling
11823 functions using @code{jal} is more efficient but requires the caller
11824 and callee to be in the same 256 megabyte segment.
11826 This option has no effect on abicalls code. The default is
11827 @option{-mno-long-calls}.
11833 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
11834 instructions, as provided by the R4650 ISA@.
11837 @itemx -mno-fused-madd
11838 @opindex mfused-madd
11839 @opindex mno-fused-madd
11840 Enable (disable) use of the floating point multiply-accumulate
11841 instructions, when they are available. The default is
11842 @option{-mfused-madd}.
11844 When multiply-accumulate instructions are used, the intermediate
11845 product is calculated to infinite precision and is not subject to
11846 the FCSR Flush to Zero bit. This may be undesirable in some
11851 Tell the MIPS assembler to not run its preprocessor over user
11852 assembler files (with a @samp{.s} suffix) when assembling them.
11855 @itemx -mno-fix-r4000
11856 @opindex mfix-r4000
11857 @opindex mno-fix-r4000
11858 Work around certain R4000 CPU errata:
11861 A double-word or a variable shift may give an incorrect result if executed
11862 immediately after starting an integer division.
11864 A double-word or a variable shift may give an incorrect result if executed
11865 while an integer multiplication is in progress.
11867 An integer division may give an incorrect result if started in a delay slot
11868 of a taken branch or a jump.
11872 @itemx -mno-fix-r4400
11873 @opindex mfix-r4400
11874 @opindex mno-fix-r4400
11875 Work around certain R4400 CPU errata:
11878 A double-word or a variable shift may give an incorrect result if executed
11879 immediately after starting an integer division.
11883 @itemx -mno-fix-vr4120
11884 @opindex mfix-vr4120
11885 Work around certain VR4120 errata:
11888 @code{dmultu} does not always produce the correct result.
11890 @code{div} and @code{ddiv} do not always produce the correct result if one
11891 of the operands is negative.
11893 The workarounds for the division errata rely on special functions in
11894 @file{libgcc.a}. At present, these functions are only provided by
11895 the @code{mips64vr*-elf} configurations.
11897 Other VR4120 errata require a nop to be inserted between certain pairs of
11898 instructions. These errata are handled by the assembler, not by GCC itself.
11901 @opindex mfix-vr4130
11902 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
11903 workarounds are implemented by the assembler rather than by GCC,
11904 although GCC will avoid using @code{mflo} and @code{mfhi} if the
11905 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
11906 instructions are available instead.
11909 @itemx -mno-fix-sb1
11911 Work around certain SB-1 CPU core errata.
11912 (This flag currently works around the SB-1 revision 2
11913 ``F1'' and ``F2'' floating point errata.)
11915 @item -mflush-func=@var{func}
11916 @itemx -mno-flush-func
11917 @opindex mflush-func
11918 Specifies the function to call to flush the I and D caches, or to not
11919 call any such function. If called, the function must take the same
11920 arguments as the common @code{_flush_func()}, that is, the address of the
11921 memory range for which the cache is being flushed, the size of the
11922 memory range, and the number 3 (to flush both caches). The default
11923 depends on the target GCC was configured for, but commonly is either
11924 @samp{_flush_func} or @samp{__cpu_flush}.
11926 @item mbranch-cost=@var{num}
11927 @opindex mbranch-cost
11928 Set the cost of branches to roughly @var{num} ``simple'' instructions.
11929 This cost is only a heuristic and is not guaranteed to produce
11930 consistent results across releases. A zero cost redundantly selects
11931 the default, which is based on the @option{-mtune} setting.
11933 @item -mbranch-likely
11934 @itemx -mno-branch-likely
11935 @opindex mbranch-likely
11936 @opindex mno-branch-likely
11937 Enable or disable use of Branch Likely instructions, regardless of the
11938 default for the selected architecture. By default, Branch Likely
11939 instructions may be generated if they are supported by the selected
11940 architecture. An exception is for the MIPS32 and MIPS64 architectures
11941 and processors which implement those architectures; for those, Branch
11942 Likely instructions will not be generated by default because the MIPS32
11943 and MIPS64 architectures specifically deprecate their use.
11945 @item -mfp-exceptions
11946 @itemx -mno-fp-exceptions
11947 @opindex mfp-exceptions
11948 Specifies whether FP exceptions are enabled. This affects how we schedule
11949 FP instructions for some processors. The default is that FP exceptions are
11952 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11953 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11956 @item -mvr4130-align
11957 @itemx -mno-vr4130-align
11958 @opindex mvr4130-align
11959 The VR4130 pipeline is two-way superscalar, but can only issue two
11960 instructions together if the first one is 8-byte aligned. When this
11961 option is enabled, GCC will align pairs of instructions that it
11962 thinks should execute in parallel.
11964 This option only has an effect when optimizing for the VR4130.
11965 It normally makes code faster, but at the expense of making it bigger.
11966 It is enabled by default at optimization level @option{-O3}.
11970 @subsection MMIX Options
11971 @cindex MMIX Options
11973 These options are defined for the MMIX:
11977 @itemx -mno-libfuncs
11979 @opindex mno-libfuncs
11980 Specify that intrinsic library functions are being compiled, passing all
11981 values in registers, no matter the size.
11984 @itemx -mno-epsilon
11986 @opindex mno-epsilon
11987 Generate floating-point comparison instructions that compare with respect
11988 to the @code{rE} epsilon register.
11990 @item -mabi=mmixware
11992 @opindex mabi-mmixware
11994 Generate code that passes function parameters and return values that (in
11995 the called function) are seen as registers @code{$0} and up, as opposed to
11996 the GNU ABI which uses global registers @code{$231} and up.
11998 @item -mzero-extend
11999 @itemx -mno-zero-extend
12000 @opindex mzero-extend
12001 @opindex mno-zero-extend
12002 When reading data from memory in sizes shorter than 64 bits, use (do not
12003 use) zero-extending load instructions by default, rather than
12004 sign-extending ones.
12007 @itemx -mno-knuthdiv
12009 @opindex mno-knuthdiv
12010 Make the result of a division yielding a remainder have the same sign as
12011 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
12012 remainder follows the sign of the dividend. Both methods are
12013 arithmetically valid, the latter being almost exclusively used.
12015 @item -mtoplevel-symbols
12016 @itemx -mno-toplevel-symbols
12017 @opindex mtoplevel-symbols
12018 @opindex mno-toplevel-symbols
12019 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
12020 code can be used with the @code{PREFIX} assembly directive.
12024 Generate an executable in the ELF format, rather than the default
12025 @samp{mmo} format used by the @command{mmix} simulator.
12027 @item -mbranch-predict
12028 @itemx -mno-branch-predict
12029 @opindex mbranch-predict
12030 @opindex mno-branch-predict
12031 Use (do not use) the probable-branch instructions, when static branch
12032 prediction indicates a probable branch.
12034 @item -mbase-addresses
12035 @itemx -mno-base-addresses
12036 @opindex mbase-addresses
12037 @opindex mno-base-addresses
12038 Generate (do not generate) code that uses @emph{base addresses}. Using a
12039 base address automatically generates a request (handled by the assembler
12040 and the linker) for a constant to be set up in a global register. The
12041 register is used for one or more base address requests within the range 0
12042 to 255 from the value held in the register. The generally leads to short
12043 and fast code, but the number of different data items that can be
12044 addressed is limited. This means that a program that uses lots of static
12045 data may require @option{-mno-base-addresses}.
12047 @item -msingle-exit
12048 @itemx -mno-single-exit
12049 @opindex msingle-exit
12050 @opindex mno-single-exit
12051 Force (do not force) generated code to have a single exit point in each
12055 @node MN10300 Options
12056 @subsection MN10300 Options
12057 @cindex MN10300 options
12059 These @option{-m} options are defined for Matsushita MN10300 architectures:
12064 Generate code to avoid bugs in the multiply instructions for the MN10300
12065 processors. This is the default.
12067 @item -mno-mult-bug
12068 @opindex mno-mult-bug
12069 Do not generate code to avoid bugs in the multiply instructions for the
12070 MN10300 processors.
12074 Generate code which uses features specific to the AM33 processor.
12078 Do not generate code which uses features specific to the AM33 processor. This
12081 @item -mreturn-pointer-on-d0
12082 @opindex mreturn-pointer-on-d0
12083 When generating a function which returns a pointer, return the pointer
12084 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
12085 only in a0, and attempts to call such functions without a prototype
12086 would result in errors. Note that this option is on by default; use
12087 @option{-mno-return-pointer-on-d0} to disable it.
12091 Do not link in the C run-time initialization object file.
12095 Indicate to the linker that it should perform a relaxation optimization pass
12096 to shorten branches, calls and absolute memory addresses. This option only
12097 has an effect when used on the command line for the final link step.
12099 This option makes symbolic debugging impossible.
12103 @subsection MT Options
12106 These @option{-m} options are defined for Morpho MT architectures:
12110 @item -march=@var{cpu-type}
12112 Generate code that will run on @var{cpu-type}, which is the name of a system
12113 representing a certain processor type. Possible values for
12114 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
12115 @samp{ms1-16-003} and @samp{ms2}.
12117 When this option is not used, the default is @option{-march=ms1-16-002}.
12121 Use byte loads and stores when generating code.
12125 Do not use byte loads and stores when generating code.
12129 Use simulator runtime
12133 Do not link in the C run-time initialization object file
12134 @file{crti.o}. Other run-time initialization and termination files
12135 such as @file{startup.o} and @file{exit.o} are still included on the
12136 linker command line.
12140 @node PDP-11 Options
12141 @subsection PDP-11 Options
12142 @cindex PDP-11 Options
12144 These options are defined for the PDP-11:
12149 Use hardware FPP floating point. This is the default. (FIS floating
12150 point on the PDP-11/40 is not supported.)
12153 @opindex msoft-float
12154 Do not use hardware floating point.
12158 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
12162 Return floating-point results in memory. This is the default.
12166 Generate code for a PDP-11/40.
12170 Generate code for a PDP-11/45. This is the default.
12174 Generate code for a PDP-11/10.
12176 @item -mbcopy-builtin
12177 @opindex bcopy-builtin
12178 Use inline @code{movmemhi} patterns for copying memory. This is the
12183 Do not use inline @code{movmemhi} patterns for copying memory.
12189 Use 16-bit @code{int}. This is the default.
12195 Use 32-bit @code{int}.
12198 @itemx -mno-float32
12200 @opindex mno-float32
12201 Use 64-bit @code{float}. This is the default.
12204 @itemx -mno-float64
12206 @opindex mno-float64
12207 Use 32-bit @code{float}.
12211 Use @code{abshi2} pattern. This is the default.
12215 Do not use @code{abshi2} pattern.
12217 @item -mbranch-expensive
12218 @opindex mbranch-expensive
12219 Pretend that branches are expensive. This is for experimenting with
12220 code generation only.
12222 @item -mbranch-cheap
12223 @opindex mbranch-cheap
12224 Do not pretend that branches are expensive. This is the default.
12228 Generate code for a system with split I&D@.
12232 Generate code for a system without split I&D@. This is the default.
12236 Use Unix assembler syntax. This is the default when configured for
12237 @samp{pdp11-*-bsd}.
12241 Use DEC assembler syntax. This is the default when configured for any
12242 PDP-11 target other than @samp{pdp11-*-bsd}.
12245 @node PowerPC Options
12246 @subsection PowerPC Options
12247 @cindex PowerPC options
12249 These are listed under @xref{RS/6000 and PowerPC Options}.
12251 @node RS/6000 and PowerPC Options
12252 @subsection IBM RS/6000 and PowerPC Options
12253 @cindex RS/6000 and PowerPC Options
12254 @cindex IBM RS/6000 and PowerPC Options
12256 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
12263 @itemx -mno-powerpc
12264 @itemx -mpowerpc-gpopt
12265 @itemx -mno-powerpc-gpopt
12266 @itemx -mpowerpc-gfxopt
12267 @itemx -mno-powerpc-gfxopt
12269 @itemx -mno-powerpc64
12273 @itemx -mno-popcntb
12285 @opindex mno-power2
12287 @opindex mno-powerpc
12288 @opindex mpowerpc-gpopt
12289 @opindex mno-powerpc-gpopt
12290 @opindex mpowerpc-gfxopt
12291 @opindex mno-powerpc-gfxopt
12292 @opindex mpowerpc64
12293 @opindex mno-powerpc64
12297 @opindex mno-popcntb
12303 @opindex mno-mfpgpr
12306 GCC supports two related instruction set architectures for the
12307 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
12308 instructions supported by the @samp{rios} chip set used in the original
12309 RS/6000 systems and the @dfn{PowerPC} instruction set is the
12310 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
12311 the IBM 4xx, 6xx, and follow-on microprocessors.
12313 Neither architecture is a subset of the other. However there is a
12314 large common subset of instructions supported by both. An MQ
12315 register is included in processors supporting the POWER architecture.
12317 You use these options to specify which instructions are available on the
12318 processor you are using. The default value of these options is
12319 determined when configuring GCC@. Specifying the
12320 @option{-mcpu=@var{cpu_type}} overrides the specification of these
12321 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
12322 rather than the options listed above.
12324 The @option{-mpower} option allows GCC to generate instructions that
12325 are found only in the POWER architecture and to use the MQ register.
12326 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
12327 to generate instructions that are present in the POWER2 architecture but
12328 not the original POWER architecture.
12330 The @option{-mpowerpc} option allows GCC to generate instructions that
12331 are found only in the 32-bit subset of the PowerPC architecture.
12332 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
12333 GCC to use the optional PowerPC architecture instructions in the
12334 General Purpose group, including floating-point square root. Specifying
12335 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
12336 use the optional PowerPC architecture instructions in the Graphics
12337 group, including floating-point select.
12339 The @option{-mmfcrf} option allows GCC to generate the move from
12340 condition register field instruction implemented on the POWER4
12341 processor and other processors that support the PowerPC V2.01
12343 The @option{-mpopcntb} option allows GCC to generate the popcount and
12344 double precision FP reciprocal estimate instruction implemented on the
12345 POWER5 processor and other processors that support the PowerPC V2.02
12347 The @option{-mfprnd} option allows GCC to generate the FP round to
12348 integer instructions implemented on the POWER5+ processor and other
12349 processors that support the PowerPC V2.03 architecture.
12350 The @option{-mcmpb} option allows GCC to generate the compare bytes
12351 instruction implemented on the POWER6 processor and other processors
12352 that support the PowerPC V2.05 architecture.
12353 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
12354 general purpose register instructions implemented on the POWER6X
12355 processor and other processors that support the extended PowerPC V2.05
12357 The @option{-mdfp} option allows GCC to generate the decimal floating
12358 point instructions implemented on some POWER processors.
12360 The @option{-mpowerpc64} option allows GCC to generate the additional
12361 64-bit instructions that are found in the full PowerPC64 architecture
12362 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
12363 @option{-mno-powerpc64}.
12365 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
12366 will use only the instructions in the common subset of both
12367 architectures plus some special AIX common-mode calls, and will not use
12368 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
12369 permits GCC to use any instruction from either architecture and to
12370 allow use of the MQ register; specify this for the Motorola MPC601.
12372 @item -mnew-mnemonics
12373 @itemx -mold-mnemonics
12374 @opindex mnew-mnemonics
12375 @opindex mold-mnemonics
12376 Select which mnemonics to use in the generated assembler code. With
12377 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
12378 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
12379 assembler mnemonics defined for the POWER architecture. Instructions
12380 defined in only one architecture have only one mnemonic; GCC uses that
12381 mnemonic irrespective of which of these options is specified.
12383 GCC defaults to the mnemonics appropriate for the architecture in
12384 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
12385 value of these option. Unless you are building a cross-compiler, you
12386 should normally not specify either @option{-mnew-mnemonics} or
12387 @option{-mold-mnemonics}, but should instead accept the default.
12389 @item -mcpu=@var{cpu_type}
12391 Set architecture type, register usage, choice of mnemonics, and
12392 instruction scheduling parameters for machine type @var{cpu_type}.
12393 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
12394 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
12395 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
12396 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
12397 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
12398 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
12399 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
12400 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
12401 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
12402 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
12404 @option{-mcpu=common} selects a completely generic processor. Code
12405 generated under this option will run on any POWER or PowerPC processor.
12406 GCC will use only the instructions in the common subset of both
12407 architectures, and will not use the MQ register. GCC assumes a generic
12408 processor model for scheduling purposes.
12410 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
12411 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
12412 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
12413 types, with an appropriate, generic processor model assumed for
12414 scheduling purposes.
12416 The other options specify a specific processor. Code generated under
12417 those options will run best on that processor, and may not run at all on
12420 The @option{-mcpu} options automatically enable or disable the
12423 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
12424 -mnew-mnemonics -mpopcntb -mpower -mpower2 -mpowerpc64 @gol
12425 -mpowerpc-gpopt -mpowerpc-gfxopt -mstring -mmulhw -mdlmzb -mmfpgpr}
12427 The particular options set for any particular CPU will vary between
12428 compiler versions, depending on what setting seems to produce optimal
12429 code for that CPU; it doesn't necessarily reflect the actual hardware's
12430 capabilities. If you wish to set an individual option to a particular
12431 value, you may specify it after the @option{-mcpu} option, like
12432 @samp{-mcpu=970 -mno-altivec}.
12434 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
12435 not enabled or disabled by the @option{-mcpu} option at present because
12436 AIX does not have full support for these options. You may still
12437 enable or disable them individually if you're sure it'll work in your
12440 @item -mtune=@var{cpu_type}
12442 Set the instruction scheduling parameters for machine type
12443 @var{cpu_type}, but do not set the architecture type, register usage, or
12444 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
12445 values for @var{cpu_type} are used for @option{-mtune} as for
12446 @option{-mcpu}. If both are specified, the code generated will use the
12447 architecture, registers, and mnemonics set by @option{-mcpu}, but the
12448 scheduling parameters set by @option{-mtune}.
12454 Generate code to compute division as reciprocal estimate and iterative
12455 refinement, creating opportunities for increased throughput. This
12456 feature requires: optional PowerPC Graphics instruction set for single
12457 precision and FRE instruction for double precision, assuming divides
12458 cannot generate user-visible traps, and the domain values not include
12459 Infinities, denormals or zero denominator.
12462 @itemx -mno-altivec
12464 @opindex mno-altivec
12465 Generate code that uses (does not use) AltiVec instructions, and also
12466 enable the use of built-in functions that allow more direct access to
12467 the AltiVec instruction set. You may also need to set
12468 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
12474 @opindex mno-vrsave
12475 Generate VRSAVE instructions when generating AltiVec code.
12478 @opindex msecure-plt
12479 Generate code that allows ld and ld.so to build executables and shared
12480 libraries with non-exec .plt and .got sections. This is a PowerPC
12481 32-bit SYSV ABI option.
12485 Generate code that uses a BSS .plt section that ld.so fills in, and
12486 requires .plt and .got sections that are both writable and executable.
12487 This is a PowerPC 32-bit SYSV ABI option.
12493 This switch enables or disables the generation of ISEL instructions.
12495 @item -misel=@var{yes/no}
12496 This switch has been deprecated. Use @option{-misel} and
12497 @option{-mno-isel} instead.
12503 This switch enables or disables the generation of SPE simd
12506 @item -mspe=@var{yes/no}
12507 This option has been deprecated. Use @option{-mspe} and
12508 @option{-mno-spe} instead.
12510 @item -mfloat-gprs=@var{yes/single/double/no}
12511 @itemx -mfloat-gprs
12512 @opindex mfloat-gprs
12513 This switch enables or disables the generation of floating point
12514 operations on the general purpose registers for architectures that
12517 The argument @var{yes} or @var{single} enables the use of
12518 single-precision floating point operations.
12520 The argument @var{double} enables the use of single and
12521 double-precision floating point operations.
12523 The argument @var{no} disables floating point operations on the
12524 general purpose registers.
12526 This option is currently only available on the MPC854x.
12532 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
12533 targets (including GNU/Linux). The 32-bit environment sets int, long
12534 and pointer to 32 bits and generates code that runs on any PowerPC
12535 variant. The 64-bit environment sets int to 32 bits and long and
12536 pointer to 64 bits, and generates code for PowerPC64, as for
12537 @option{-mpowerpc64}.
12540 @itemx -mno-fp-in-toc
12541 @itemx -mno-sum-in-toc
12542 @itemx -mminimal-toc
12544 @opindex mno-fp-in-toc
12545 @opindex mno-sum-in-toc
12546 @opindex mminimal-toc
12547 Modify generation of the TOC (Table Of Contents), which is created for
12548 every executable file. The @option{-mfull-toc} option is selected by
12549 default. In that case, GCC will allocate at least one TOC entry for
12550 each unique non-automatic variable reference in your program. GCC
12551 will also place floating-point constants in the TOC@. However, only
12552 16,384 entries are available in the TOC@.
12554 If you receive a linker error message that saying you have overflowed
12555 the available TOC space, you can reduce the amount of TOC space used
12556 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
12557 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
12558 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
12559 generate code to calculate the sum of an address and a constant at
12560 run-time instead of putting that sum into the TOC@. You may specify one
12561 or both of these options. Each causes GCC to produce very slightly
12562 slower and larger code at the expense of conserving TOC space.
12564 If you still run out of space in the TOC even when you specify both of
12565 these options, specify @option{-mminimal-toc} instead. This option causes
12566 GCC to make only one TOC entry for every file. When you specify this
12567 option, GCC will produce code that is slower and larger but which
12568 uses extremely little TOC space. You may wish to use this option
12569 only on files that contain less frequently executed code.
12575 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
12576 @code{long} type, and the infrastructure needed to support them.
12577 Specifying @option{-maix64} implies @option{-mpowerpc64} and
12578 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
12579 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
12582 @itemx -mno-xl-compat
12583 @opindex mxl-compat
12584 @opindex mno-xl-compat
12585 Produce code that conforms more closely to IBM XL compiler semantics
12586 when using AIX-compatible ABI. Pass floating-point arguments to
12587 prototyped functions beyond the register save area (RSA) on the stack
12588 in addition to argument FPRs. Do not assume that most significant
12589 double in 128-bit long double value is properly rounded when comparing
12590 values and converting to double. Use XL symbol names for long double
12593 The AIX calling convention was extended but not initially documented to
12594 handle an obscure K&R C case of calling a function that takes the
12595 address of its arguments with fewer arguments than declared. IBM XL
12596 compilers access floating point arguments which do not fit in the
12597 RSA from the stack when a subroutine is compiled without
12598 optimization. Because always storing floating-point arguments on the
12599 stack is inefficient and rarely needed, this option is not enabled by
12600 default and only is necessary when calling subroutines compiled by IBM
12601 XL compilers without optimization.
12605 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
12606 application written to use message passing with special startup code to
12607 enable the application to run. The system must have PE installed in the
12608 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
12609 must be overridden with the @option{-specs=} option to specify the
12610 appropriate directory location. The Parallel Environment does not
12611 support threads, so the @option{-mpe} option and the @option{-pthread}
12612 option are incompatible.
12614 @item -malign-natural
12615 @itemx -malign-power
12616 @opindex malign-natural
12617 @opindex malign-power
12618 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
12619 @option{-malign-natural} overrides the ABI-defined alignment of larger
12620 types, such as floating-point doubles, on their natural size-based boundary.
12621 The option @option{-malign-power} instructs GCC to follow the ABI-specified
12622 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
12624 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
12628 @itemx -mhard-float
12629 @opindex msoft-float
12630 @opindex mhard-float
12631 Generate code that does not use (uses) the floating-point register set.
12632 Software floating point emulation is provided if you use the
12633 @option{-msoft-float} option, and pass the option to GCC when linking.
12636 @itemx -mno-multiple
12638 @opindex mno-multiple
12639 Generate code that uses (does not use) the load multiple word
12640 instructions and the store multiple word instructions. These
12641 instructions are generated by default on POWER systems, and not
12642 generated on PowerPC systems. Do not use @option{-mmultiple} on little
12643 endian PowerPC systems, since those instructions do not work when the
12644 processor is in little endian mode. The exceptions are PPC740 and
12645 PPC750 which permit the instructions usage in little endian mode.
12650 @opindex mno-string
12651 Generate code that uses (does not use) the load string instructions
12652 and the store string word instructions to save multiple registers and
12653 do small block moves. These instructions are generated by default on
12654 POWER systems, and not generated on PowerPC systems. Do not use
12655 @option{-mstring} on little endian PowerPC systems, since those
12656 instructions do not work when the processor is in little endian mode.
12657 The exceptions are PPC740 and PPC750 which permit the instructions
12658 usage in little endian mode.
12663 @opindex mno-update
12664 Generate code that uses (does not use) the load or store instructions
12665 that update the base register to the address of the calculated memory
12666 location. These instructions are generated by default. If you use
12667 @option{-mno-update}, there is a small window between the time that the
12668 stack pointer is updated and the address of the previous frame is
12669 stored, which means code that walks the stack frame across interrupts or
12670 signals may get corrupted data.
12673 @itemx -mno-fused-madd
12674 @opindex mfused-madd
12675 @opindex mno-fused-madd
12676 Generate code that uses (does not use) the floating point multiply and
12677 accumulate instructions. These instructions are generated by default if
12678 hardware floating is used.
12684 Generate code that uses (does not use) the half-word multiply and
12685 multiply-accumulate instructions on the IBM 405 and 440 processors.
12686 These instructions are generated by default when targetting those
12693 Generate code that uses (does not use) the string-search @samp{dlmzb}
12694 instruction on the IBM 405 and 440 processors. This instruction is
12695 generated by default when targetting those processors.
12697 @item -mno-bit-align
12699 @opindex mno-bit-align
12700 @opindex mbit-align
12701 On System V.4 and embedded PowerPC systems do not (do) force structures
12702 and unions that contain bit-fields to be aligned to the base type of the
12705 For example, by default a structure containing nothing but 8
12706 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
12707 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
12708 the structure would be aligned to a 1 byte boundary and be one byte in
12711 @item -mno-strict-align
12712 @itemx -mstrict-align
12713 @opindex mno-strict-align
12714 @opindex mstrict-align
12715 On System V.4 and embedded PowerPC systems do not (do) assume that
12716 unaligned memory references will be handled by the system.
12718 @item -mrelocatable
12719 @itemx -mno-relocatable
12720 @opindex mrelocatable
12721 @opindex mno-relocatable
12722 On embedded PowerPC systems generate code that allows (does not allow)
12723 the program to be relocated to a different address at runtime. If you
12724 use @option{-mrelocatable} on any module, all objects linked together must
12725 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
12727 @item -mrelocatable-lib
12728 @itemx -mno-relocatable-lib
12729 @opindex mrelocatable-lib
12730 @opindex mno-relocatable-lib
12731 On embedded PowerPC systems generate code that allows (does not allow)
12732 the program to be relocated to a different address at runtime. Modules
12733 compiled with @option{-mrelocatable-lib} can be linked with either modules
12734 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
12735 with modules compiled with the @option{-mrelocatable} options.
12741 On System V.4 and embedded PowerPC systems do not (do) assume that
12742 register 2 contains a pointer to a global area pointing to the addresses
12743 used in the program.
12746 @itemx -mlittle-endian
12748 @opindex mlittle-endian
12749 On System V.4 and embedded PowerPC systems compile code for the
12750 processor in little endian mode. The @option{-mlittle-endian} option is
12751 the same as @option{-mlittle}.
12754 @itemx -mbig-endian
12756 @opindex mbig-endian
12757 On System V.4 and embedded PowerPC systems compile code for the
12758 processor in big endian mode. The @option{-mbig-endian} option is
12759 the same as @option{-mbig}.
12761 @item -mdynamic-no-pic
12762 @opindex mdynamic-no-pic
12763 On Darwin and Mac OS X systems, compile code so that it is not
12764 relocatable, but that its external references are relocatable. The
12765 resulting code is suitable for applications, but not shared
12768 @item -mprioritize-restricted-insns=@var{priority}
12769 @opindex mprioritize-restricted-insns
12770 This option controls the priority that is assigned to
12771 dispatch-slot restricted instructions during the second scheduling
12772 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
12773 @var{no/highest/second-highest} priority to dispatch slot restricted
12776 @item -msched-costly-dep=@var{dependence_type}
12777 @opindex msched-costly-dep
12778 This option controls which dependences are considered costly
12779 by the target during instruction scheduling. The argument
12780 @var{dependence_type} takes one of the following values:
12781 @var{no}: no dependence is costly,
12782 @var{all}: all dependences are costly,
12783 @var{true_store_to_load}: a true dependence from store to load is costly,
12784 @var{store_to_load}: any dependence from store to load is costly,
12785 @var{number}: any dependence which latency >= @var{number} is costly.
12787 @item -minsert-sched-nops=@var{scheme}
12788 @opindex minsert-sched-nops
12789 This option controls which nop insertion scheme will be used during
12790 the second scheduling pass. The argument @var{scheme} takes one of the
12792 @var{no}: Don't insert nops.
12793 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
12794 according to the scheduler's grouping.
12795 @var{regroup_exact}: Insert nops to force costly dependent insns into
12796 separate groups. Insert exactly as many nops as needed to force an insn
12797 to a new group, according to the estimated processor grouping.
12798 @var{number}: Insert nops to force costly dependent insns into
12799 separate groups. Insert @var{number} nops to force an insn to a new group.
12802 @opindex mcall-sysv
12803 On System V.4 and embedded PowerPC systems compile code using calling
12804 conventions that adheres to the March 1995 draft of the System V
12805 Application Binary Interface, PowerPC processor supplement. This is the
12806 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
12808 @item -mcall-sysv-eabi
12809 @opindex mcall-sysv-eabi
12810 Specify both @option{-mcall-sysv} and @option{-meabi} options.
12812 @item -mcall-sysv-noeabi
12813 @opindex mcall-sysv-noeabi
12814 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
12816 @item -mcall-solaris
12817 @opindex mcall-solaris
12818 On System V.4 and embedded PowerPC systems compile code for the Solaris
12822 @opindex mcall-linux
12823 On System V.4 and embedded PowerPC systems compile code for the
12824 Linux-based GNU system.
12828 On System V.4 and embedded PowerPC systems compile code for the
12829 Hurd-based GNU system.
12831 @item -mcall-netbsd
12832 @opindex mcall-netbsd
12833 On System V.4 and embedded PowerPC systems compile code for the
12834 NetBSD operating system.
12836 @item -maix-struct-return
12837 @opindex maix-struct-return
12838 Return all structures in memory (as specified by the AIX ABI)@.
12840 @item -msvr4-struct-return
12841 @opindex msvr4-struct-return
12842 Return structures smaller than 8 bytes in registers (as specified by the
12845 @item -mabi=@var{abi-type}
12847 Extend the current ABI with a particular extension, or remove such extension.
12848 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
12849 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
12853 Extend the current ABI with SPE ABI extensions. This does not change
12854 the default ABI, instead it adds the SPE ABI extensions to the current
12858 @opindex mabi=no-spe
12859 Disable Booke SPE ABI extensions for the current ABI@.
12861 @item -mabi=ibmlongdouble
12862 @opindex mabi=ibmlongdouble
12863 Change the current ABI to use IBM extended precision long double.
12864 This is a PowerPC 32-bit SYSV ABI option.
12866 @item -mabi=ieeelongdouble
12867 @opindex mabi=ieeelongdouble
12868 Change the current ABI to use IEEE extended precision long double.
12869 This is a PowerPC 32-bit Linux ABI option.
12872 @itemx -mno-prototype
12873 @opindex mprototype
12874 @opindex mno-prototype
12875 On System V.4 and embedded PowerPC systems assume that all calls to
12876 variable argument functions are properly prototyped. Otherwise, the
12877 compiler must insert an instruction before every non prototyped call to
12878 set or clear bit 6 of the condition code register (@var{CR}) to
12879 indicate whether floating point values were passed in the floating point
12880 registers in case the function takes a variable arguments. With
12881 @option{-mprototype}, only calls to prototyped variable argument functions
12882 will set or clear the bit.
12886 On embedded PowerPC systems, assume that the startup module is called
12887 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
12888 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
12893 On embedded PowerPC systems, assume that the startup module is called
12894 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
12899 On embedded PowerPC systems, assume that the startup module is called
12900 @file{crt0.o} and the standard C libraries are @file{libads.a} and
12903 @item -myellowknife
12904 @opindex myellowknife
12905 On embedded PowerPC systems, assume that the startup module is called
12906 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
12911 On System V.4 and embedded PowerPC systems, specify that you are
12912 compiling for a VxWorks system.
12916 Specify that you are compiling for the WindISS simulation environment.
12920 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
12921 header to indicate that @samp{eabi} extended relocations are used.
12927 On System V.4 and embedded PowerPC systems do (do not) adhere to the
12928 Embedded Applications Binary Interface (eabi) which is a set of
12929 modifications to the System V.4 specifications. Selecting @option{-meabi}
12930 means that the stack is aligned to an 8 byte boundary, a function
12931 @code{__eabi} is called to from @code{main} to set up the eabi
12932 environment, and the @option{-msdata} option can use both @code{r2} and
12933 @code{r13} to point to two separate small data areas. Selecting
12934 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
12935 do not call an initialization function from @code{main}, and the
12936 @option{-msdata} option will only use @code{r13} to point to a single
12937 small data area. The @option{-meabi} option is on by default if you
12938 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
12941 @opindex msdata=eabi
12942 On System V.4 and embedded PowerPC systems, put small initialized
12943 @code{const} global and static data in the @samp{.sdata2} section, which
12944 is pointed to by register @code{r2}. Put small initialized
12945 non-@code{const} global and static data in the @samp{.sdata} section,
12946 which is pointed to by register @code{r13}. Put small uninitialized
12947 global and static data in the @samp{.sbss} section, which is adjacent to
12948 the @samp{.sdata} section. The @option{-msdata=eabi} option is
12949 incompatible with the @option{-mrelocatable} option. The
12950 @option{-msdata=eabi} option also sets the @option{-memb} option.
12953 @opindex msdata=sysv
12954 On System V.4 and embedded PowerPC systems, put small global and static
12955 data in the @samp{.sdata} section, which is pointed to by register
12956 @code{r13}. Put small uninitialized global and static data in the
12957 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
12958 The @option{-msdata=sysv} option is incompatible with the
12959 @option{-mrelocatable} option.
12961 @item -msdata=default
12963 @opindex msdata=default
12965 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12966 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12967 same as @option{-msdata=sysv}.
12970 @opindex msdata-data
12971 On System V.4 and embedded PowerPC systems, put small global
12972 data in the @samp{.sdata} section. Put small uninitialized global
12973 data in the @samp{.sbss} section. Do not use register @code{r13}
12974 to address small data however. This is the default behavior unless
12975 other @option{-msdata} options are used.
12979 @opindex msdata=none
12981 On embedded PowerPC systems, put all initialized global and static data
12982 in the @samp{.data} section, and all uninitialized data in the
12983 @samp{.bss} section.
12987 @cindex smaller data references (PowerPC)
12988 @cindex .sdata/.sdata2 references (PowerPC)
12989 On embedded PowerPC systems, put global and static items less than or
12990 equal to @var{num} bytes into the small data or bss sections instead of
12991 the normal data or bss section. By default, @var{num} is 8. The
12992 @option{-G @var{num}} switch is also passed to the linker.
12993 All modules should be compiled with the same @option{-G @var{num}} value.
12996 @itemx -mno-regnames
12998 @opindex mno-regnames
12999 On System V.4 and embedded PowerPC systems do (do not) emit register
13000 names in the assembly language output using symbolic forms.
13003 @itemx -mno-longcall
13005 @opindex mno-longcall
13006 By default assume that all calls are far away so that a longer more
13007 expensive calling sequence is required. This is required for calls
13008 further than 32 megabytes (33,554,432 bytes) from the current location.
13009 A short call will be generated if the compiler knows
13010 the call cannot be that far away. This setting can be overridden by
13011 the @code{shortcall} function attribute, or by @code{#pragma
13014 Some linkers are capable of detecting out-of-range calls and generating
13015 glue code on the fly. On these systems, long calls are unnecessary and
13016 generate slower code. As of this writing, the AIX linker can do this,
13017 as can the GNU linker for PowerPC/64. It is planned to add this feature
13018 to the GNU linker for 32-bit PowerPC systems as well.
13020 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
13021 callee, L42'', plus a ``branch island'' (glue code). The two target
13022 addresses represent the callee and the ``branch island''. The
13023 Darwin/PPC linker will prefer the first address and generate a ``bl
13024 callee'' if the PPC ``bl'' instruction will reach the callee directly;
13025 otherwise, the linker will generate ``bl L42'' to call the ``branch
13026 island''. The ``branch island'' is appended to the body of the
13027 calling function; it computes the full 32-bit address of the callee
13030 On Mach-O (Darwin) systems, this option directs the compiler emit to
13031 the glue for every direct call, and the Darwin linker decides whether
13032 to use or discard it.
13034 In the future, we may cause GCC to ignore all longcall specifications
13035 when the linker is known to generate glue.
13039 Adds support for multithreading with the @dfn{pthreads} library.
13040 This option sets flags for both the preprocessor and linker.
13044 @node S/390 and zSeries Options
13045 @subsection S/390 and zSeries Options
13046 @cindex S/390 and zSeries Options
13048 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
13052 @itemx -msoft-float
13053 @opindex mhard-float
13054 @opindex msoft-float
13055 Use (do not use) the hardware floating-point instructions and registers
13056 for floating-point operations. When @option{-msoft-float} is specified,
13057 functions in @file{libgcc.a} will be used to perform floating-point
13058 operations. When @option{-mhard-float} is specified, the compiler
13059 generates IEEE floating-point instructions. This is the default.
13061 @item -mlong-double-64
13062 @itemx -mlong-double-128
13063 @opindex mlong-double-64
13064 @opindex mlong-double-128
13065 These switches control the size of @code{long double} type. A size
13066 of 64bit makes the @code{long double} type equivalent to the @code{double}
13067 type. This is the default.
13070 @itemx -mno-backchain
13071 @opindex mbackchain
13072 @opindex mno-backchain
13073 Store (do not store) the address of the caller's frame as backchain pointer
13074 into the callee's stack frame.
13075 A backchain may be needed to allow debugging using tools that do not understand
13076 DWARF-2 call frame information.
13077 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
13078 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
13079 the backchain is placed into the topmost word of the 96/160 byte register
13082 In general, code compiled with @option{-mbackchain} is call-compatible with
13083 code compiled with @option{-mmo-backchain}; however, use of the backchain
13084 for debugging purposes usually requires that the whole binary is built with
13085 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
13086 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13087 to build a linux kernel use @option{-msoft-float}.
13089 The default is to not maintain the backchain.
13091 @item -mpacked-stack
13092 @item -mno-packed-stack
13093 @opindex mpacked-stack
13094 @opindex mno-packed-stack
13095 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
13096 specified, the compiler uses the all fields of the 96/160 byte register save
13097 area only for their default purpose; unused fields still take up stack space.
13098 When @option{-mpacked-stack} is specified, register save slots are densely
13099 packed at the top of the register save area; unused space is reused for other
13100 purposes, allowing for more efficient use of the available stack space.
13101 However, when @option{-mbackchain} is also in effect, the topmost word of
13102 the save area is always used to store the backchain, and the return address
13103 register is always saved two words below the backchain.
13105 As long as the stack frame backchain is not used, code generated with
13106 @option{-mpacked-stack} is call-compatible with code generated with
13107 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
13108 S/390 or zSeries generated code that uses the stack frame backchain at run
13109 time, not just for debugging purposes. Such code is not call-compatible
13110 with code compiled with @option{-mpacked-stack}. Also, note that the
13111 combination of @option{-mbackchain},
13112 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
13113 to build a linux kernel use @option{-msoft-float}.
13115 The default is to not use the packed stack layout.
13118 @itemx -mno-small-exec
13119 @opindex msmall-exec
13120 @opindex mno-small-exec
13121 Generate (or do not generate) code using the @code{bras} instruction
13122 to do subroutine calls.
13123 This only works reliably if the total executable size does not
13124 exceed 64k. The default is to use the @code{basr} instruction instead,
13125 which does not have this limitation.
13131 When @option{-m31} is specified, generate code compliant to the
13132 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
13133 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
13134 particular to generate 64-bit instructions. For the @samp{s390}
13135 targets, the default is @option{-m31}, while the @samp{s390x}
13136 targets default to @option{-m64}.
13142 When @option{-mzarch} is specified, generate code using the
13143 instructions available on z/Architecture.
13144 When @option{-mesa} is specified, generate code using the
13145 instructions available on ESA/390. Note that @option{-mesa} is
13146 not possible with @option{-m64}.
13147 When generating code compliant to the GNU/Linux for S/390 ABI,
13148 the default is @option{-mesa}. When generating code compliant
13149 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
13155 Generate (or do not generate) code using the @code{mvcle} instruction
13156 to perform block moves. When @option{-mno-mvcle} is specified,
13157 use a @code{mvc} loop instead. This is the default unless optimizing for
13164 Print (or do not print) additional debug information when compiling.
13165 The default is to not print debug information.
13167 @item -march=@var{cpu-type}
13169 Generate code that will run on @var{cpu-type}, which is the name of a system
13170 representing a certain processor type. Possible values for
13171 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
13172 When generating code using the instructions available on z/Architecture,
13173 the default is @option{-march=z900}. Otherwise, the default is
13174 @option{-march=g5}.
13176 @item -mtune=@var{cpu-type}
13178 Tune to @var{cpu-type} everything applicable about the generated code,
13179 except for the ABI and the set of available instructions.
13180 The list of @var{cpu-type} values is the same as for @option{-march}.
13181 The default is the value used for @option{-march}.
13184 @itemx -mno-tpf-trace
13185 @opindex mtpf-trace
13186 @opindex mno-tpf-trace
13187 Generate code that adds (does not add) in TPF OS specific branches to trace
13188 routines in the operating system. This option is off by default, even
13189 when compiling for the TPF OS@.
13192 @itemx -mno-fused-madd
13193 @opindex mfused-madd
13194 @opindex mno-fused-madd
13195 Generate code that uses (does not use) the floating point multiply and
13196 accumulate instructions. These instructions are generated by default if
13197 hardware floating point is used.
13199 @item -mwarn-framesize=@var{framesize}
13200 @opindex mwarn-framesize
13201 Emit a warning if the current function exceeds the given frame size. Because
13202 this is a compile time check it doesn't need to be a real problem when the program
13203 runs. It is intended to identify functions which most probably cause
13204 a stack overflow. It is useful to be used in an environment with limited stack
13205 size e.g.@: the linux kernel.
13207 @item -mwarn-dynamicstack
13208 @opindex mwarn-dynamicstack
13209 Emit a warning if the function calls alloca or uses dynamically
13210 sized arrays. This is generally a bad idea with a limited stack size.
13212 @item -mstack-guard=@var{stack-guard}
13213 @item -mstack-size=@var{stack-size}
13214 @opindex mstack-guard
13215 @opindex mstack-size
13216 If these options are provided the s390 back end emits additional instructions in
13217 the function prologue which trigger a trap if the stack size is @var{stack-guard}
13218 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
13219 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
13220 the frame size of the compiled function is chosen.
13221 These options are intended to be used to help debugging stack overflow problems.
13222 The additionally emitted code causes only little overhead and hence can also be
13223 used in production like systems without greater performance degradation. The given
13224 values have to be exact powers of 2 and @var{stack-size} has to be greater than
13225 @var{stack-guard} without exceeding 64k.
13226 In order to be efficient the extra code makes the assumption that the stack starts
13227 at an address aligned to the value given by @var{stack-size}.
13228 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
13231 @node Score Options
13232 @subsection Score Options
13233 @cindex Score Options
13235 These options are defined for Score implementations:
13240 Compile code for big endian mode. This is the default.
13244 Compile code for little endian mode.
13248 Disable generate bcnz instruction.
13252 Enable generate unaligned load and store instruction.
13256 Enable the use of multiply-accumulate instructions. Disabled by default.
13260 Specify the SCORE5 as the target architecture.
13264 Specify the SCORE5U of the target architecture.
13268 Specify the SCORE7 as the target architecture. This is the default.
13272 Specify the SCORE7D as the target architecture.
13276 @subsection SH Options
13278 These @samp{-m} options are defined for the SH implementations:
13283 Generate code for the SH1.
13287 Generate code for the SH2.
13290 Generate code for the SH2e.
13294 Generate code for the SH3.
13298 Generate code for the SH3e.
13302 Generate code for the SH4 without a floating-point unit.
13304 @item -m4-single-only
13305 @opindex m4-single-only
13306 Generate code for the SH4 with a floating-point unit that only
13307 supports single-precision arithmetic.
13311 Generate code for the SH4 assuming the floating-point unit is in
13312 single-precision mode by default.
13316 Generate code for the SH4.
13320 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
13321 floating-point unit is not used.
13323 @item -m4a-single-only
13324 @opindex m4a-single-only
13325 Generate code for the SH4a, in such a way that no double-precision
13326 floating point operations are used.
13329 @opindex m4a-single
13330 Generate code for the SH4a assuming the floating-point unit is in
13331 single-precision mode by default.
13335 Generate code for the SH4a.
13339 Same as @option{-m4a-nofpu}, except that it implicitly passes
13340 @option{-dsp} to the assembler. GCC doesn't generate any DSP
13341 instructions at the moment.
13345 Compile code for the processor in big endian mode.
13349 Compile code for the processor in little endian mode.
13353 Align doubles at 64-bit boundaries. Note that this changes the calling
13354 conventions, and thus some functions from the standard C library will
13355 not work unless you recompile it first with @option{-mdalign}.
13359 Shorten some address references at link time, when possible; uses the
13360 linker option @option{-relax}.
13364 Use 32-bit offsets in @code{switch} tables. The default is to use
13369 Enable the use of the instruction @code{fmovd}.
13373 Comply with the calling conventions defined by Renesas.
13377 Comply with the calling conventions defined by Renesas.
13381 Comply with the calling conventions defined for GCC before the Renesas
13382 conventions were available. This option is the default for all
13383 targets of the SH toolchain except for @samp{sh-symbianelf}.
13386 @opindex mnomacsave
13387 Mark the @code{MAC} register as call-clobbered, even if
13388 @option{-mhitachi} is given.
13392 Increase IEEE-compliance of floating-point code.
13393 At the moment, this is equivalent to @option{-fno-finite-math-only}.
13394 When generating 16 bit SH opcodes, getting IEEE-conforming results for
13395 comparisons of NANs / infinities incurs extra overhead in every
13396 floating point comparison, therefore the default is set to
13397 @option{-ffinite-math-only}.
13399 @item -minline-ic_invalidate
13400 @opindex minline-ic_invalidate
13401 Inline code to invalidate instruction cache entries after setting up
13402 nested function trampolines.
13403 This option has no effect if -musermode is in effect and the selected
13404 code generation option (e.g. -m4) does not allow the use of the icbi
13406 If the selected code generation option does not allow the use of the icbi
13407 instruction, and -musermode is not in effect, the inlined code will
13408 manipulate the instruction cache address array directly with an associative
13409 write. This not only requires privileged mode, but it will also
13410 fail if the cache line had been mapped via the TLB and has become unmapped.
13414 Dump instruction size and location in the assembly code.
13417 @opindex mpadstruct
13418 This option is deprecated. It pads structures to multiple of 4 bytes,
13419 which is incompatible with the SH ABI@.
13423 Optimize for space instead of speed. Implied by @option{-Os}.
13426 @opindex mprefergot
13427 When generating position-independent code, emit function calls using
13428 the Global Offset Table instead of the Procedure Linkage Table.
13432 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
13433 if the inlined code would not work in user mode.
13434 This is the default when the target is @code{sh-*-linux*}.
13436 @item -multcost=@var{number}
13437 @opindex multcost=@var{number}
13438 Set the cost to assume for a multiply insn.
13440 @item -mdiv=@var{strategy}
13441 @opindex mdiv=@var{strategy}
13442 Set the division strategy to use for SHmedia code. @var{strategy} must be
13443 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
13444 inv:call2, inv:fp .
13445 "fp" performs the operation in floating point. This has a very high latency,
13446 but needs only a few instructions, so it might be a good choice if
13447 your code has enough easily exploitable ILP to allow the compiler to
13448 schedule the floating point instructions together with other instructions.
13449 Division by zero causes a floating point exception.
13450 "inv" uses integer operations to calculate the inverse of the divisor,
13451 and then multiplies the dividend with the inverse. This strategy allows
13452 cse and hoisting of the inverse calculation. Division by zero calculates
13453 an unspecified result, but does not trap.
13454 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
13455 have been found, or if the entire operation has been hoisted to the same
13456 place, the last stages of the inverse calculation are intertwined with the
13457 final multiply to reduce the overall latency, at the expense of using a few
13458 more instructions, and thus offering fewer scheduling opportunities with
13460 "call" calls a library function that usually implements the inv:minlat
13462 This gives high code density for m5-*media-nofpu compilations.
13463 "call2" uses a different entry point of the same library function, where it
13464 assumes that a pointer to a lookup table has already been set up, which
13465 exposes the pointer load to cse / code hoisting optimizations.
13466 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
13467 code generation, but if the code stays unoptimized, revert to the "call",
13468 "call2", or "fp" strategies, respectively. Note that the
13469 potentially-trapping side effect of division by zero is carried by a
13470 separate instruction, so it is possible that all the integer instructions
13471 are hoisted out, but the marker for the side effect stays where it is.
13472 A recombination to fp operations or a call is not possible in that case.
13473 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
13474 that the inverse calculation was nor separated from the multiply, they speed
13475 up division where the dividend fits into 20 bits (plus sign where applicable),
13476 by inserting a test to skip a number of operations in this case; this test
13477 slows down the case of larger dividends. inv20u assumes the case of a such
13478 a small dividend to be unlikely, and inv20l assumes it to be likely.
13480 @item -mdivsi3_libfunc=@var{name}
13481 @opindex mdivsi3_libfunc=@var{name}
13482 Set the name of the library function used for 32 bit signed division to
13483 @var{name}. This only affect the name used in the call and inv:call
13484 division strategies, and the compiler will still expect the same
13485 sets of input/output/clobbered registers as if this option was not present.
13487 @item -madjust-unroll
13488 @opindex madjust-unroll
13489 Throttle unrolling to avoid thrashing target registers.
13490 This option only has an effect if the gcc code base supports the
13491 TARGET_ADJUST_UNROLL_MAX target hook.
13493 @item -mindexed-addressing
13494 @opindex mindexed-addressing
13495 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
13496 This is only safe if the hardware and/or OS implement 32 bit wrap-around
13497 semantics for the indexed addressing mode. The architecture allows the
13498 implementation of processors with 64 bit MMU, which the OS could use to
13499 get 32 bit addressing, but since no current hardware implementation supports
13500 this or any other way to make the indexed addressing mode safe to use in
13501 the 32 bit ABI, the default is -mno-indexed-addressing.
13503 @item -mgettrcost=@var{number}
13504 @opindex mgettrcost=@var{number}
13505 Set the cost assumed for the gettr instruction to @var{number}.
13506 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
13510 Assume pt* instructions won't trap. This will generally generate better
13511 scheduled code, but is unsafe on current hardware. The current architecture
13512 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
13513 This has the unintentional effect of making it unsafe to schedule ptabs /
13514 ptrel before a branch, or hoist it out of a loop. For example,
13515 __do_global_ctors, a part of libgcc that runs constructors at program
13516 startup, calls functions in a list which is delimited by @minus{}1. With the
13517 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
13518 That means that all the constructors will be run a bit quicker, but when
13519 the loop comes to the end of the list, the program crashes because ptabs
13520 loads @minus{}1 into a target register. Since this option is unsafe for any
13521 hardware implementing the current architecture specification, the default
13522 is -mno-pt-fixed. Unless the user specifies a specific cost with
13523 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
13524 this deters register allocation using target registers for storing
13527 @item -minvalid-symbols
13528 @opindex minvalid-symbols
13529 Assume symbols might be invalid. Ordinary function symbols generated by
13530 the compiler will always be valid to load with movi/shori/ptabs or
13531 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
13532 to generate symbols that will cause ptabs / ptrel to trap.
13533 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
13534 It will then prevent cross-basic-block cse, hoisting and most scheduling
13535 of symbol loads. The default is @option{-mno-invalid-symbols}.
13538 @node SPARC Options
13539 @subsection SPARC Options
13540 @cindex SPARC options
13542 These @samp{-m} options are supported on the SPARC:
13545 @item -mno-app-regs
13547 @opindex mno-app-regs
13549 Specify @option{-mapp-regs} to generate output using the global registers
13550 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
13553 To be fully SVR4 ABI compliant at the cost of some performance loss,
13554 specify @option{-mno-app-regs}. You should compile libraries and system
13555 software with this option.
13558 @itemx -mhard-float
13560 @opindex mhard-float
13561 Generate output containing floating point instructions. This is the
13565 @itemx -msoft-float
13567 @opindex msoft-float
13568 Generate output containing library calls for floating point.
13569 @strong{Warning:} the requisite libraries are not available for all SPARC
13570 targets. Normally the facilities of the machine's usual C compiler are
13571 used, but this cannot be done directly in cross-compilation. You must make
13572 your own arrangements to provide suitable library functions for
13573 cross-compilation. The embedded targets @samp{sparc-*-aout} and
13574 @samp{sparclite-*-*} do provide software floating point support.
13576 @option{-msoft-float} changes the calling convention in the output file;
13577 therefore, it is only useful if you compile @emph{all} of a program with
13578 this option. In particular, you need to compile @file{libgcc.a}, the
13579 library that comes with GCC, with @option{-msoft-float} in order for
13582 @item -mhard-quad-float
13583 @opindex mhard-quad-float
13584 Generate output containing quad-word (long double) floating point
13587 @item -msoft-quad-float
13588 @opindex msoft-quad-float
13589 Generate output containing library calls for quad-word (long double)
13590 floating point instructions. The functions called are those specified
13591 in the SPARC ABI@. This is the default.
13593 As of this writing, there are no SPARC implementations that have hardware
13594 support for the quad-word floating point instructions. They all invoke
13595 a trap handler for one of these instructions, and then the trap handler
13596 emulates the effect of the instruction. Because of the trap handler overhead,
13597 this is much slower than calling the ABI library routines. Thus the
13598 @option{-msoft-quad-float} option is the default.
13600 @item -mno-unaligned-doubles
13601 @itemx -munaligned-doubles
13602 @opindex mno-unaligned-doubles
13603 @opindex munaligned-doubles
13604 Assume that doubles have 8 byte alignment. This is the default.
13606 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
13607 alignment only if they are contained in another type, or if they have an
13608 absolute address. Otherwise, it assumes they have 4 byte alignment.
13609 Specifying this option avoids some rare compatibility problems with code
13610 generated by other compilers. It is not the default because it results
13611 in a performance loss, especially for floating point code.
13613 @item -mno-faster-structs
13614 @itemx -mfaster-structs
13615 @opindex mno-faster-structs
13616 @opindex mfaster-structs
13617 With @option{-mfaster-structs}, the compiler assumes that structures
13618 should have 8 byte alignment. This enables the use of pairs of
13619 @code{ldd} and @code{std} instructions for copies in structure
13620 assignment, in place of twice as many @code{ld} and @code{st} pairs.
13621 However, the use of this changed alignment directly violates the SPARC
13622 ABI@. Thus, it's intended only for use on targets where the developer
13623 acknowledges that their resulting code will not be directly in line with
13624 the rules of the ABI@.
13626 @item -mimpure-text
13627 @opindex mimpure-text
13628 @option{-mimpure-text}, used in addition to @option{-shared}, tells
13629 the compiler to not pass @option{-z text} to the linker when linking a
13630 shared object. Using this option, you can link position-dependent
13631 code into a shared object.
13633 @option{-mimpure-text} suppresses the ``relocations remain against
13634 allocatable but non-writable sections'' linker error message.
13635 However, the necessary relocations will trigger copy-on-write, and the
13636 shared object is not actually shared across processes. Instead of
13637 using @option{-mimpure-text}, you should compile all source code with
13638 @option{-fpic} or @option{-fPIC}.
13640 This option is only available on SunOS and Solaris.
13642 @item -mcpu=@var{cpu_type}
13644 Set the instruction set, register set, and instruction scheduling parameters
13645 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
13646 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
13647 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
13648 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
13649 @samp{ultrasparc3}, and @samp{niagara}.
13651 Default instruction scheduling parameters are used for values that select
13652 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
13653 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
13655 Here is a list of each supported architecture and their supported
13660 v8: supersparc, hypersparc
13661 sparclite: f930, f934, sparclite86x
13663 v9: ultrasparc, ultrasparc3, niagara
13666 By default (unless configured otherwise), GCC generates code for the V7
13667 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
13668 additionally optimizes it for the Cypress CY7C602 chip, as used in the
13669 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
13670 SPARCStation 1, 2, IPX etc.
13672 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
13673 architecture. The only difference from V7 code is that the compiler emits
13674 the integer multiply and integer divide instructions which exist in SPARC-V8
13675 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
13676 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
13679 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
13680 the SPARC architecture. This adds the integer multiply, integer divide step
13681 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
13682 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
13683 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
13684 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
13685 MB86934 chip, which is the more recent SPARClite with FPU@.
13687 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
13688 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
13689 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
13690 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
13691 optimizes it for the TEMIC SPARClet chip.
13693 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
13694 architecture. This adds 64-bit integer and floating-point move instructions,
13695 3 additional floating-point condition code registers and conditional move
13696 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
13697 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
13698 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
13699 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
13700 @option{-mcpu=niagara}, the compiler additionally optimizes it for
13701 Sun UltraSPARC T1 chips.
13703 @item -mtune=@var{cpu_type}
13705 Set the instruction scheduling parameters for machine type
13706 @var{cpu_type}, but do not set the instruction set or register set that the
13707 option @option{-mcpu=@var{cpu_type}} would.
13709 The same values for @option{-mcpu=@var{cpu_type}} can be used for
13710 @option{-mtune=@var{cpu_type}}, but the only useful values are those
13711 that select a particular cpu implementation. Those are @samp{cypress},
13712 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
13713 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
13714 @samp{ultrasparc3}, and @samp{niagara}.
13719 @opindex mno-v8plus
13720 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
13721 difference from the V8 ABI is that the global and out registers are
13722 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
13723 mode for all SPARC-V9 processors.
13729 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
13730 Visual Instruction Set extensions. The default is @option{-mno-vis}.
13733 These @samp{-m} options are supported in addition to the above
13734 on SPARC-V9 processors in 64-bit environments:
13737 @item -mlittle-endian
13738 @opindex mlittle-endian
13739 Generate code for a processor running in little-endian mode. It is only
13740 available for a few configurations and most notably not on Solaris and Linux.
13746 Generate code for a 32-bit or 64-bit environment.
13747 The 32-bit environment sets int, long and pointer to 32 bits.
13748 The 64-bit environment sets int to 32 bits and long and pointer
13751 @item -mcmodel=medlow
13752 @opindex mcmodel=medlow
13753 Generate code for the Medium/Low code model: 64-bit addresses, programs
13754 must be linked in the low 32 bits of memory. Programs can be statically
13755 or dynamically linked.
13757 @item -mcmodel=medmid
13758 @opindex mcmodel=medmid
13759 Generate code for the Medium/Middle code model: 64-bit addresses, programs
13760 must be linked in the low 44 bits of memory, the text and data segments must
13761 be less than 2GB in size and the data segment must be located within 2GB of
13764 @item -mcmodel=medany
13765 @opindex mcmodel=medany
13766 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
13767 may be linked anywhere in memory, the text and data segments must be less
13768 than 2GB in size and the data segment must be located within 2GB of the
13771 @item -mcmodel=embmedany
13772 @opindex mcmodel=embmedany
13773 Generate code for the Medium/Anywhere code model for embedded systems:
13774 64-bit addresses, the text and data segments must be less than 2GB in
13775 size, both starting anywhere in memory (determined at link time). The
13776 global register %g4 points to the base of the data segment. Programs
13777 are statically linked and PIC is not supported.
13780 @itemx -mno-stack-bias
13781 @opindex mstack-bias
13782 @opindex mno-stack-bias
13783 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
13784 frame pointer if present, are offset by @minus{}2047 which must be added back
13785 when making stack frame references. This is the default in 64-bit mode.
13786 Otherwise, assume no such offset is present.
13789 These switches are supported in addition to the above on Solaris:
13794 Add support for multithreading using the Solaris threads library. This
13795 option sets flags for both the preprocessor and linker. This option does
13796 not affect the thread safety of object code produced by the compiler or
13797 that of libraries supplied with it.
13801 Add support for multithreading using the POSIX threads library. This
13802 option sets flags for both the preprocessor and linker. This option does
13803 not affect the thread safety of object code produced by the compiler or
13804 that of libraries supplied with it.
13808 This is a synonym for @option{-pthreads}.
13812 @subsection SPU Options
13813 @cindex SPU options
13815 These @samp{-m} options are supported on the SPU:
13819 @itemx -merror-reloc
13820 @opindex mwarn-reloc
13821 @opindex merror-reloc
13823 The loader for SPU does not handle dynamic relocations. By default, GCC
13824 will give an error when it generates code that requires a dynamic
13825 relocation. @option{-mno-error-reloc} disables the error,
13826 @option{-mwarn-reloc} will generate a warning instead.
13829 @itemx -munsafe-dma
13831 @opindex munsafe-dma
13833 Instructions which initiate or test completion of DMA must not be
13834 reordered with respect to loads and stores of the memory which is being
13835 accessed. Users typically address this problem using the volatile
13836 keyword, but that can lead to inefficient code in places where the
13837 memory is known to not change. Rather than mark the memory as volatile
13838 we treat the DMA instructions as potentially effecting all memory. With
13839 @option{-munsafe-dma} users must use the volatile keyword to protect
13842 @item -mbranch-hints
13843 @opindex mbranch-hints
13845 By default, GCC will generate a branch hint instruction to avoid
13846 pipeline stalls for always taken or probably taken branches. A hint
13847 will not be generated closer than 8 instructions away from its branch.
13848 There is little reason to disable them, except for debugging purposes,
13849 or to make an object a little bit smaller.
13853 @opindex msmall-mem
13854 @opindex mlarge-mem
13856 By default, GCC generates code assuming that addresses are never larger
13857 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
13858 a full 32 bit address.
13863 By default, GCC links against startup code that assumes the SPU-style
13864 main function interface (which has an unconventional parameter list).
13865 With @option{-mstdmain}, GCC will link your program against startup
13866 code that assumes a C99-style interface to @code{main}, including a
13867 local copy of @code{argv} strings.
13869 @item -mfixed-range=@var{register-range}
13870 @opindex mfixed-range
13871 Generate code treating the given register range as fixed registers.
13872 A fixed register is one that the register allocator can not use. This is
13873 useful when compiling kernel code. A register range is specified as
13874 two registers separated by a dash. Multiple register ranges can be
13875 specified separated by a comma.
13879 @node System V Options
13880 @subsection Options for System V
13882 These additional options are available on System V Release 4 for
13883 compatibility with other compilers on those systems:
13888 Create a shared object.
13889 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
13893 Identify the versions of each tool used by the compiler, in a
13894 @code{.ident} assembler directive in the output.
13898 Refrain from adding @code{.ident} directives to the output file (this is
13901 @item -YP,@var{dirs}
13903 Search the directories @var{dirs}, and no others, for libraries
13904 specified with @option{-l}.
13906 @item -Ym,@var{dir}
13908 Look in the directory @var{dir} to find the M4 preprocessor.
13909 The assembler uses this option.
13910 @c This is supposed to go with a -Yd for predefined M4 macro files, but
13911 @c the generic assembler that comes with Solaris takes just -Ym.
13914 @node TMS320C3x/C4x Options
13915 @subsection TMS320C3x/C4x Options
13916 @cindex TMS320C3x/C4x Options
13918 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
13922 @item -mcpu=@var{cpu_type}
13924 Set the instruction set, register set, and instruction scheduling
13925 parameters for machine type @var{cpu_type}. Supported values for
13926 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
13927 @samp{c44}. The default is @samp{c40} to generate code for the
13932 @itemx -msmall-memory
13934 @opindex mbig-memory
13936 @opindex msmall-memory
13938 Generates code for the big or small memory model. The small memory
13939 model assumed that all data fits into one 64K word page. At run-time
13940 the data page (DP) register must be set to point to the 64K page
13941 containing the .bss and .data program sections. The big memory model is
13942 the default and requires reloading of the DP register for every direct
13949 Allow (disallow) allocation of general integer operands into the block
13950 count register BK@.
13956 Enable (disable) generation of code using decrement and branch,
13957 DBcond(D), instructions. This is enabled by default for the C4x. To be
13958 on the safe side, this is disabled for the C3x, since the maximum
13959 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
13960 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
13961 that it can utilize the decrement and branch instruction, but will give
13962 up if there is more than one memory reference in the loop. Thus a loop
13963 where the loop counter is decremented can generate slightly more
13964 efficient code, in cases where the RPTB instruction cannot be utilized.
13966 @item -mdp-isr-reload
13968 @opindex mdp-isr-reload
13970 Force the DP register to be saved on entry to an interrupt service
13971 routine (ISR), reloaded to point to the data section, and restored on
13972 exit from the ISR@. This should not be required unless someone has
13973 violated the small memory model by modifying the DP register, say within
13980 For the C3x use the 24-bit MPYI instruction for integer multiplies
13981 instead of a library call to guarantee 32-bit results. Note that if one
13982 of the operands is a constant, then the multiplication will be performed
13983 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
13984 then squaring operations are performed inline instead of a library call.
13987 @itemx -mno-fast-fix
13989 @opindex mno-fast-fix
13990 The C3x/C4x FIX instruction to convert a floating point value to an
13991 integer value chooses the nearest integer less than or equal to the
13992 floating point value rather than to the nearest integer. Thus if the
13993 floating point number is negative, the result will be incorrectly
13994 truncated an additional code is necessary to detect and correct this
13995 case. This option can be used to disable generation of the additional
13996 code required to correct the result.
14002 Enable (disable) generation of repeat block sequences using the RPTB
14003 instruction for zero overhead looping. The RPTB construct is only used
14004 for innermost loops that do not call functions or jump across the loop
14005 boundaries. There is no advantage having nested RPTB loops due to the
14006 overhead required to save and restore the RC, RS, and RE registers.
14007 This is enabled by default with @option{-O2}.
14009 @item -mrpts=@var{count}
14013 Enable (disable) the use of the single instruction repeat instruction
14014 RPTS@. If a repeat block contains a single instruction, and the loop
14015 count can be guaranteed to be less than the value @var{count}, GCC will
14016 emit a RPTS instruction instead of a RPTB@. If no value is specified,
14017 then a RPTS will be emitted even if the loop count cannot be determined
14018 at compile time. Note that the repeated instruction following RPTS does
14019 not have to be reloaded from memory each iteration, thus freeing up the
14020 CPU buses for operands. However, since interrupts are blocked by this
14021 instruction, it is disabled by default.
14023 @item -mloop-unsigned
14024 @itemx -mno-loop-unsigned
14025 @opindex mloop-unsigned
14026 @opindex mno-loop-unsigned
14027 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
14028 is @math{2^{31} + 1} since these instructions test if the iteration count is
14029 negative to terminate the loop. If the iteration count is unsigned
14030 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
14031 exceeded. This switch allows an unsigned iteration count.
14035 Try to emit an assembler syntax that the TI assembler (asm30) is happy
14036 with. This also enforces compatibility with the API employed by the TI
14037 C3x C compiler. For example, long doubles are passed as structures
14038 rather than in floating point registers.
14044 Generate code that uses registers (stack) for passing arguments to functions.
14045 By default, arguments are passed in registers where possible rather
14046 than by pushing arguments on to the stack.
14048 @item -mparallel-insns
14049 @itemx -mno-parallel-insns
14050 @opindex mparallel-insns
14051 @opindex mno-parallel-insns
14052 Allow the generation of parallel instructions. This is enabled by
14053 default with @option{-O2}.
14055 @item -mparallel-mpy
14056 @itemx -mno-parallel-mpy
14057 @opindex mparallel-mpy
14058 @opindex mno-parallel-mpy
14059 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
14060 provided @option{-mparallel-insns} is also specified. These instructions have
14061 tight register constraints which can pessimize the code generation
14062 of large functions.
14067 @subsection V850 Options
14068 @cindex V850 Options
14070 These @samp{-m} options are defined for V850 implementations:
14074 @itemx -mno-long-calls
14075 @opindex mlong-calls
14076 @opindex mno-long-calls
14077 Treat all calls as being far away (near). If calls are assumed to be
14078 far away, the compiler will always load the functions address up into a
14079 register, and call indirect through the pointer.
14085 Do not optimize (do optimize) basic blocks that use the same index
14086 pointer 4 or more times to copy pointer into the @code{ep} register, and
14087 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
14088 option is on by default if you optimize.
14090 @item -mno-prolog-function
14091 @itemx -mprolog-function
14092 @opindex mno-prolog-function
14093 @opindex mprolog-function
14094 Do not use (do use) external functions to save and restore registers
14095 at the prologue and epilogue of a function. The external functions
14096 are slower, but use less code space if more than one function saves
14097 the same number of registers. The @option{-mprolog-function} option
14098 is on by default if you optimize.
14102 Try to make the code as small as possible. At present, this just turns
14103 on the @option{-mep} and @option{-mprolog-function} options.
14105 @item -mtda=@var{n}
14107 Put static or global variables whose size is @var{n} bytes or less into
14108 the tiny data area that register @code{ep} points to. The tiny data
14109 area can hold up to 256 bytes in total (128 bytes for byte references).
14111 @item -msda=@var{n}
14113 Put static or global variables whose size is @var{n} bytes or less into
14114 the small data area that register @code{gp} points to. The small data
14115 area can hold up to 64 kilobytes.
14117 @item -mzda=@var{n}
14119 Put static or global variables whose size is @var{n} bytes or less into
14120 the first 32 kilobytes of memory.
14124 Specify that the target processor is the V850.
14127 @opindex mbig-switch
14128 Generate code suitable for big switch tables. Use this option only if
14129 the assembler/linker complain about out of range branches within a switch
14134 This option will cause r2 and r5 to be used in the code generated by
14135 the compiler. This setting is the default.
14137 @item -mno-app-regs
14138 @opindex mno-app-regs
14139 This option will cause r2 and r5 to be treated as fixed registers.
14143 Specify that the target processor is the V850E1. The preprocessor
14144 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
14145 this option is used.
14149 Specify that the target processor is the V850E@. The preprocessor
14150 constant @samp{__v850e__} will be defined if this option is used.
14152 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
14153 are defined then a default target processor will be chosen and the
14154 relevant @samp{__v850*__} preprocessor constant will be defined.
14156 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
14157 defined, regardless of which processor variant is the target.
14159 @item -mdisable-callt
14160 @opindex mdisable-callt
14161 This option will suppress generation of the CALLT instruction for the
14162 v850e and v850e1 flavors of the v850 architecture. The default is
14163 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
14168 @subsection VAX Options
14169 @cindex VAX options
14171 These @samp{-m} options are defined for the VAX:
14176 Do not output certain jump instructions (@code{aobleq} and so on)
14177 that the Unix assembler for the VAX cannot handle across long
14182 Do output those jump instructions, on the assumption that you
14183 will assemble with the GNU assembler.
14187 Output code for g-format floating point numbers instead of d-format.
14190 @node VxWorks Options
14191 @subsection VxWorks Options
14192 @cindex VxWorks Options
14194 The options in this section are defined for all VxWorks targets.
14195 Options specific to the target hardware are listed with the other
14196 options for that target.
14201 GCC can generate code for both VxWorks kernels and real time processes
14202 (RTPs). This option switches from the former to the latter. It also
14203 defines the preprocessor macro @code{__RTP__}.
14206 @opindex non-static
14207 Link an RTP executable against shared libraries rather than static
14208 libraries. The options @option{-static} and @option{-shared} can
14209 also be used for RTPs (@pxref{Link Options}); @option{-static}
14216 These options are passed down to the linker. They are defined for
14217 compatibility with Diab.
14220 @opindex Xbind-lazy
14221 Enable lazy binding of function calls. This option is equivalent to
14222 @option{-Wl,-z,now} and is defined for compatibility with Diab.
14226 Disable lazy binding of function calls. This option is the default and
14227 is defined for compatibility with Diab.
14230 @node x86-64 Options
14231 @subsection x86-64 Options
14232 @cindex x86-64 options
14234 These are listed under @xref{i386 and x86-64 Options}.
14236 @node Xstormy16 Options
14237 @subsection Xstormy16 Options
14238 @cindex Xstormy16 Options
14240 These options are defined for Xstormy16:
14245 Choose startup files and linker script suitable for the simulator.
14248 @node Xtensa Options
14249 @subsection Xtensa Options
14250 @cindex Xtensa Options
14252 These options are supported for Xtensa targets:
14256 @itemx -mno-const16
14258 @opindex mno-const16
14259 Enable or disable use of @code{CONST16} instructions for loading
14260 constant values. The @code{CONST16} instruction is currently not a
14261 standard option from Tensilica. When enabled, @code{CONST16}
14262 instructions are always used in place of the standard @code{L32R}
14263 instructions. The use of @code{CONST16} is enabled by default only if
14264 the @code{L32R} instruction is not available.
14267 @itemx -mno-fused-madd
14268 @opindex mfused-madd
14269 @opindex mno-fused-madd
14270 Enable or disable use of fused multiply/add and multiply/subtract
14271 instructions in the floating-point option. This has no effect if the
14272 floating-point option is not also enabled. Disabling fused multiply/add
14273 and multiply/subtract instructions forces the compiler to use separate
14274 instructions for the multiply and add/subtract operations. This may be
14275 desirable in some cases where strict IEEE 754-compliant results are
14276 required: the fused multiply add/subtract instructions do not round the
14277 intermediate result, thereby producing results with @emph{more} bits of
14278 precision than specified by the IEEE standard. Disabling fused multiply
14279 add/subtract instructions also ensures that the program output is not
14280 sensitive to the compiler's ability to combine multiply and add/subtract
14283 @item -mtext-section-literals
14284 @itemx -mno-text-section-literals
14285 @opindex mtext-section-literals
14286 @opindex mno-text-section-literals
14287 Control the treatment of literal pools. The default is
14288 @option{-mno-text-section-literals}, which places literals in a separate
14289 section in the output file. This allows the literal pool to be placed
14290 in a data RAM/ROM, and it also allows the linker to combine literal
14291 pools from separate object files to remove redundant literals and
14292 improve code size. With @option{-mtext-section-literals}, the literals
14293 are interspersed in the text section in order to keep them as close as
14294 possible to their references. This may be necessary for large assembly
14297 @item -mtarget-align
14298 @itemx -mno-target-align
14299 @opindex mtarget-align
14300 @opindex mno-target-align
14301 When this option is enabled, GCC instructs the assembler to
14302 automatically align instructions to reduce branch penalties at the
14303 expense of some code density. The assembler attempts to widen density
14304 instructions to align branch targets and the instructions following call
14305 instructions. If there are not enough preceding safe density
14306 instructions to align a target, no widening will be performed. The
14307 default is @option{-mtarget-align}. These options do not affect the
14308 treatment of auto-aligned instructions like @code{LOOP}, which the
14309 assembler will always align, either by widening density instructions or
14310 by inserting no-op instructions.
14313 @itemx -mno-longcalls
14314 @opindex mlongcalls
14315 @opindex mno-longcalls
14316 When this option is enabled, GCC instructs the assembler to translate
14317 direct calls to indirect calls unless it can determine that the target
14318 of a direct call is in the range allowed by the call instruction. This
14319 translation typically occurs for calls to functions in other source
14320 files. Specifically, the assembler translates a direct @code{CALL}
14321 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
14322 The default is @option{-mno-longcalls}. This option should be used in
14323 programs where the call target can potentially be out of range. This
14324 option is implemented in the assembler, not the compiler, so the
14325 assembly code generated by GCC will still show direct call
14326 instructions---look at the disassembled object code to see the actual
14327 instructions. Note that the assembler will use an indirect call for
14328 every cross-file call, not just those that really will be out of range.
14331 @node zSeries Options
14332 @subsection zSeries Options
14333 @cindex zSeries options
14335 These are listed under @xref{S/390 and zSeries Options}.
14337 @node Code Gen Options
14338 @section Options for Code Generation Conventions
14339 @cindex code generation conventions
14340 @cindex options, code generation
14341 @cindex run-time options
14343 These machine-independent options control the interface conventions
14344 used in code generation.
14346 Most of them have both positive and negative forms; the negative form
14347 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
14348 one of the forms is listed---the one which is not the default. You
14349 can figure out the other form by either removing @samp{no-} or adding
14353 @item -fbounds-check
14354 @opindex fbounds-check
14355 For front-ends that support it, generate additional code to check that
14356 indices used to access arrays are within the declared range. This is
14357 currently only supported by the Java and Fortran front-ends, where
14358 this option defaults to true and false respectively.
14362 This option generates traps for signed overflow on addition, subtraction,
14363 multiplication operations.
14367 This option instructs the compiler to assume that signed arithmetic
14368 overflow of addition, subtraction and multiplication wraps around
14369 using twos-complement representation. This flag enables some optimizations
14370 and disables others. This option is enabled by default for the Java
14371 front-end, as required by the Java language specification.
14374 @opindex fexceptions
14375 Enable exception handling. Generates extra code needed to propagate
14376 exceptions. For some targets, this implies GCC will generate frame
14377 unwind information for all functions, which can produce significant data
14378 size overhead, although it does not affect execution. If you do not
14379 specify this option, GCC will enable it by default for languages like
14380 C++ which normally require exception handling, and disable it for
14381 languages like C that do not normally require it. However, you may need
14382 to enable this option when compiling C code that needs to interoperate
14383 properly with exception handlers written in C++. You may also wish to
14384 disable this option if you are compiling older C++ programs that don't
14385 use exception handling.
14387 @item -fnon-call-exceptions
14388 @opindex fnon-call-exceptions
14389 Generate code that allows trapping instructions to throw exceptions.
14390 Note that this requires platform-specific runtime support that does
14391 not exist everywhere. Moreover, it only allows @emph{trapping}
14392 instructions to throw exceptions, i.e.@: memory references or floating
14393 point instructions. It does not allow exceptions to be thrown from
14394 arbitrary signal handlers such as @code{SIGALRM}.
14396 @item -funwind-tables
14397 @opindex funwind-tables
14398 Similar to @option{-fexceptions}, except that it will just generate any needed
14399 static data, but will not affect the generated code in any other way.
14400 You will normally not enable this option; instead, a language processor
14401 that needs this handling would enable it on your behalf.
14403 @item -fasynchronous-unwind-tables
14404 @opindex fasynchronous-unwind-tables
14405 Generate unwind table in dwarf2 format, if supported by target machine. The
14406 table is exact at each instruction boundary, so it can be used for stack
14407 unwinding from asynchronous events (such as debugger or garbage collector).
14409 @item -fpcc-struct-return
14410 @opindex fpcc-struct-return
14411 Return ``short'' @code{struct} and @code{union} values in memory like
14412 longer ones, rather than in registers. This convention is less
14413 efficient, but it has the advantage of allowing intercallability between
14414 GCC-compiled files and files compiled with other compilers, particularly
14415 the Portable C Compiler (pcc).
14417 The precise convention for returning structures in memory depends
14418 on the target configuration macros.
14420 Short structures and unions are those whose size and alignment match
14421 that of some integer type.
14423 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
14424 switch is not binary compatible with code compiled with the
14425 @option{-freg-struct-return} switch.
14426 Use it to conform to a non-default application binary interface.
14428 @item -freg-struct-return
14429 @opindex freg-struct-return
14430 Return @code{struct} and @code{union} values in registers when possible.
14431 This is more efficient for small structures than
14432 @option{-fpcc-struct-return}.
14434 If you specify neither @option{-fpcc-struct-return} nor
14435 @option{-freg-struct-return}, GCC defaults to whichever convention is
14436 standard for the target. If there is no standard convention, GCC
14437 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
14438 the principal compiler. In those cases, we can choose the standard, and
14439 we chose the more efficient register return alternative.
14441 @strong{Warning:} code compiled with the @option{-freg-struct-return}
14442 switch is not binary compatible with code compiled with the
14443 @option{-fpcc-struct-return} switch.
14444 Use it to conform to a non-default application binary interface.
14446 @item -fshort-enums
14447 @opindex fshort-enums
14448 Allocate to an @code{enum} type only as many bytes as it needs for the
14449 declared range of possible values. Specifically, the @code{enum} type
14450 will be equivalent to the smallest integer type which has enough room.
14452 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
14453 code that is not binary compatible with code generated without that switch.
14454 Use it to conform to a non-default application binary interface.
14456 @item -fshort-double
14457 @opindex fshort-double
14458 Use the same size for @code{double} as for @code{float}.
14460 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
14461 code that is not binary compatible with code generated without that switch.
14462 Use it to conform to a non-default application binary interface.
14464 @item -fshort-wchar
14465 @opindex fshort-wchar
14466 Override the underlying type for @samp{wchar_t} to be @samp{short
14467 unsigned int} instead of the default for the target. This option is
14468 useful for building programs to run under WINE@.
14470 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
14471 code that is not binary compatible with code generated without that switch.
14472 Use it to conform to a non-default application binary interface.
14475 @opindex fno-common
14476 In C, allocate even uninitialized global variables in the data section of the
14477 object file, rather than generating them as common blocks. This has the
14478 effect that if the same variable is declared (without @code{extern}) in
14479 two different compilations, you will get an error when you link them.
14480 The only reason this might be useful is if you wish to verify that the
14481 program will work on other systems which always work this way.
14485 Ignore the @samp{#ident} directive.
14487 @item -finhibit-size-directive
14488 @opindex finhibit-size-directive
14489 Don't output a @code{.size} assembler directive, or anything else that
14490 would cause trouble if the function is split in the middle, and the
14491 two halves are placed at locations far apart in memory. This option is
14492 used when compiling @file{crtstuff.c}; you should not need to use it
14495 @item -fverbose-asm
14496 @opindex fverbose-asm
14497 Put extra commentary information in the generated assembly code to
14498 make it more readable. This option is generally only of use to those
14499 who actually need to read the generated assembly code (perhaps while
14500 debugging the compiler itself).
14502 @option{-fno-verbose-asm}, the default, causes the
14503 extra information to be omitted and is useful when comparing two assembler
14506 @item -frecord-gcc-switches
14507 @opindex frecord-gcc-switches
14508 This switch causes the command line that was used to invoke the
14509 compiler to be recorded into the object file that is being created.
14510 This switch is only implemented on some targets and the exact format
14511 of the recording is target and binary file format dependent, but it
14512 usually takes the form of a section containing ASCII text. This
14513 switch is related to the @option{-fverbose-asm} switch, but that
14514 switch only records information in the assembler output file as
14515 comments, so it never reaches the object file.
14519 @cindex global offset table
14521 Generate position-independent code (PIC) suitable for use in a shared
14522 library, if supported for the target machine. Such code accesses all
14523 constant addresses through a global offset table (GOT)@. The dynamic
14524 loader resolves the GOT entries when the program starts (the dynamic
14525 loader is not part of GCC; it is part of the operating system). If
14526 the GOT size for the linked executable exceeds a machine-specific
14527 maximum size, you get an error message from the linker indicating that
14528 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
14529 instead. (These maximums are 8k on the SPARC and 32k
14530 on the m68k and RS/6000. The 386 has no such limit.)
14532 Position-independent code requires special support, and therefore works
14533 only on certain machines. For the 386, GCC supports PIC for System V
14534 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
14535 position-independent.
14537 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14542 If supported for the target machine, emit position-independent code,
14543 suitable for dynamic linking and avoiding any limit on the size of the
14544 global offset table. This option makes a difference on the m68k,
14545 PowerPC and SPARC@.
14547 Position-independent code requires special support, and therefore works
14548 only on certain machines.
14550 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
14557 These options are similar to @option{-fpic} and @option{-fPIC}, but
14558 generated position independent code can be only linked into executables.
14559 Usually these options are used when @option{-pie} GCC option will be
14560 used during linking.
14562 @option{-fpie} and @option{-fPIE} both define the macros
14563 @code{__pie__} and @code{__PIE__}. The macros have the value 1
14564 for @option{-fpie} and 2 for @option{-fPIE}.
14566 @item -fno-jump-tables
14567 @opindex fno-jump-tables
14568 Do not use jump tables for switch statements even where it would be
14569 more efficient than other code generation strategies. This option is
14570 of use in conjunction with @option{-fpic} or @option{-fPIC} for
14571 building code which forms part of a dynamic linker and cannot
14572 reference the address of a jump table. On some targets, jump tables
14573 do not require a GOT and this option is not needed.
14575 @item -ffixed-@var{reg}
14577 Treat the register named @var{reg} as a fixed register; generated code
14578 should never refer to it (except perhaps as a stack pointer, frame
14579 pointer or in some other fixed role).
14581 @var{reg} must be the name of a register. The register names accepted
14582 are machine-specific and are defined in the @code{REGISTER_NAMES}
14583 macro in the machine description macro file.
14585 This flag does not have a negative form, because it specifies a
14588 @item -fcall-used-@var{reg}
14589 @opindex fcall-used
14590 Treat the register named @var{reg} as an allocable register that is
14591 clobbered by function calls. It may be allocated for temporaries or
14592 variables that do not live across a call. Functions compiled this way
14593 will not save and restore the register @var{reg}.
14595 It is an error to used this flag with the frame pointer or stack pointer.
14596 Use of this flag for other registers that have fixed pervasive roles in
14597 the machine's execution model will produce disastrous results.
14599 This flag does not have a negative form, because it specifies a
14602 @item -fcall-saved-@var{reg}
14603 @opindex fcall-saved
14604 Treat the register named @var{reg} as an allocable register saved by
14605 functions. It may be allocated even for temporaries or variables that
14606 live across a call. Functions compiled this way will save and restore
14607 the register @var{reg} if they use it.
14609 It is an error to used this flag with the frame pointer or stack pointer.
14610 Use of this flag for other registers that have fixed pervasive roles in
14611 the machine's execution model will produce disastrous results.
14613 A different sort of disaster will result from the use of this flag for
14614 a register in which function values may be returned.
14616 This flag does not have a negative form, because it specifies a
14619 @item -fpack-struct[=@var{n}]
14620 @opindex fpack-struct
14621 Without a value specified, pack all structure members together without
14622 holes. When a value is specified (which must be a small power of two), pack
14623 structure members according to this value, representing the maximum
14624 alignment (that is, objects with default alignment requirements larger than
14625 this will be output potentially unaligned at the next fitting location.
14627 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
14628 code that is not binary compatible with code generated without that switch.
14629 Additionally, it makes the code suboptimal.
14630 Use it to conform to a non-default application binary interface.
14632 @item -finstrument-functions
14633 @opindex finstrument-functions
14634 Generate instrumentation calls for entry and exit to functions. Just
14635 after function entry and just before function exit, the following
14636 profiling functions will be called with the address of the current
14637 function and its call site. (On some platforms,
14638 @code{__builtin_return_address} does not work beyond the current
14639 function, so the call site information may not be available to the
14640 profiling functions otherwise.)
14643 void __cyg_profile_func_enter (void *this_fn,
14645 void __cyg_profile_func_exit (void *this_fn,
14649 The first argument is the address of the start of the current function,
14650 which may be looked up exactly in the symbol table.
14652 This instrumentation is also done for functions expanded inline in other
14653 functions. The profiling calls will indicate where, conceptually, the
14654 inline function is entered and exited. This means that addressable
14655 versions of such functions must be available. If all your uses of a
14656 function are expanded inline, this may mean an additional expansion of
14657 code size. If you use @samp{extern inline} in your C code, an
14658 addressable version of such functions must be provided. (This is
14659 normally the case anyways, but if you get lucky and the optimizer always
14660 expands the functions inline, you might have gotten away without
14661 providing static copies.)
14663 A function may be given the attribute @code{no_instrument_function}, in
14664 which case this instrumentation will not be done. This can be used, for
14665 example, for the profiling functions listed above, high-priority
14666 interrupt routines, and any functions from which the profiling functions
14667 cannot safely be called (perhaps signal handlers, if the profiling
14668 routines generate output or allocate memory).
14670 @item -fstack-check
14671 @opindex fstack-check
14672 Generate code to verify that you do not go beyond the boundary of the
14673 stack. You should specify this flag if you are running in an
14674 environment with multiple threads, but only rarely need to specify it in
14675 a single-threaded environment since stack overflow is automatically
14676 detected on nearly all systems if there is only one stack.
14678 Note that this switch does not actually cause checking to be done; the
14679 operating system must do that. The switch causes generation of code
14680 to ensure that the operating system sees the stack being extended.
14682 @item -fstack-limit-register=@var{reg}
14683 @itemx -fstack-limit-symbol=@var{sym}
14684 @itemx -fno-stack-limit
14685 @opindex fstack-limit-register
14686 @opindex fstack-limit-symbol
14687 @opindex fno-stack-limit
14688 Generate code to ensure that the stack does not grow beyond a certain value,
14689 either the value of a register or the address of a symbol. If the stack
14690 would grow beyond the value, a signal is raised. For most targets,
14691 the signal is raised before the stack overruns the boundary, so
14692 it is possible to catch the signal without taking special precautions.
14694 For instance, if the stack starts at absolute address @samp{0x80000000}
14695 and grows downwards, you can use the flags
14696 @option{-fstack-limit-symbol=__stack_limit} and
14697 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
14698 of 128KB@. Note that this may only work with the GNU linker.
14700 @cindex aliasing of parameters
14701 @cindex parameters, aliased
14702 @item -fargument-alias
14703 @itemx -fargument-noalias
14704 @itemx -fargument-noalias-global
14705 @itemx -fargument-noalias-anything
14706 @opindex fargument-alias
14707 @opindex fargument-noalias
14708 @opindex fargument-noalias-global
14709 @opindex fargument-noalias-anything
14710 Specify the possible relationships among parameters and between
14711 parameters and global data.
14713 @option{-fargument-alias} specifies that arguments (parameters) may
14714 alias each other and may alias global storage.@*
14715 @option{-fargument-noalias} specifies that arguments do not alias
14716 each other, but may alias global storage.@*
14717 @option{-fargument-noalias-global} specifies that arguments do not
14718 alias each other and do not alias global storage.
14719 @option{-fargument-noalias-anything} specifies that arguments do not
14720 alias any other storage.
14722 Each language will automatically use whatever option is required by
14723 the language standard. You should not need to use these options yourself.
14725 @item -fleading-underscore
14726 @opindex fleading-underscore
14727 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
14728 change the way C symbols are represented in the object file. One use
14729 is to help link with legacy assembly code.
14731 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
14732 generate code that is not binary compatible with code generated without that
14733 switch. Use it to conform to a non-default application binary interface.
14734 Not all targets provide complete support for this switch.
14736 @item -ftls-model=@var{model}
14737 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
14738 The @var{model} argument should be one of @code{global-dynamic},
14739 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
14741 The default without @option{-fpic} is @code{initial-exec}; with
14742 @option{-fpic} the default is @code{global-dynamic}.
14744 @item -fvisibility=@var{default|internal|hidden|protected}
14745 @opindex fvisibility
14746 Set the default ELF image symbol visibility to the specified option---all
14747 symbols will be marked with this unless overridden within the code.
14748 Using this feature can very substantially improve linking and
14749 load times of shared object libraries, produce more optimized
14750 code, provide near-perfect API export and prevent symbol clashes.
14751 It is @strong{strongly} recommended that you use this in any shared objects
14754 Despite the nomenclature, @code{default} always means public ie;
14755 available to be linked against from outside the shared object.
14756 @code{protected} and @code{internal} are pretty useless in real-world
14757 usage so the only other commonly used option will be @code{hidden}.
14758 The default if @option{-fvisibility} isn't specified is
14759 @code{default}, i.e., make every
14760 symbol public---this causes the same behavior as previous versions of
14763 A good explanation of the benefits offered by ensuring ELF
14764 symbols have the correct visibility is given by ``How To Write
14765 Shared Libraries'' by Ulrich Drepper (which can be found at
14766 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
14767 solution made possible by this option to marking things hidden when
14768 the default is public is to make the default hidden and mark things
14769 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
14770 and @code{__attribute__ ((visibility("default")))} instead of
14771 @code{__declspec(dllexport)} you get almost identical semantics with
14772 identical syntax. This is a great boon to those working with
14773 cross-platform projects.
14775 For those adding visibility support to existing code, you may find
14776 @samp{#pragma GCC visibility} of use. This works by you enclosing
14777 the declarations you wish to set visibility for with (for example)
14778 @samp{#pragma GCC visibility push(hidden)} and
14779 @samp{#pragma GCC visibility pop}.
14780 Bear in mind that symbol visibility should be viewed @strong{as
14781 part of the API interface contract} and thus all new code should
14782 always specify visibility when it is not the default ie; declarations
14783 only for use within the local DSO should @strong{always} be marked explicitly
14784 as hidden as so to avoid PLT indirection overheads---making this
14785 abundantly clear also aids readability and self-documentation of the code.
14786 Note that due to ISO C++ specification requirements, operator new and
14787 operator delete must always be of default visibility.
14789 Be aware that headers from outside your project, in particular system
14790 headers and headers from any other library you use, may not be
14791 expecting to be compiled with visibility other than the default. You
14792 may need to explicitly say @samp{#pragma GCC visibility push(default)}
14793 before including any such headers.
14795 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
14796 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
14797 no modifications. However, this means that calls to @samp{extern}
14798 functions with no explicit visibility will use the PLT, so it is more
14799 effective to use @samp{__attribute ((visibility))} and/or
14800 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
14801 declarations should be treated as hidden.
14803 Note that @samp{-fvisibility} does affect C++ vague linkage
14804 entities. This means that, for instance, an exception class that will
14805 be thrown between DSOs must be explicitly marked with default
14806 visibility so that the @samp{type_info} nodes will be unified between
14809 An overview of these techniques, their benefits and how to use them
14810 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
14816 @node Environment Variables
14817 @section Environment Variables Affecting GCC
14818 @cindex environment variables
14820 @c man begin ENVIRONMENT
14821 This section describes several environment variables that affect how GCC
14822 operates. Some of them work by specifying directories or prefixes to use
14823 when searching for various kinds of files. Some are used to specify other
14824 aspects of the compilation environment.
14826 Note that you can also specify places to search using options such as
14827 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
14828 take precedence over places specified using environment variables, which
14829 in turn take precedence over those specified by the configuration of GCC@.
14830 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
14831 GNU Compiler Collection (GCC) Internals}.
14836 @c @itemx LC_COLLATE
14838 @c @itemx LC_MONETARY
14839 @c @itemx LC_NUMERIC
14844 @c @findex LC_COLLATE
14845 @findex LC_MESSAGES
14846 @c @findex LC_MONETARY
14847 @c @findex LC_NUMERIC
14851 These environment variables control the way that GCC uses
14852 localization information that allow GCC to work with different
14853 national conventions. GCC inspects the locale categories
14854 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
14855 so. These locale categories can be set to any value supported by your
14856 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
14857 Kingdom encoded in UTF-8.
14859 The @env{LC_CTYPE} environment variable specifies character
14860 classification. GCC uses it to determine the character boundaries in
14861 a string; this is needed for some multibyte encodings that contain quote
14862 and escape characters that would otherwise be interpreted as a string
14865 The @env{LC_MESSAGES} environment variable specifies the language to
14866 use in diagnostic messages.
14868 If the @env{LC_ALL} environment variable is set, it overrides the value
14869 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
14870 and @env{LC_MESSAGES} default to the value of the @env{LANG}
14871 environment variable. If none of these variables are set, GCC
14872 defaults to traditional C English behavior.
14876 If @env{TMPDIR} is set, it specifies the directory to use for temporary
14877 files. GCC uses temporary files to hold the output of one stage of
14878 compilation which is to be used as input to the next stage: for example,
14879 the output of the preprocessor, which is the input to the compiler
14882 @item GCC_EXEC_PREFIX
14883 @findex GCC_EXEC_PREFIX
14884 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
14885 names of the subprograms executed by the compiler. No slash is added
14886 when this prefix is combined with the name of a subprogram, but you can
14887 specify a prefix that ends with a slash if you wish.
14889 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
14890 an appropriate prefix to use based on the pathname it was invoked with.
14892 If GCC cannot find the subprogram using the specified prefix, it
14893 tries looking in the usual places for the subprogram.
14895 The default value of @env{GCC_EXEC_PREFIX} is
14896 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
14897 the installed compiler. In many cases @var{prefix} is the value
14898 of @code{prefix} when you ran the @file{configure} script.
14900 Other prefixes specified with @option{-B} take precedence over this prefix.
14902 This prefix is also used for finding files such as @file{crt0.o} that are
14905 In addition, the prefix is used in an unusual way in finding the
14906 directories to search for header files. For each of the standard
14907 directories whose name normally begins with @samp{/usr/local/lib/gcc}
14908 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
14909 replacing that beginning with the specified prefix to produce an
14910 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
14911 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
14912 These alternate directories are searched first; the standard directories
14913 come next. If a standard directory begins with the configured
14914 @var{prefix} then the value of @var{prefix} is replaced by
14915 @env{GCC_EXEC_PREFIX} when looking for header files.
14917 @item COMPILER_PATH
14918 @findex COMPILER_PATH
14919 The value of @env{COMPILER_PATH} is a colon-separated list of
14920 directories, much like @env{PATH}. GCC tries the directories thus
14921 specified when searching for subprograms, if it can't find the
14922 subprograms using @env{GCC_EXEC_PREFIX}.
14925 @findex LIBRARY_PATH
14926 The value of @env{LIBRARY_PATH} is a colon-separated list of
14927 directories, much like @env{PATH}. When configured as a native compiler,
14928 GCC tries the directories thus specified when searching for special
14929 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
14930 using GCC also uses these directories when searching for ordinary
14931 libraries for the @option{-l} option (but directories specified with
14932 @option{-L} come first).
14936 @cindex locale definition
14937 This variable is used to pass locale information to the compiler. One way in
14938 which this information is used is to determine the character set to be used
14939 when character literals, string literals and comments are parsed in C and C++.
14940 When the compiler is configured to allow multibyte characters,
14941 the following values for @env{LANG} are recognized:
14945 Recognize JIS characters.
14947 Recognize SJIS characters.
14949 Recognize EUCJP characters.
14952 If @env{LANG} is not defined, or if it has some other value, then the
14953 compiler will use mblen and mbtowc as defined by the default locale to
14954 recognize and translate multibyte characters.
14958 Some additional environments variables affect the behavior of the
14961 @include cppenv.texi
14965 @node Precompiled Headers
14966 @section Using Precompiled Headers
14967 @cindex precompiled headers
14968 @cindex speed of compilation
14970 Often large projects have many header files that are included in every
14971 source file. The time the compiler takes to process these header files
14972 over and over again can account for nearly all of the time required to
14973 build the project. To make builds faster, GCC allows users to
14974 `precompile' a header file; then, if builds can use the precompiled
14975 header file they will be much faster.
14977 To create a precompiled header file, simply compile it as you would any
14978 other file, if necessary using the @option{-x} option to make the driver
14979 treat it as a C or C++ header file. You will probably want to use a
14980 tool like @command{make} to keep the precompiled header up-to-date when
14981 the headers it contains change.
14983 A precompiled header file will be searched for when @code{#include} is
14984 seen in the compilation. As it searches for the included file
14985 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
14986 compiler looks for a precompiled header in each directory just before it
14987 looks for the include file in that directory. The name searched for is
14988 the name specified in the @code{#include} with @samp{.gch} appended. If
14989 the precompiled header file can't be used, it is ignored.
14991 For instance, if you have @code{#include "all.h"}, and you have
14992 @file{all.h.gch} in the same directory as @file{all.h}, then the
14993 precompiled header file will be used if possible, and the original
14994 header will be used otherwise.
14996 Alternatively, you might decide to put the precompiled header file in a
14997 directory and use @option{-I} to ensure that directory is searched
14998 before (or instead of) the directory containing the original header.
14999 Then, if you want to check that the precompiled header file is always
15000 used, you can put a file of the same name as the original header in this
15001 directory containing an @code{#error} command.
15003 This also works with @option{-include}. So yet another way to use
15004 precompiled headers, good for projects not designed with precompiled
15005 header files in mind, is to simply take most of the header files used by
15006 a project, include them from another header file, precompile that header
15007 file, and @option{-include} the precompiled header. If the header files
15008 have guards against multiple inclusion, they will be skipped because
15009 they've already been included (in the precompiled header).
15011 If you need to precompile the same header file for different
15012 languages, targets, or compiler options, you can instead make a
15013 @emph{directory} named like @file{all.h.gch}, and put each precompiled
15014 header in the directory, perhaps using @option{-o}. It doesn't matter
15015 what you call the files in the directory, every precompiled header in
15016 the directory will be considered. The first precompiled header
15017 encountered in the directory that is valid for this compilation will
15018 be used; they're searched in no particular order.
15020 There are many other possibilities, limited only by your imagination,
15021 good sense, and the constraints of your build system.
15023 A precompiled header file can be used only when these conditions apply:
15027 Only one precompiled header can be used in a particular compilation.
15030 A precompiled header can't be used once the first C token is seen. You
15031 can have preprocessor directives before a precompiled header; you can
15032 even include a precompiled header from inside another header, so long as
15033 there are no C tokens before the @code{#include}.
15036 The precompiled header file must be produced for the same language as
15037 the current compilation. You can't use a C precompiled header for a C++
15041 The precompiled header file must have been produced by the same compiler
15042 binary as the current compilation is using.
15045 Any macros defined before the precompiled header is included must
15046 either be defined in the same way as when the precompiled header was
15047 generated, or must not affect the precompiled header, which usually
15048 means that they don't appear in the precompiled header at all.
15050 The @option{-D} option is one way to define a macro before a
15051 precompiled header is included; using a @code{#define} can also do it.
15052 There are also some options that define macros implicitly, like
15053 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
15056 @item If debugging information is output when using the precompiled
15057 header, using @option{-g} or similar, the same kind of debugging information
15058 must have been output when building the precompiled header. However,
15059 a precompiled header built using @option{-g} can be used in a compilation
15060 when no debugging information is being output.
15062 @item The same @option{-m} options must generally be used when building
15063 and using the precompiled header. @xref{Submodel Options},
15064 for any cases where this rule is relaxed.
15066 @item Each of the following options must be the same when building and using
15067 the precompiled header:
15069 @gccoptlist{-fexceptions -funit-at-a-time}
15072 Some other command-line options starting with @option{-f},
15073 @option{-p}, or @option{-O} must be defined in the same way as when
15074 the precompiled header was generated. At present, it's not clear
15075 which options are safe to change and which are not; the safest choice
15076 is to use exactly the same options when generating and using the
15077 precompiled header. The following are known to be safe:
15079 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
15080 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
15081 -fsched-verbose=<number> -fschedule-insns -fvisibility= @gol
15086 For all of these except the last, the compiler will automatically
15087 ignore the precompiled header if the conditions aren't met. If you
15088 find an option combination that doesn't work and doesn't cause the
15089 precompiled header to be ignored, please consider filing a bug report,
15092 If you do use differing options when generating and using the
15093 precompiled header, the actual behavior will be a mixture of the
15094 behavior for the options. For instance, if you use @option{-g} to
15095 generate the precompiled header but not when using it, you may or may
15096 not get debugging information for routines in the precompiled header.
15098 @node Running Protoize
15099 @section Running Protoize
15101 The program @code{protoize} is an optional part of GCC@. You can use
15102 it to add prototypes to a program, thus converting the program to ISO
15103 C in one respect. The companion program @code{unprotoize} does the
15104 reverse: it removes argument types from any prototypes that are found.
15106 When you run these programs, you must specify a set of source files as
15107 command line arguments. The conversion programs start out by compiling
15108 these files to see what functions they define. The information gathered
15109 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
15111 After scanning comes actual conversion. The specified files are all
15112 eligible to be converted; any files they include (whether sources or
15113 just headers) are eligible as well.
15115 But not all the eligible files are converted. By default,
15116 @code{protoize} and @code{unprotoize} convert only source and header
15117 files in the current directory. You can specify additional directories
15118 whose files should be converted with the @option{-d @var{directory}}
15119 option. You can also specify particular files to exclude with the
15120 @option{-x @var{file}} option. A file is converted if it is eligible, its
15121 directory name matches one of the specified directory names, and its
15122 name within the directory has not been excluded.
15124 Basic conversion with @code{protoize} consists of rewriting most
15125 function definitions and function declarations to specify the types of
15126 the arguments. The only ones not rewritten are those for varargs
15129 @code{protoize} optionally inserts prototype declarations at the
15130 beginning of the source file, to make them available for any calls that
15131 precede the function's definition. Or it can insert prototype
15132 declarations with block scope in the blocks where undeclared functions
15135 Basic conversion with @code{unprotoize} consists of rewriting most
15136 function declarations to remove any argument types, and rewriting
15137 function definitions to the old-style pre-ISO form.
15139 Both conversion programs print a warning for any function declaration or
15140 definition that they can't convert. You can suppress these warnings
15143 The output from @code{protoize} or @code{unprotoize} replaces the
15144 original source file. The original file is renamed to a name ending
15145 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
15146 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
15147 for DOS) file already exists, then the source file is simply discarded.
15149 @code{protoize} and @code{unprotoize} both depend on GCC itself to
15150 scan the program and collect information about the functions it uses.
15151 So neither of these programs will work until GCC is installed.
15153 Here is a table of the options you can use with @code{protoize} and
15154 @code{unprotoize}. Each option works with both programs unless
15158 @item -B @var{directory}
15159 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
15160 usual directory (normally @file{/usr/local/lib}). This file contains
15161 prototype information about standard system functions. This option
15162 applies only to @code{protoize}.
15164 @item -c @var{compilation-options}
15165 Use @var{compilation-options} as the options when running @command{gcc} to
15166 produce the @samp{.X} files. The special option @option{-aux-info} is
15167 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
15169 Note that the compilation options must be given as a single argument to
15170 @code{protoize} or @code{unprotoize}. If you want to specify several
15171 @command{gcc} options, you must quote the entire set of compilation options
15172 to make them a single word in the shell.
15174 There are certain @command{gcc} arguments that you cannot use, because they
15175 would produce the wrong kind of output. These include @option{-g},
15176 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
15177 the @var{compilation-options}, they are ignored.
15180 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
15181 systems) instead of @samp{.c}. This is convenient if you are converting
15182 a C program to C++. This option applies only to @code{protoize}.
15185 Add explicit global declarations. This means inserting explicit
15186 declarations at the beginning of each source file for each function
15187 that is called in the file and was not declared. These declarations
15188 precede the first function definition that contains a call to an
15189 undeclared function. This option applies only to @code{protoize}.
15191 @item -i @var{string}
15192 Indent old-style parameter declarations with the string @var{string}.
15193 This option applies only to @code{protoize}.
15195 @code{unprotoize} converts prototyped function definitions to old-style
15196 function definitions, where the arguments are declared between the
15197 argument list and the initial @samp{@{}. By default, @code{unprotoize}
15198 uses five spaces as the indentation. If you want to indent with just
15199 one space instead, use @option{-i " "}.
15202 Keep the @samp{.X} files. Normally, they are deleted after conversion
15206 Add explicit local declarations. @code{protoize} with @option{-l} inserts
15207 a prototype declaration for each function in each block which calls the
15208 function without any declaration. This option applies only to
15212 Make no real changes. This mode just prints information about the conversions
15213 that would have been done without @option{-n}.
15216 Make no @samp{.save} files. The original files are simply deleted.
15217 Use this option with caution.
15219 @item -p @var{program}
15220 Use the program @var{program} as the compiler. Normally, the name
15221 @file{gcc} is used.
15224 Work quietly. Most warnings are suppressed.
15227 Print the version number, just like @option{-v} for @command{gcc}.
15230 If you need special compiler options to compile one of your program's
15231 source files, then you should generate that file's @samp{.X} file
15232 specially, by running @command{gcc} on that source file with the
15233 appropriate options and the option @option{-aux-info}. Then run
15234 @code{protoize} on the entire set of files. @code{protoize} will use
15235 the existing @samp{.X} file because it is newer than the source file.
15239 gcc -Dfoo=bar file1.c -aux-info file1.X
15244 You need to include the special files along with the rest in the
15245 @code{protoize} command, even though their @samp{.X} files already
15246 exist, because otherwise they won't get converted.
15248 @xref{Protoize Caveats}, for more information on how to use
15249 @code{protoize} successfully.